Claims:1. A processor implemented method for optimization and predictive maintenance of bioreactor systems, comprising:
receiving, via one or more hardware processors of a digital twin bioreactor system, a plurality of input data from one or more sources, wherein the plurality of input data corresponds to a bioreactor system (202);
pre-processing, via the one or more hardware processors, the plurality of input data and classifying the plurality of pre-processed input data into a plurality of set of variables (204);
forecasting, via one or more corresponding models executed by the one or more hardware processors, one or more expected profiles for each variable comprised in a first set of variables from the plurality of set of variables (206);
fetching, via the one or more hardware processors, one or more relevant actuation profiles of each variable comprised in a second set of variables from the plurality of set of variables (208);
predicting, via the one or more hardware processors, one or more expected profiles of each variable comprised in a third set of variables from the plurality set of variables, wherein the one or more relevant actuation profiles and the one or more expected profiles are fetched and predicted respectively by using one or more corresponding models (210);
monitoring, via the one or more hardware processors, one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or more forecasted expected profiles and the one or more predicted expected profiles respectively (212);
re-estimating, via the one or more hardware processors, the one or more relevant actuation profiles by performing dynamic optimization on the bioreactor system using an optimization model and estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the real-time status of data associated with the first set of variables and the third set of variables (214); and
recommending, via the one or more hardware processors, the one or more re-estimated relevant actuation profiles to a bioreactor system by the digital twin bioreactor system (216).

2. The processor implemented method of claim 1, wherein the one or more sources comprise at least one of a digital twin bioreactor plant, a knowledge database, one or more databases, and a server.

3. The processor implemented method of claim 1, wherein the plurality of input data comprises real-time data and non-real-time data.

4. The processor implemented method of claim 1, wherein the plurality of set of variables comprises a set of disturbance variables, a set of process variables, a set of manipulated variables, a set of design parameters, and a set of material properties.

5. The processor implemented method of claim 1, wherein the one or more relevant actuation profiles of each variable comprised in the second set of variables are selected based on at least one of one or more design parameters, and one or more material properties associated with the bioreactor system.

6. The processor implemented method of claim 1, further comprising:
performing one or more of:
fine-tuning one or more model parameters associated with the digital twin bioreactor system;
retraining the one or more corresponding models of the digital twin bioreactor system using a dataset; or
re-building the one or more corresponding models of the digital twin bioreactor system, wherein the step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values from one or more corresponding real-time values.

7. The processor implemented method of claim 1, further comprising:
performing a comparison of the trajectory of second set of variables with one or more reference trajectories stored in a knowledge database; and
modifying the optimization model based on the comparison.

8. The processor implemented method of claim 1, further comprising:
detecting, via the one or more hardware processors, one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set of variables;
performing a root cause analysis on the one or more detected faults in the bioreactor system; and
recommending one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis being performed.

9. A system (100) for optimization and predictive maintenance of bioreactor systems, comprising:
a memory (102) storing instructions;
one or more communication interfaces (106); and
one or more hardware processors (104) coupled to the memory (102) via the one or more communication interfaces (106), wherein the one or more hardware processors (104) are configured by the instructions to:
receiving a plurality of input data from one or more sources, wherein the plurality of input data corresponds to a bioreactor system;
pre-process the plurality of input data and classifying the plurality of pre-processed input data into a plurality of set of variables; forecasting, via one or more corresponding models executed by the one or more hardware processors, one or more expected profiles for each variable comprised in a first set of variables from the plurality of set of variables;
fetch one or more relevant actuation profiles of each variable comprised in a second set of variables from the plurality of set of variables;
predict one or more expected profiles of each variable comprised in a third set of variables from the plurality set of variables, wherein the one or more relevant actuation profiles and the one or more expected profiles are fetched and predicted respectively by using one or more corresponding models; monitoring, via the one or more hardware processors, one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or more forecasted expected profiles and the one or more predicted expected profiles respectively;
re-estimate the one or more relevant actuation profiles by performing dynamic optimization on the bioreactor system using an optimization model and estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the real-time status of data associated with the first set of variables and the third set of variables; and
recommend the one or more re-estimated relevant actuation profiles to a bioreactor system by a digital twin bioreactor system.

10. The system of claim 9, wherein the one or more sources comprise at least one of a digital twin bioreactor plant, a knowledge database, one or more databases, and a server.

11. The system of claim 9, wherein the plurality of input data comprises real-time data and non-real-time data.

12. The system of claim 9, wherein the plurality of set of variables comprises a set of disturbance variables, a set of process variables, a set of manipulated variables, a set of design parameters, and a set of material properties.

13. The system of claim 9, wherein the one or more relevant actuation profiles of each variable comprised in the second set of variables are selected based on at least one of one or more design parameters, and one or more material properties associated with the bioreactor system.

14. The system of claim 9, wherein the one or more hardware processors are further configured by the instructions to:
perform one or more of:
fine-tuning one or more model parameters associated with the digital twin bioreactor system;
retraining the one or more corresponding models of the digital twin bioreactor system using a dataset; or
re-building the one or more corresponding models of the digital twin bioreactor system, wherein the step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values from their real-time values.

15. The system of claim 9, wherein the one or more hardware processors are further configured by the instructions to:
perform a comparison of the trajectory of second set of variables with one or more reference trajectories stored in a knowledge database; and
modify the optimization model based on the comparison.

16. The system of claim 9, wherein the one or more hardware processors are further configured by the instructions to:
detect one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set of variables;
perform a root cause analysis on the one or more detected faults in the bioreactor system; and
recommend one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis being performed. , Description:FORM 2

THE PATENTS ACT, 1970 (39 of 1970)
&
THE PATENT RULES, 2003

COMPLETE SPECIFICATION
(See Section 10 and Rule 13)

Title of invention:
SYSTEM AND METHOD FOR OPTIMIZATION AND PREDICTIVE MAINTENANCE OF BIOREACTOR SYSTEMS

Applicant:
Tata Consultancy Services Limited
A company Incorporated in India under the Companies Act, 1956 Having address:
Nirmal Building, 9th Floor, Nariman Point, Mumbai 400021, Maharashtra, India

The following specification particularly describes the invention and the manner in which it is to be performed.

TECHNICAL FIELD
[01] The disclosure herein generally relates to bioreactor systems, and, more particularly, to system and method for optimization and predictive maintenance of bioreactor systems.
5
BACKGROUND
[02] Most of the available optimization and control techniques are applicable to bioreactor, but the complex biological system and non-linear process dynamics behavior makes optimization a tedious task. Human intervention and operational
10 experience are often required to achieve a high-quality product. The unavailability of hard sensors to measure some process parameters and delay in arrival of offline measurement values makes the system control more rigid.

SUMMARY
15 [003] Embodiments of the present disclosure present technological improvements as solutions to one or more of the above-mentioned technical problems recognized by the inventors in conventional systems. For example, in one aspect, there is provided a processor implemented method for optimization and predictive maintenance of bioreactor systems. The method comprises receiving, via one or more
20 hardware processors of a digital twin bioreactor system, a plurality of input data from one or more sources, wherein the plurality of input data corresponds to a bioreactor system; pre-processing, via the one or more hardware processors, the plurality of input data and classifying the plurality of pre-processed input data into a plurality of set of variables; forecasting, via one or more corresponding models executed by the one or
25 more hardware processors, one or more expected profiles for each variable comprised in a first set of variables from the plurality of set of variables; fetching, via the one or more hardware processors, one or more relevant actuation profiles of each variable comprised in a second set of variables from the plurality of set of variables; predicting,

via the one or more hardware processors, one or more expected profiles of each variable comprised in a third set of variables from the plurality set of variables, wherein the one or more relevant actuation profiles and the one or more expected profiles are fetched and predicted respectively by using one or more corresponding models; monitoring,
5 via the one or more hardware processors, one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or more forecasted expected profiles and the one or more predicted expected profiles respectively; re-estimating, via the one or more hardware processors, the one or more relevant actuation profiles by performing dynamic
10 optimization on the bioreactor system using an optimization model and estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the real-time status of data associated with the first set of variables and the third set of variables; and recommending, via the one or more hardware processors, the one or more re-estimated relevant actuation profiles to a
15 bioreactor system by the digital twin bioreactor system.
[04] In an embodiment, the one or more sources comprise at least one of a digital twin bioreactor plant, a knowledge database, one or more databases, and a server.
[05] In an embodiment, the plurality of input data comprises real-time data
20 and non-real-time data.
[06] In an embodiment, the plurality of set of variables comprises a set of disturbance variables, a set of process variables, a set of manipulated variables, a set of design parameters, and a set of material properties.
[07] In an embodiment, the one or more relevant actuation profiles of each
25 variable comprised in the second set of variables are selected based on at least one of one or more design parameters, and one or more material properties associated with the bioreactor system.

[08] In an embodiment, the method further comprises performing one or more of: fine-tuning one or more model parameters associated with the digital twin bioreactor system; retraining the one or more corresponding models of the digital twin bioreactor system using a dataset; or re-building the one or more corresponding models
5 of the digital twin bioreactor system, wherein the step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values from their real-time values.
[09] In an embodiment, the method further comprises performing a comparison of the trajectory of second set of variables with one or more reference
10 trajectories stored in a knowledge database; and modifying the optimization model based on the comparison.
[10] In an embodiment, the method further comprises detecting, via the one or more hardware processors, one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set of variables;
15 performing a root cause analysis on the one or more detected faults in the bioreactor system; and recommending one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis being performed.
[11] In another aspect, there is provided a system for optimization and
20 predictive maintenance of bioreactor systems, comprising: a memory storing instructions; one or more communication interfaces; and one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to: receive a plurality of input data from one or more sources, wherein the plurality of
25 input data corresponds to a bioreactor system; pre-process the plurality of input data and classifying the plurality of pre-processed input data into a plurality of set of variables; forecasting, via one or more corresponding models executed by the one or more hardware processors, one or more expected profiles for each variable comprised

in a first set of variables from the plurality of set of variables; fetch one or more relevant actuation profiles of each variable comprised in a second set of variables from the plurality of set of variables; predict one or more expected profiles of each variable comprised in a third set of variables from the plurality set of variables, wherein the one
5 or more relevant actuation profiles and the one or more expected profiles are fetched and predicted respectively by using one or more corresponding models; monitoring, via the one or more hardware processors, one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or more forecasted expected profiles and the one or more
10 predicted expected profiles respectively; re-estimate the one or more relevant actuation profiles by performing dynamic optimization on the bioreactor system using an optimization model and estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the real-time status of data associated with the first set of variables and the third set of variables; and
15 recommend the one or more re-estimated relevant actuation profiles to a bioreactor system by a digital twin bioreactor system.
[12] In an embodiment, the one or more sources comprise at least one of a digital twin bioreactor plant, a knowledge database, one or more databases, and a server.
20 [013] In an embodiment, the plurality of input data comprises real-time data and non-real-time data.
[014] In an embodiment, the plurality of set of variables comprises a set of disturbance variables, a set of process variables, a set of manipulated variables, a set of design parameters, and a set of material properties.
25 [015] In an embodiment, the one or more relevant actuation profiles of each variable comprised in the second set of variables are selected based on at least one of one or more design parameters, and one or more material properties associated with the bioreactor system.

[16] In an embodiment, the one or more hardware processors are further configured by the instructions to: perform one or more of: fine-tuning one or more model parameters associated with the digital twin bioreactor system; retraining the one or more corresponding models of the digital twin bioreactor system using a dataset; or
5 re-building the one or more corresponding models of the digital twin bioreactor system, wherein the step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values from their real-time values.
[17] In an embodiment, the one or more hardware processors are further configured by the instructions to: perform a comparison of the trajectory of second set
10 of variables with one or more reference trajectories stored in a knowledge database; and modify the optimization model based on the comparison.
[18] In an embodiment, the one or more hardware processors are further configured by the instructions to: detect one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set
15 of variables; perform a root cause analysis on the one or more detected faults in the bioreactor system; and recommend one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis being performed.
[19] In yet another aspect, there are provided one or more non-transitory
20 machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause a method for selective path sensitive interval analysis by receiving, via the one or more hardware processors of a digital twin bioreactor system, a plurality of input data from one or more sources, wherein the plurality of input data corresponds to a bioreactor system; pre-processing,
25 via the one or more hardware processors, the plurality of input data and classifying the plurality of pre-processed input data into a plurality of set of variables; forecasting, via one or more corresponding models executed by the one or more hardware processors, one or more expected profiles for each variable comprised in a first set of variables

from the plurality of set of variables; fetching, via the one or more hardware processors, one or more relevant actuation profiles of each variable comprised in a second set of variables from the plurality of set of variables; predicting, via the one or more hardware processors, one or more expected profiles of each variable comprised in a third set of
5 variables from the plurality set of variables, wherein the one or more relevant actuation profiles and the one or more expected profiles are fetched and predicted respectively by using one or more corresponding models; monitoring, via the one or more hardware processors, one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or
10 more forecasted expected profiles and the one or more predicted expected profiles respectively; re-estimating, via the one or more hardware processors, the one or more relevant actuation profiles by performing dynamic optimization on the bioreactor system using an optimization model and estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the
15 real-time status of data associated with the first set of variables and the third set of variables; and recommending, via the one or more hardware processors, the one or more re-estimated relevant actuation profiles to a bioreactor system by the digital twin bioreactor system.
[20] In an embodiment, the one or more sources comprise at least one of a
20 digital twin bioreactor plant, a knowledge database, one or more databases, and a server.
[21] In an embodiment, the plurality of input data comprises real-time data and non-real-time data.
[22] In an embodiment, the plurality of set of variables comprises a set of
25 disturbance variables, a set of process variables, a set of manipulated variables, a set of design parameters, and a set of material properties.
[23] In an embodiment, the one or more relevant actuation profiles of each variable comprised in the second set of variables are selected based on at least one of

one or more design parameters, and one or more material properties associated with the bioreactor system.
[24] In an embodiment, the method further comprises performing one or more of: fine-tuning one or more model parameters associated with the digital twin
5 bioreactor system; retraining the one or more corresponding models of the digital twin bioreactor system using a dataset; or re-building the one or more corresponding models of the digital twin bioreactor system, wherein the step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values from their real-time values.
10 [025] In an embodiment, the method further comprises performing a comparison of the trajectory of second set of variables with one or more reference trajectories stored in a knowledge database; and modifying the optimization model based on the comparison.
[26] In an embodiment, the method further comprises detecting, via the one
15 or more hardware processors, one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set of variables; performing a root cause analysis on the one or more detected faults in the bioreactor system; and recommending one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis
20 being performed.
[27] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

25 BRIEF DESCRIPTION OF THE DRAWINGS
[28] The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and, together with the description, serve to explain the disclosed principles:

[29] FIG. 1 depicts an exemplary system for optimization and predictive maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure.
[30] FIG. 2 depicts an exemplary high level block diagram of the digital twin
5 bioreactor system for optimization and predictive maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure.
[31] FIG. 3 depicts an exemplary detailed block diagram of the digital twin bioreactor system for optimization and predictive maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure.
10 [032] FIGS. 4A-4B depict an exemplary flow chart illustrating a method for optimization and predictive maintenance of bioreactor systems, using the system of FIG. 1, in accordance with an embodiment of the present disclosure.
[33] FIG. 5 depicts an exemplary digital twin bioreactor system for a biopharmaceutical reactor with models incorporated to capture phenomena in the
15 biopharmaceutical reactor, in accordance with an embodiment of the present disclosure.
[34] FIG. 6 depicts an exemplary block diagram of a dynamic optimization method of the bioreactor system, in accordance with an embodiment of the present disclosure.
20 [035] FIG. 7 depicts various components and physical sensors in the bioreactor system, in accordance with an embodiment of the present disclosure.
[36] FIG. 8 depicts velocity and volume fraction of gas phase in the bioreactor system, in accordance with an embodiment of the present disclosure.
[37] FIG. 9A depicts a graphical representation illustrating a comparison of
25 predicted axial velocity with reference values of the bioreactor system, in accordance with an embodiment of the present disclosure.

[38] FIG. 9B depicts a graphical representation illustrating a comparison of predicted radial velocity with reference values of the bioreactor system, in accordance with an embodiment of the present disclosure.
[39] FIG. 10A depicts a glucose concentration profile predicted by the digital
5 twin bioreactor system of FIG. 1, in accordance with an embodiment of the present disclosure.
[40] FIG. 10B depicts a glutamine concentration profile predicted by the digital twin bioreactor system of FIG. 1, in accordance with an embodiment of the present disclosure.
10 [041] FIG. 10C depicts a lactate concentration profile predicted by the digital twin bioreactor system of FIG. 1, in accordance with an embodiment of the present disclosure.
[42] FIG. 10D depicts an ammonia concentration profile predicted by the digital twin bioreactor system of FIG. 1, in accordance with an embodiment of the
15 present disclosure.
[43] FIG. 10E depicts a viable cell density profile predicted by the digital twin bioreactor system of FIG. 1, in accordance with an embodiment of the present disclosure.

20 DETAILED DESCRIPTION OF EMBODIMENTS
[44] Exemplary embodiments are described with reference to the accompanying drawings. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. Wherever convenient, the same reference numbers are used throughout the drawings to refer to the same or
25 like parts. While examples and features of disclosed principles are described herein, modifications, adaptations, and other implementations are possible without departing from the scope of the disclosed embodiments.

[45] Embodiments of the present disclosure provide systems and methods for optimization and predictive maintenance of bioreactor systems which are used to produce therapeutic proteins such as monoclonal antibody (mAb), antibodies for diagnostics, viral vaccines, stem cells for regenerative therapy, and the like. The
5 bioreactor system mentioned here are capable of handling cell lines such as, but are not limited to, Chinese Hamster Ovary (CHO) cells, Baby Hamster Kidney (BHK) cells, mouse myeloma cells, and the like. In an embodiment of the present disclosure, the expression ‘bioreactor system’ may also be referred as ‘bioreactor’, ‘reactor’, and interchangeably used herein.
10 [046] Based on the flow rate in and out of the bioreactor systems the mode of operation is classified as batch, fed batch, and continuous operation. The selection of operation mode is based on type of cell line used, targeted amount of final product, and the like. The bioreactor system mentioned in the present disclosure is operated in batch or fed-batch mode. In a batch process, the cell lines are inoculated to a fixed volume
15 of medium in a bioreactor vessel. The process lasts until the nutrients are completely consumed by the cell line and the broth is removed at the end of the run. Fed-batch process is a modified version of batch process, the nutrients (also referred as substrates and interchangeably used herein) are added to the vessel in increments throughout the process duration. The nutrient feeding rate is determined by the substrate limitation and
20 demand by the cell line.
[47] Cells isolated from animal tissues (also referred to as mammalian cells and interchangeably used herein) are used as host cells for production of various therapeutic proteins like mAb and virus vaccines. The bioreactor systems should provide a biologically active and controlled environment for healthy culture of
25 mammalian cells and there by obtaining a high rate of protein production. The procedure of maintaining an adequate environment for cell growth include:
1. Identifying and monitoring key parameters such as temperature, pressure, pH, dissolved oxygen, off gas carbon dioxide, off gas oxygen, viable cell density,

nutrient addition, secondary metabolite concentration, protein produced, and the like.
2. Controlling manipulated variables such as nutrient addition, aeration rate, agitation rate, acid addition, base addition, coolant flow rate, off gas release
5 rate, and the like.
3. Preventing contamination and maintaining purity of the medium.
[48] The mammalian cells are extremely sensitive to local gradients and any physical or chemical stimuli can affect the performance of a bioreactor system and the amount of protein produced. In general, mammalian cells exhibit an optimal behavior
10 at a physiological temperature of 37 degree Celsius and hence even a small change in temperature can lead to variation(s) in metabolism rate of the cells. The temperature inside the bioreactor system is typically measured using a temperature sensor and the same is maintained at desired set point by adjusting the flow rate and temperature of a coolant that is circulated around the bioreactor system. For successful completion of a
15 batch or fed batch process cycle, optimal control of pH is important. For a mammalian cell cultivation, the pH is maintained in the range of 7.0 to 7.4. In the absence of efficient control system, pH shifts to highly acidic range. Typically, the pH values are measured using probes and are controlled by sparging CO2 or by adding acidic/basic solution to the bioreactor system. In cell culture, oxygen is a key substrate required for
20 cell metabolism and is available in the form of dissolved oxygen. The level of dissolved oxygen is monitored by a sensor and air is sparged into the bioreactor system in an estimated flow rate to obtain the desired value of dissolved oxygen. If the dissolved oxygen level inside the reactor is beyond the desired value, nitrogen is sparged into the bioreactor system to strip out excess oxygen. The agitation system used in the
25 bioreactor system typically consists of impellers, a drive mechanism, and a motor. The agitation system generates efficient mixing in the bioreactor system to maintain a homogenous distribution of temperature, pressure, dissolved oxygen, and pH.

[49] Pressure inside the bioreactor system is a major factor which affects the saturation concentration of gases like carbon dioxide dissolved in the medium, which adversely affects the cell metabolism. High pressure inside the reactor vessel is also a concern in terms of safety and contamination. The foam build-up in the head space of
5 the reactor can lead to clogging of top vents and pressure rise in the reactor. A proper foam detection system and addition of anti-foaming agents helps in controlling the excessive foaming.
[50] Optimum feeding strategy is a key factor in cell metabolism and bioreactor performance. The mammalian cell medium is fed with optimum quantity of
10 carbon source and nitrogen source along with other amino acids. In a mammalian cell culture glucose and glutamine are the common substrates. During the cell metabolism these substrates are taken up by the cells and lactate, ammonium and the like are formed as secondary metabolites. The excessive accumulation of secondary metabolites causes cell death and thereby leads to a poor protein formation rate. Accumulation of lactate
15 in mammalian cell culture is found to have inhibitory effect in cell growth and protein production. In the bioreactor system, common key performance indicators are protein titer, total cell density, and viable cell density. Total cell density gives the total number of cells in the bioreactor and viable cell density indicates the number of alive and productive cells.
20 [051] In general, for measuring the concentration of substrate and secondary metabolites, molecular spectroscopy techniques can be used. The measurement technique needs separate sterile equipment for sampling and laboratory testing and takes a relatively long time. The accuracy of the above-mentioned off-line measurement techniques depends on the data used for calibrating the measurement
25 technique. Mammalian cell metabolism shows high variability across batches, which may not be considered by offline measurement calibration techniques. Product titer is also measured using spectroscopy techniques; total cell density and viable cell density are measured either by offline cell counter or via in-line process sensors.

[52] Compared to microbial cells, mammalian cells have a weak outer membrane which make them susceptible to small fluctuation in the bioreactor conditions. Mammalian cells are classified as shear sensitive cells and hence, high agitation or aeration rates can cause cell damage. At the same time, adequate aeration
5 and mixing is required for maintaining a biologically active environment inside the reactor. The heterogenous nature of cells within a batch and unpredictability of cell metabolism during cell culture makes optimization and control more challenging. While many of the key parameters such as substrate concentration, metabolite concentration and the like are measured off-line, the time lag in receiving off-line
10 measurement values may lead to delayed decision making by the plant operators and sub-optimized bioreactor system. While mammalian cells have a relatively slow growth rate, the culture duration is high; the control and optimization method should take advantage of this long culture time.
[53] The two important factors which affect the performance of the
15 bioreactor system are (i) quality of the cell line used (ii) bioprocess control and optimization. Optimization of bioreactor systems using simple process models and manual control may lead to sub-optimal performance; the complexity of the bioreactor system demands a comprehensive tool like a digital twin bioreactor system, which can be made available in real time, working in tandem with the actual bioreactor. Most of
20 the current optimization and control techniques used in biopharmaceutical industries operate on the principle of static optimization which do not update values of manipulated variables with time or as the need may be. Because of this, it may not be possible for constant monitoring of process performance and for taking immediate corrective actions. This affects important key performance indicators (KPIs) such as
25 product titer, viable cell density, total cell density, and the like. Most of the models for bioreactors used in industries run on approximate or simplified physics based or data- based models. Due to such model implementations, the complexity of the process is not captured well enough. Such approximate models are not capable of accurate

predictions. In most of the biopharmaceutical industries, physical sensors for important quality attributes are not present. Samples must be sent to a laboratory for measurements which is time consuming. Many vendors have tried to implement technologies for online/inline measurement of quality attributes, but these attempts
5 have not been fully successful and the challenges in controlling and optimization of such equipment continue to remain. Considering the importance of controlling key process parameters and monitoring key performance indicators in real time, the embodiments of present disclosure provide an optimization and predictive maintenance of bioreactor systems which address the following areas.
10 1. Monitoring, controlling and dynamic optimization of the bioreactor process in real time.
2. Prediction or forecast of key parameters as required.
3. Reduction of process variability in a batch and improvement in protein production.
15 4. Prediction of key process parameters like volumetric mass transfer co-efficient, substrate utilization rate, metabolite formation rate, viable cell density and the like.
5. Detection of any faults in the bioreactor system, identification of root causes for the faults and recommending rectification strategies for the faults.
20 [054] More specifically, the present disclosure is related to development of a digital twin for optimization and control of the bioreactor system, involving a system of integrated hardware and software components. The digital twin developed in the present disclosure considers various phenomena such as, but are not limited to, gas- hydrodynamics, interphase-mass transfer, cell metabolism, cell variability, product
25 formation and quality, and the like. The system of the present disclosure is implemented using real time and non-real time data. The system further includes a model repository with various data-based and physics-based models for predictions and optimization of bioreactor system. An appropriate model can be selected as per

requirement. Furthermore, a knowledge-database is comprised in the system that stores profiles for various manipulated variables and fault rectification strategies, wherein appropriate manipulated variables are selected based on current state of bioreactor conditions. In a further aspect, expected profiles of disturbance variables are predicted.
5 Such expected profiles include, but are not limited to, ambient temperature and the like. Similarly, the system of the present disclosure also predicts relevant actuation profiles for various manipulated variables or fetches the same from a knowledgebase. These are monitored over time for dynamic control and optimization of the bioreactor system. In other words, the bioreactor system’s performance is monitored in real-time (or near-
10 real-time) and the same is communicated to various systems for any re-estimation of manipulated variables if needed.
[55] The digital twin as described by the present disclosure also includes multiple physics-based models for predicting key parameters such as volumetric mass transfer co-efficient shear stress, substrate utilization rate, metabolite formation rate,
15 viable cell density, pH and amount of protein produced.
[56] The output from one model predicted value is used as input to other. For instance, the shear stress predicted by gas-hydrodynamics model is an input for cell population balance model. Likewise, the predicted Kla value is a measure of the mixing efficiency and an input to gas component solubility model. The prediction approach as
20 described herein by the present disclosure for substrate concentration, secondary metabolites, viable cell density and biomass concentration prediction is an effective alternative for time consuming off-line analysis.
[57] Referring now to the drawings, and more particularly to FIGS. 1 through 10E, where similar reference characters denote corresponding features
25 consistently throughout the figures, there are shown preferred embodiments and these embodiments are described in the context of the following exemplary system and/or method.

[58] FIG. 1 depicts an exemplary system 100 for optimization and predictive maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure. In an embodiment, the system 100 may also be referred as a bioreactor digital twin, a digital twin, or digital twin bioreactor system, and interchangeably used
5 herein. In an embodiment, the system 100 includes one or more hardware processors 104, communication interface device(s) or input/output (I/O) interface(s) 106 (also referred as interface(s)), and one or more data storage devices or memory 102 operatively coupled to the one or more hardware processors 104. The one or more processors 104 may be one or more software processing components and/or hardware
10 processors. In an embodiment, the hardware processors can be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and/or any devices that manipulate signals based on operational instructions. Among other capabilities, the processor(s) is/are configured to fetch and execute computer-readable instructions
15 stored in the memory. In an embodiment, the system 100 can be implemented in a variety of computing systems, such as laptop computers, notebooks, hand-held devices, workstations, mainframe computers, servers, a network cloud, and the like.
[59] The I/O interface device(s) 106 can include a variety of software and hardware interfaces, for example, a web interface, a graphical user interface, and the
20 like and can facilitate multiple communications within a wide variety of networks N/W and protocol types, including wired networks, for example, LAN, cable, etc., and wireless networks, such as WLAN, cellular, or satellite. In an embodiment, the I/O interface device(s) can include one or more ports for connecting a number of devices to one another or to another server.
25 [060] The memory 102 may include any computer-readable medium known in the art including, for example, volatile memory, such as static random-access memory (SRAM) and dynamic-random access memory (DRAM), and/or non-volatile memory, such as read only memory (ROM), erasable programmable ROM, flash

memories, hard disks, optical disks, and magnetic tapes. In an embodiment, a database 108 is comprised in the memory 102, wherein the database 108 comprises details on input received, information on input data being pre-processing, predicted values of various variables, various profiles associated with the variables, wherein the profiles
5 comprise expected profiles, actuation profiles, and the like. The variables comprise one or more of disturbance variables, process variables, manipulated variables, design parameters, material properties, and the like.
[61] The information stored in the database 108 further comprises details on user defined constraints and/or objectives. The database 108 further comprises details
10 of variables, profiles, constraints, and objectives being monitored for change, re- estimated values of various variables, physics-based models, data-based models, optimization model(s), and the like. Furthermore, the database 108 stores changes implemented in the bioreactor system and optimal set of variables being recommended for dynamic control and optimization of the bioreactor system. The memory 102 further
15 comprises (or may further comprise) information pertaining to input(s)/output(s) of each step performed by the systems and methods of the present disclosure. In other words, input(s) fed at each step and output(s) generated at each step are comprised in the memory 102 and can be utilized in further processing and analysis.
[62] FIG. 2, with reference to FIG. 1, depicts an exemplary high level block
20 diagram of the digital twin bioreactor system 100 for optimization and predictive maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure.
[63] FIG. 3, with reference to FIGS. 1 and 2, depicts an exemplary detailed block diagram of the digital twin bioreactor system 100 for optimization and predictive
25 maintenance of bioreactor systems, in accordance with an embodiment of the present disclosure. As can be seen from FIG. 3, the system 100 comprises a bioreactor system which needs to be dynamically controlled and optimized for better performance. The system 100 further comprises various other components, such as but are not limited to,

bioreactor plant automation systems for obtaining and providing real-time data from the bioreactor system along with one or more recommendations. The system 100 further comprises one or more bioreactor plant data sources that transmit real-time or non-real-time data to the digital twin bioreactor system 100 for optimization and
5 predictive maintenance of bioreactor systems. The system 100 further includes one or more repositories (e.g., the database 108 of FIG. 1) such as knowledge database, one or more databases, model repository, and the like to store various information associated with the operation and processes of the bioreactor system (and/or digital twin bioreactor system). The system 100 is further configured with a user interface for
10 ease of access to information for inputting and outputting various information. Such information includes input fed to, and output generated from various sub- systems/components of the system 100 (and/or the bioreactor system). In one embodiment, the system 100 includes various modules components such as a receiving module, a data preprocessing module, a soft-sensing module, a self-monitoring
15 module, a self-learning module, a self-optimization module, a user-defined simulation module, a control and optimization module for dynamic control and optimization of the bioreactor system and recommendation of optimum set points of manipulated variables associated with the bioreactor system for better and optimal performance, and a fault detection module for corrective action for faults. The soft-sensing module
20 comprises a forecasting module and a prediction module. In an embodiment of the present disclosure, the receiving module, the data preprocessing module, a soft-sensing module, a self-monitoring module, a self-learning module, a self-optimization module, a user-defined simulation module, a control and optimization module, the forecasting module and the prediction module are implemented as at least one of a logically self-
25 contained part of a software program, a self-contained hardware component, and/or, a self-contained hardware component with a logically self-contained part of a software program embedded into each of the hardware component that when executed perform the above method described herein.

[64] FIGS. 4A-4B, with reference to FIG. 1, depict an exemplary flow chart illustrating a method for optimization and predictive maintenance of bioreactor systems, using the system 100 of FIG. 1, in accordance with an embodiment of the present disclosure. In an embodiment, the system(s) 100 comprises one or more data
5 storage devices or the memory 102 operatively coupled to the one or more hardware processors 104 and is configured to store instructions for execution of steps of the method by the one or more processors 104. The steps of the method of the present disclosure will now be explained with reference to components of the system 100 of FIGS. 1-3, and the flow diagram as depicted in FIGS. 4A-4B. In an embodiment, at
10 step 202 of the present disclosure, the one or more hardware processors 104 receive, via the receiving module of the bioreactor digital twin 100 of FIG. 3, a plurality of input data (also referred as input data and interchangeably used herein) from one or more sources (also referred as source or sources and interchangeably used herein), wherein the plurality of input data corresponds to a bioreactor system (or the input data
15 may be related to one or more processes and/or systems of the bioreactor system). The one or more sources comprise at least one of a bioreactor plant(s), a knowledge database, one or more databases, a server, and the like. In an embodiment, the plurality of input data is of the bioreactor system. Bioreactor system is a system whose performance is measure by the digital twin bioreactor system. Both the Bioreactor
20 system and the digital twin bioreactor system are part of the bioreactor plant(s). In an embodiment of the present disclosure, a plant automation system (e.g., bioreactor plant automation system such as a distributed control system (DCS)) interacts with one or more systems such as, but are not limited to, a laboratory information management system, a bioreactor digital twin via an Open Platform Communications (OPC) server
25 (e.g., server as depicted in FIG. 2). During interaction, input data is collected. For instance, the plurality of input data from the one or more sources, comprising material properties of substrates, metabolites biological properties of cell lines used, real-time data from bioreactor, ambient temperature and pressure values, user input(s) like

constraints and objectives for the optimization problem, or combinations thereof. Some of the sources further include laboratory information management systems (LIMS), knowledgebase, the databases, model repository for obtaining input data.
[65] At step 204 of the present disclosure, the one or more hardware
5 processors 104 preprocess, via the preprocessing module of the bioreactor digital twin 100 of FIG. 3, the plurality of input data and classify the plurality of pre-processed input data into a plurality of set of variables. In one embodiment, the system 100 may implement known in the art preprocessing techniques for preprocessing the plurality of input data. Preprocessing of the input data comprises one or more steps such as removal
10 of outlier(s), missing pattern analysis, filling missing value using suitable techniques, if required, and the like. The input data coming into the digital twin system from various in-situ sensor, on-line and off-line analyzer are then appropriately classified into the plurality of set of variables, such as, but are not limited to, one or more disturbance variables, one or more process variables, one or more manipulated
15 variables, one or more design parameters, one or more material properties, and the like.
[66] At step 206 of the present disclosure, the one or more hardware processors 104 forecast, via one or more corresponding models executed by the one or more hardware processors, one or more expected profiles for each variable comprised in a first set of variables from the plurality of set of variables. More specifically, the
20 hardware processors 104 invoke the forecasting module of FIG. 3 for forecasting the one or more expected profiles for each variable comprised in the first set of variables. The first set of variables are one or more disturbance variables. In an embodiment, the model repository comprised in the database 108 stores the one or more corresponding models. For instance, the one or more corresponding models such as, but are not limited
25 to, physics-based model(s) and data-based model(s) which capture all the bioreactor phenomena are comprised in the model repository. The model repository further stores a gas-hydrodynamics model which uses agitation rate, aeration rate, and the like as inputs. The gas-hydrodynamics model working in tandem with the bubble size

distribution model predicts volumetric mass transfer co-efficient. The volumetric mass transfer co-efficient is a measure of mixing efficiency in the bioreactor. Other models for pH prediction, cell metabolism and cell population balance are stored in the model repository. All the above models are coupled and work together dynamically as per the
5 requirement. Data-based models are employed for predicting the profiles of disturbance variables. The data-based models are continually updated with recent data using a self-learning algorithm as known in the art.
[67] At step 208 of the present disclosure, the one or more hardware processors 104 fetch one or more relevant actuation profiles of each variable comprised
10 in a second set of variables from the plurality of set of variables. More specifically, the second set of variables are manipulated variables wherein the one or more relevant actuation profiles of each manipulated variable are fetched. In an embodiment of the present disclosure, the one or more relevant actuation profiles may also be referred as most appropriate actuation profiles and interchangeably used herein. The one or more
15 relevant actuation profiles of each variable comprised in the second set of variables/manipulated are selected based on one or more design parameters, and/or one or more material properties associated with the digital twin bioreactor system. Examples of the appropriate actuation profiles comprise, but are not limited to, agitation rate/speed (in RPM), aeration (LPM), substrate addition, acid/base addition
20 flow rate, cooling jacket water flow, exhaust gas release rate. These profiles are fetched from the knowledgebase, in one embodiment of the present disclosure.
[68] At step 210 of the present disclosure, the one or more hardware processors 104 predict one or more expected profiles of each variable comprised in a third set of variables from the plurality set of variables. More specifically, the third set
25 of variables are process variables wherein the one or more most appropriate profiles of each process variable are predicted. Examples of the one or more expected profiles, comprise, but are not limited to, pH, Dissolved oxygen (DO), Dissolved carbon dioxide (DCO2), temperature, pressure, volume, exhaust O2/CO2, metabolite concentration.

These expected profiles are predicted using one or more respective models stored in the model repository. To predict the one or more process variables, prediction module is invoked by the one or more hardware processors 104.
[69] At step 212 of the present disclosure, the one or more hardware
5 processors 104 monitor one or more changes in the real-time status of data associated with the first set of variables (e.g., the one or more disturbance variables) and the third set of variables (e.g., the one or more process variables). The step of monitoring the one or more changes in the real-time status of data associated with the first set of variables and the third set of variables based on a comparison with the one or more
10 forecasted expected profiles and the one or more predicted expected profiles respectively. The disturbance variables and the process variables are monitored in real- time or near-real-time for observing one or more changes. More specifically, at step 212, the one or more hardware processors 104 invoke the self-monitoring module of FIG. 3 that monitors and verifies/validates the forecasted set of disturbance variables
15 with measured values. The self-monitoring module further computes a difference between predicted and measured values (either real time or non-real time) values of process variables. In other words, the step of monitoring real-time status of data associated with the first set of variables and the third set of variables comprises performing a comparison of (i) the first set of variables and the third set of variables
20 being predicted in real-time and (ii) one or more corresponding measured values associated with the first set of variables and the third set of variables for determination any deviations/changes, etc. The third set of variables like bioreactor system temperature, pH, pressure, dissolved oxygen, dissolved carbon dioxide and the like are measured using physical sensors attached to the bioreactor system. The metric used for
25 deviation can be either mean square error or absolute deviation or relative deviation or percentage deviation, etc. In case the deviation of the predicted values of first set of variables and third set of variables is less than a pre-defined threshold of the given variable, the prediction and forecasting models do not require any modification.

However, in case the deviation is larger than the pre-defined threshold, the parameters of prediction and forecasting models needs to re-tune such that the prediction accuracy is improved. In an embodiment, the value of the pre-defined threshold is different for each variable in the plurality of first set and third set of variables. For example, if the
5 percentage deviation of threshold for the ambient temperature from the first set of variables is 15%, any deviation higher than or equal to 15% between forecasted value of ambient temperature and actual measured value of ambient temperature initiates self-learning. Likewise, for pH value in the third set of variables, if the percentage deviation between precited value of pH and actual value of pH exceeds the threshold
10 value of 5%, the self-learning for pH prediction model is enabled. Similarly, the self- learning is enabled for other prediction models based on deviation in their respective variables. For instance, the self-learning module receives the deviation of the first set of variables and the third set of variables being predicted in real time or non-real time with the measured value. Based on the deviation reported, the self-learning module is
15 configured to perform one or more of the following:
1. Update the models of the forecasting module, if the difference between expected profile and actual profile of first set of variables is more than the pre- defined threshold.
2. Update the models of the prediction module, if the difference between expected
20 profile and actual profile of third set of variables is more than the pre-defined threshold.
[70] At step 214 of the present disclosure, the one or more hardware processors 104 re-estimate the one or more relevant actuation profiles by performing dynamic optimization on the bioreactor system using an optimization model and
25 estimating a trajectory of the second set of variables for a specific time period based on the one or more identified changes in the real-time status of data associated with the first set of variables and the third set of variables. The trajectory of second set of variables is then compared with one or more reference trajectories stored in the

knowledge database. Based on the comparison, the optimization model is modified. In one embodiment of the present disclosure, the optimization model is modified if the difference between optimized trajectory of the second set of variables and reference trajectory comprised in the knowledge database exceeds beyond a threshold value.
5 [071] The above steps of 212 and 214 are better understood by way of following description:
[072] In an embodiment, the steps 212 through 214 are performed by the control and optimization module of FIG. 3, wherein the one or more hardware processor 104 recommend the optimized set of the second set of variables to the digital
10 twin bioreactor system. The control and optimization module of FIG. 1 performs real- time dynamic optimization of the given second set of variables (manipulated variables) while maximizing the performance of the bioreactor system. The output of the dynamic optimization is the trajectories of the manipulated variables that needs to be implemented in the given control horizon. However, to perform the real-time dynamic
15 optimization, the numerical values of disturbance variables are also required. Depending on the sensitivity of the disturbance variable, they can be either assumed to be constant during the control horizon or they can be forecasted using the forecasting models comprised in the model repository. The dynamic optimization also requires the transient model for performance of the bioreactor. To apply constraints for practicality
20 and to maintain KPIs within limits, transient models (comprised in the model repository) of constraints are also passed to the dynamic optimization model (or the control and optimization module of FIG. 3). The output from the dynamic optimization is implemented and then the state of the system is tracked for next iteration of dynamic optimization.
25 [073] In an embodiment of the present disclosure, the actuation profiles of the manipulated variables are implemented in real time with the help of one or more actuators comprised in the bioreactor system. In an embodiment, the steps 212 through 214 are performed by the control and optimization module of FIG. 1, wherein the

control and optimization module optimizes the bioreactor system performance to get maximum yield and product quality by adjusting the second set of variables in the optimization model. In an embodiment, at steps 212, the control and optimization module identifies/monitors one or more changes in the real-time status of data
5 associated with the first set of variables and third set of variables being monitored. In an embodiment, dynamic optimization is performed and the trajectory of the second set of variables (manipulated variables) for a time period of prediction horizon is estimated based on the observed changes in the first set of variables and the third set of variables in the past and by forecasting the first set of variables and predicting the
10 third set of variables for a time period of prediction horizon. However, the control and optimization module recommends trajectory of optimum values of the manipulated variables for a time period of control horizon to the plant automation system/the bioreactor system for better control and performance. All the changes are strictly in accordance with the constraints provided by the user, in one example embodiment. The
15 control and optimization module further performs real-time dynamic optimization for the next control horizon while the bioreactor system is implementing these actuation profiles.
[74] The control and optimization module performs an online control for maximizing the yield and quality of the product as described above. In the present
20 disclosure, the one or more hardware processors 104 further performs an offline trajectory optimization and save the trajectory of the second set of variables in the knowledge database. The one or more hardware processors 104 compares the assumed values of first set of variables (or the disturbance variables) in an offline trajectory optimization model with the measured values in the real-time optimization. The
25 optimization and control module try to maintain the third set of variables (or the process variables) at set-points prescribed by the offline optimization by adjusting the second set of variables (manipulated variables) and the values of second set of variables from offline optimization are identified and the one or more relevant actuation profiles are

re-estimated based on the deviation of forecasted values of first set of variables (or the disturbance variables) from the assumed values during offline optimization. The control and optimization module takes input from the user to either go with real-time dynamic optimization or go with the implementation of trajectory optimization in real-
5 time. In other words, at step 216 the one or more re-estimated relevant actuation profiles are recommended to a bioreactor system by the digital twin bioreactor system.
[75] The self-monitoring module further receives the optimized second set of variables (or the re-estimated relevant actuation profiles) from the control and optimization module, which are compared with the data of manipulated variable profile
10 available in the knowledge data base. The recommended optimized trajectory of variables includes one or more of agitation rate, aeration rate, nutrient supply rate, acid or base addition rate, anti-foam addition rate, coolant flow rate (also referred as cooling jacket water flow rate and interchangeably used herein) and Exhaust gas release rate. If the recommended re-estimated relevant actuation profiles of the second set of
15 variables are deviating from the expected profiles beyond a predefined threshold value, the self-optimization module is triggered. The threshold for deviations is different for each variable in the second set of variables. The threshold value can be expressed in terms of absolute deviation or percentage deviation. The reference trajectory for second set of variables depends on the reactor design parameters, working volume and material
20 properties. The acceptable range of manipulated variables, reference trajectories, and the like are stored in the knowledge database.
[76] For example, considering the design parameters and material properties, wherein the acceptable range of impeller speed or agitation rate in a bioreactor is limited at 350 RPM.
25 Case 1: if the control and optimization module recommends an agitation rate of 300 RPM for a period of prediction horizon.

Self-monitoring module compares the optimized trajectory of 300 RPM with the critical agitation rate of 350 RPM. Here the self-optimization is not triggered as the recommended agitation rate lies within the acceptable limit.
Case 2: if the control and optimization module recommends an agitation rate of 450
5 RPM for a period of prediction horizon.
Self-monitoring module compares the optimized trajectory of 450 RPM with the critical agitation rate of 350 RPM. Here the self-optimization is triggered as the recommended agitation rate exceeds the acceptable limit.
[77] Also, for a given working volume, design parameters and material
10 properties if the reference trajectory for nutrient feed flow is 1 liter per hour for a period of 24 Hours. and the acceptable deviation is up to 10% lower or higher from the reference trajectory. If the nutrient supply is less than 0.9 liter per hour, the bioreactor system will move to a nutrient under fed state. Likewise, if the optimized nutrient supply is higher than 1.1 liter per hour, the nutrients will accumulate in the reactor and
15 lead to further complexity. Therefore, the monitoring module will trigger self- optimization for an optimized nutrient flow rate below 0.9 liter per hour or above 1.1 liter per hour. Similarly, if the reference trajectory for aeration rate is 2 liter per minute for a period of 24 Hours and the acceptable deviation is up to 10% lower or higher from the reference trajectory. The monitoring module will trigger self-optimization for an
20 optimized aeration rate below 1.8 liter per minute or above 2.2 liter per minute. Likewise, the other optimized manipulated variables in the second set of variables are compared with their respective reference trajectories in knowledge database. Further, the self-optimization module tweaks the control and optimization models used in the control and optimization modules by performing one or more of the following.
25 1. Changing the objective function.
2. Changing the values of the constraints.
3. Changing the parameters such as tolerance or convergence criteria of optimization algorithm

4. Choosing a different optimization algorithm
[78] In the present disclosure, a user-defined offline simulation module is available online to investigate various scenarios that helps gain knowledge about the process. The module is also available to investigate the sensitivity of any constituent
5 in the soft-sensing module and the control and optimization module. The user-defined simulation module is configured to provide the response of the forecasting module and the predicting module for any set of the user defined model inputs, model parameters and the like. Similarly, the user-defined simulation module is configured to provide the response for any set of the user defined objective function(s), constraint values and
10 disturbance variables in the control and optimization module on the optimized second set of variables that are analyzed in the offline simulation. Though, the user-defined simulation module is available online, it may not be providing responses using real- time data from current state of bioreactor, but using the data provided by the user. This input data may or may not be of the current state of the bioreactor, in one example
15 embodiment.
[79] In an embodiment, based on the one or more re-estimated relevant actuation profiles being recommended, the following are performed by the self- learning module accordingly: In one instance, one or more model parameters associated with the digital twin bioreactor system are fine-tuned. For example, physics-
20 based cell metabolism model parameters such as cycle time, specific yield of protein, death rate of cells, growth rate of cells, rate of lysis and the like are fine-tuned. Likewise, fine-tuning can be done for gas hydrodynamics model, bubble size distribution, kla estimation model, and the like. The finetuning of model parameters can be performed on data-based models also, in one example embodiment. In another
25 instance, the one or more corresponding models of the digital twin bioreactor system are retrained using a new dataset. The data-based model used to predict the first set of variables are retrained with new set of experimental data (e.g., new set of manipulated variables, new set of disturbance variables or combination thereof). In yet another

instance, the one or more corresponding models of the digital twin bioreactor system are rebuilt. For example, if the model performance is not improved after fine-tuning or retraining, then the models are rebuilt. Rebuilding models include modifying gas hydrodynamics model by choosing a new model for calculation turbulence dissipation,
5 rebuilding bubble size distribution model, and the like, in one example embodiment of the present disclosure. The cell metabolism model is rebuilt when the digital twin system identifies a change in raw materials or new type of cell lines are used, in another example embodiment of the present disclosure.The step of fine-tuning, retraining, and re-building are performed based on the deviation in predicted and forecasted values
10 from one or more corresponding real-time values, in one embodiment of the present disclosure.
[80] The method of the present disclosure further comprises detecting, via the one or more hardware processors, one or more faults in the bioreactor system based on the one or more expected profiles of each variable comprised in the third set of
15 variables; performing a root cause analysis on the one or more detected faults in the bioreactor system; and recommending one or more corresponding corrective actions to the bioreactor system based on the one or more detected faults and the root cause analysis being performed. The above steps of fault detection, performing root cause analysis, and recommending corrective actions may be better understood by way of
20 following description:
[81] Any faults such as bioreactor contamination, fault in aeration system, fault in agitation system, sensor malfunction, controller malfunction, and the like may be detected by the system 100 (e.g., using a fault detection module – shown in FIG. 3). Faults are (or may be) identified based on the unforeseen changes observed in the
25 values of third set of variables. Based on the identified faults, the system recommends fault rectification strategies (or corresponding corrective actions) to the operator or the bioreactor system. The early detection of faults reduces operational complexity and

avoid loss of production batch. Fault detection and rectification action performed includes the following exemplary steps:
1. Comparing the real-time or near real time values of third set of variables values with the reference values available in the knowledge database.
5 2. Detecting the faults in the bioreactor system based on the comparison and indicating by means of an alarm.
3. Identifying the root cause of the fault.
4. Recommending rectification strategy based on the historical data available in knowledge database.
10 [082] The faults in bioreactor system can be identified by monitoring the variations in the process variables. The unanticipated changes in the values of process variables may be often due to some faults in the bioreactor system components or due to disturbances in the bioreactor system. The fault detection module differentiates the process variable deviation occurring due to faults and those occurring due to
15 disturbances by making use of the reference data available in the knowledge database. For example, the value of volumetric mass transfer co-efficient (Kla) is a measure of mixing efficiency in the bioreactor. The drop of Kla value below a predefined threshold value could indicate a potential fault in aeration or mixing system. The root cause for this can be development of a film over the sparger which alters the size of the air
20 bubbles or any defect in the stirrer system. The fault detection module then recommends the rectification strategy for the identified faults, like the film formation on the sparger can be removed by cleaning the sparger. Faults in the stirring system is identified by monitoring real-time rotor speed using an inbuilt tachometer or checking the motor system for any overheat. Overheating of motor and variation in the rotor
25 speed could be due to worn out shaft bearing or lack of lubrication. Based on the identified indicators, the system recommends lubrication of the rotor or replacement of the shaft bearing.

[083] Consider an example: Contamination inside the bioreactor leads to significant changes in protein production rate and biomass growth rate. Contamination inside a mammalian cell culture based bioreactor refers to any undesired foreign body or living species such as bacteria, fungi, Mycoplasmas, viruses, and the like. The
5 doubling time of certain bacteria are much less compared to typical mammalian cell lines and this high growth rate of contaminants finally leads to a majority for foreign population than desired cell culture. The source of contamination could be from the seed inoculum, misplacing of joint seals and opening ports, poor sterilization, airborne via atmosphere or gas supply and the like. The main indicators of contamination are
10 the variation in product formation rate, high substrate consumption rate, variation in acid production rate, density difference, increase in turbidity, change in color of the cell medium and the like. The fault detection module checks for any unexpected variations in the above said process variables. The turbidity values are measured using an inbuilt turbidity sensor. Based on the variations observed in the data associated with
15 process variables, the module detects the root cause of the fault by fetching appropriate information from the knowledge database. Poor cleaning may be a major reason for biological contamination wherein the pressure leak from the vessel or pipelines restricts the bioreactor system from reaching sterilization temperature during cleaning using steam. By a proper monitoring of pressure gauge values from the bioreactor and
20 pipelines, the digital twin system raises the alarm about pressure leakage. The exhaust gas filter is another potential source of microbes wherein the wet filter may aid the growth of microbes and thus contaminate the vessel. In such scenarios, corrective actions such as change in an exhaust gas filter is recommended are regular intervals or if any anomaly is identified.
25 [084] Apart from biological contamination, other typical sources of contamination are leakage of lubricants and leakage of coolant water. The mixing of lubricant or coolant with the cell culture medium affects the cell growth. The leakage usually happens via cracked pipes or worn-out O-rings at opening ports and joints. The

leakage of coolant from the cooling jacket reflects in the cooling jacket fluid pressure drop. The system detects this variation (via the fault detection module – not shown in FIGS.) and suggests appropriate corrective measures. Appropriate corrective measures may include inspection or replacement of O-rings at various joints after a predefined
5 number of sterilization cycles wherein the tight seal at joints prevent the leakage of lubricants to the cell culture medium.
[85] The bioreactor system is equipped with high-definition (HD) cameras to capture the inside image of bioreactor before and after cleaning cycle. The images captured by the camera are analyzed by the fault detection module to identify any
10 visible contamination or foreign body present. Appropriate image analysis techniques (comprised in the memory 102 or in the model repository) are fetched for analyzing the input image for any visible changes from the reference image in the knowledge database. Images captured are focused mainly on impellers, baffles, spargers, port or valve joints, sensor ports, and the like and is searched for any remnants from the
15 previous cell culture. If any abnormalities are found, the fault detection module raises an alarm and recommends for another cleaning cycle.
[86] Substrate flow drift is another possible fault in a stirred vessel bioreactor. The decrease in viable cell density could be due to low substrate feed rate, any fault in substrate feed pump or substrate flow meter. This could lead to a substrate
20 flow drift. If the substrate is not supplied as per the optimized trajectory by the digital twin system, the viable cell density and protein production rate decrease accordingly. Any abnormality detected in the measured feed flow rate, feed flow pump overheat, etc. are possible indication of substrate flow drift. The fault detection module recommends rectification strategies such as feed supply pipeline cleaning, replacement
25 or recalibration of flow meter and feed pump replacement.
[87] Physical sensors going haywire is another common fault in bioreactor system. Often sensor failures lead to more complex scenario because various controllers and the monitoring modules depend on physical sensors for various

operations. Any non-viable data from physical sensor is detected as sensor failure by the fault detection module. Based on the intensity of deviation observed in the physical sensor data the module recommends either a sensor re-calibration or a replacement. This is applicable to various pressure sensor, temperature sensor, level sensor, turbidity
5 sensor, pH sensor, dissolved oxygen sensor, flowmeters, and the like.
[88] All the above-mentioned faults are detected by analyzing the deviations observed in the process variables. The fault detection module identifies whether the process variable is deviant on the higher or the lower side of the normal value. Based on the observed abnormality and reference data available in the knowledge database,
10 the fault detection module performs the root causes analysis. For a single fault detected there can be multiples causes. For example, drop in Kla value can be either due to poor aeration or poor mixing or both. Similarly, contamination could be due to leakage of lubricant or growth of any other organisms. Based on the variation observed in the third set of variables, possible root causes are ranked based on historical data and reference
15 data available in knowledge database. The system suggests corrective measures for all possible root causes, with a higher weightage for top ranked root cause. Some of the faults in bioreactor can progressively increase the damage and the concern of safety also. Clogging of exhaust gas release port can cause pressure building inside the bioreactor (bioreactor or reactor system which is monitored by digital twin) and may
20 lead to explosion, for valve blockage and similar hazardous faults and for such occurring events the system raises an alarm along with the recommended corrective measures.
[89] FIG. 5, with reference to FIGS. 1 through 4B, depicts an exemplary digital twin bioreactor system for a biopharmaceutical reactor with models
25 incorporated to capture phenomena in the biopharmaceutical reactor, in accordance with an embodiment of the present disclosure. In FIG. 5, the bioreactor comprises a gas-hydrodynamics model, bubble size distribution model, buffer system model, species balance model for substrates and metabolites (also referred as species balance

model and interchangeably used herein), and the like. These models are integrated with each other to work together for optimizing the reactor’s performance such that viable cell density and product formation rate is enhanced. Using the method of FIGS. 4A- 4B, the reactor is dynamically optimized and controlled in real-time/near-real-time.
5 The real time sensor data, online and offline analysis data along with soft sensor data is analyzed in real time to make one or more changes in the operating course of reactor to improve efficiency. In an embodiment, both data-based and physics-based models are implemented and utilized for prediction of process parameters as described above. FIG. 5 can be better understood by way of following description.
10 [090] The gas-hydrodynamics model coupled with bubble size distribution model predicts the volumetric mass transfer co-efficient (Kla), shear forces, volume fraction of gas phase, velocity, and the like. These values give insight to control the manipulated variables such as agitation, gas flow rate, and the like. The buffer system model predicts the pH value wherein the impact of pH value in glucose uptake rate and
15 lactate production rate is monitored. A species balance model predicts the substrate uptake and product formation in the bioreactor system and is coupled with a cell population balance model for the biological cells. The shear forces and volume fraction of gas phase predicted by the gas-hydrodynamics model and viable cell density from the cell population balance model are some of the inputs for calculating substrate and
20 metabolite concentration. The dissolved oxygen concentration prediction model receives KLa value, viable cell density, and the like as model inputs. Total cell density, viable cell density, and death rate of cells are predicted by cell population balance model, which is coupled with gas-hydrodynamics and species balance model. The cell population balance model captures the total cell density, viable cell density and dead
25 cell density. The model considers the death of the living cells due to shear and bubble breakage also. This helps in accurately predicting the viable cell density and amount of final protein produced. Compared to many existing models for stirred vessel bioreactor, the integrated model presented here is quite comprehensive and captures most of the

phenomena in a bioreactor. In many of the existing models, the effect of gas- hydrodynamics is often ignored or simplified. The models that do not consider detailed gas-hydrodynamics as presented here underperform when the process conditions or design parameters changes. The gas-hydrodynamics model comprises a multiphase
5 model, a turbulence model, and a drag force model. The gas-hydrodynamics model used here is coupled with the bubble population balance model. The bubble population balance model estimates the bubble size distribution, and this helps in accurate prediction of volumetric mass transfer co-efficient. Usually, the gas-hydrodynamics model attains steady state in a short-time as compared to the cell metabolism model;
10 the integrated model in the present disclosure considers this time scale difference and effectively couples both the gas-hydrodynamics and cell metabolism model. The cell metabolism model coupled with the gas hydrodynamics model effectively predicts the substrate utilization rate, metabolite formation rate and product formation rate. All these models in the prediction module work in tandem with forecasting module and the
15 control and optimization module for performing the method of the present disclosure described herein.
[91] FIG. 6, with reference to FIGS. 1 through 5, depicts an exemplary block diagram of a dynamic optimization method of the bioreactor system, in accordance with an embodiment of the present disclosure. Targets of KPI like total protein
20 produced and bounds are received by the system 100 from the user as input. The forecasting and the prediction modules communicate the first and the third set of variables respectively to the control and optimization module. Based on the target of KPI like total protein produced set by the user, the bioreactor digital twin system fetches appropriate objective function and constraints from the database. The real time
25 and non-real time data from the bioreactor plant data is pre-processed and sent to the control and optimization module. Once all the required inputs are received the control and optimization module calculates the optimized second set of variables and recommends it to the bioreactor system as shown in FIG. 6.

[92] FIG. 7, with reference to FIGS. 1 through 6, depicts various components and physical sensors in the bioreactor system, in accordance with an embodiment of the present disclosure. The physical sensors are the sources of real time data sent to the bioreactor digital twin system, in one example embodiment of the present disclosure.
5 [093] FIG. 8, with reference to FIGS. 1 through 7, depicts velocity and volume fraction of gas phase in the bioreactor system, in accordance with an embodiment of the present disclosure. More specifically, FIG. 8 illustrates the distribution of secondary phase (gas) inside the reactor. The velocity and volume fraction are represented across a mid-plane inside the bioreactor. The velocity of gas phase is very
10 high near the impeller due to the no-slip boundary condition assumed. The centrifugal force generated by the impeller pushes the gas towards the walls of the vessel and aids in proper mixing. The gas-hydrodynamics part of physics-based model captures the gas velocity and volume fraction accurately and these values are used to predict volumetric mass transfer coefficient.
15 [094] FIG. 9A, with reference to FIGS. 1 through 8, depicts a graphical representation illustrating a comparison of predicted axial velocity with reference values of the bioreactor system, in accordance with an embodiment of the present disclosure. More specifically, FIG. 9A illustrates mean axial velocity of liquid at a fixed radial position. At the plane of impeller disc, the axial velocity is zero. Below the
20 plane of impeller disc, liquid is moving towards the reactor bottom and above the plane liquid is moving toward the top of the reactor. The predicted values are in good agreement with the mean experimental values.
[95] FIG. 9B, with reference to FIGS. 1 through 9A, depicts a comparison of predicted radial velocity with reference values of the bioreactor system, in
25 accordance with an embodiment of the present disclosure. More specifically, FIG. 9B illustrates the mean radial velocity of liquid at a fixed radial position. The liquid phase attains higher values of radial velocity at the plane of impeller disc and lower values

toward the top and bottom of reactor vessel respectively. The predicted values are in good agreement with the mean experimental values.
[96] FIGS. 10A through 10B, with reference to FIGS. 1 through 9B, depict species concentration profiles predicted by the digital twin bioreactor system 100 of
5 FIG. 1, in accordance with an embodiment of the present disclosure. More specifically, FIG. 10A depicts a glucose concentration profile predicted by the digital twin bioreactor system 100 of FIG. 1, in accordance with an embodiment of the present disclosure. FIG. 10B depicts a glutamine concentration profile predicted by the digital twin bioreactor system 100 of FIG. 1, in accordance with an embodiment of the present
10 disclosure. FIG. 10C depicts a lactate concentration profile predicted by the digital twin bioreactor system 100 of FIG. 1, in accordance with an embodiment of the present disclosure. FIG. 10D depicts an ammonia concentration profile predicted by the digital twin bioreactor system 100 of FIG. 1, in accordance with an embodiment of the present disclosure. FIG. 10E depicts a viable cell density profile predicted by the digital twin
15 bioreactor system 100 of FIG. 1, in accordance with an embodiment of the present disclosure. More specifically, FIG. 10E illustrates the changes in concentration of metabolites - glucose, glutamine, lactate, ammonium, and viable cell density. Based on the initial concentration of glucose, glutamine and viable cells, the physics-based species transport model predicts the changes in the values of concentration of
20 metabolites and the viable cell density. The digital twin dynamically updates the metabolite concentration using uptake rates and degradation of glucose and glutamine. The accurate prediction of metabolite concentration helps in planning the feeding strategy in a fed batch bioreactor and early detection of any metabolite (ammonia/lactate) accumulation.
25 [097] Embodiments of the present disclosure provide a comprehensive digital twin that involves development of algorithms and codes for optimization and maintenance of the bioreactor system, aiming to increase product titer quality, viable cell density, and product formation rate. Present disclosure provides systems and

methods that implement physics-based and data-based models which are integrated to work together and predict the bioreactor performance well in advance. Any deviation in plant performance is taken care of and the appropriate control recommendations are sent by the bioreactor digital twin (or digital twin bioreactor system) to the bioreactor
5 system. The model database contains models that capture all the phenomena inside a bioreactor. For example, the bubble size distribution model coupled with gas- hydrodynamics model can predict the bubble breakage and therefore, the cell death, due to bubble breakage. Real time dynamic optimization of the bioreactor process is described herein, where a new set of trajectories of the manipulated variables are
10 predicted for process optimization. The database 108 and the modules are capable of handling different process modes such as batch, fed-batch or continuous process and multiple mammalian cell-lines and appropriate model(s) for each mode or a specific cell line can be selected based on the requirement(s).
[98] The written description describes the subject matter herein to enable any
15 person skilled in the art to make and use the embodiments. The scope of the subject matter embodiments is defined by the claims and may include other modifications that occur to those skilled in the art. Such other modifications are intended to be within the scope of the claims if they have similar elements that do not differ from the literal language of the claims or if they include equivalent elements with insubstantial
20 differences from the literal language of the claims.
[99] It is to be understood that the scope of the protection is extended to such a program and in addition to a computer-readable means having a message therein; such computer-readable storage means contain program-code means for implementation of one or more steps of the method, when the program runs on a server
25 or mobile device or any suitable programmable device. The hardware device can be any kind of device which can be programmed including e.g., any kind of computer like a server or a personal computer, or the like, or any combination thereof. The device may also include means which could be e.g., hardware means like e.g., an application-

specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination of hardware and software means, e.g., an ASIC and an FPGA, or at least one microprocessor and at least one memory with software processing components located therein. Thus, the means can include both hardware means and software means.
5 The method embodiments described herein could be implemented in hardware and software. The device may also include software means. Alternatively, the embodiments may be implemented on different hardware devices, e.g., using a plurality of CPUs.
[0100] The embodiments herein can comprise hardware and software elements. The embodiments that are implemented in software include but are not limited to,
10 firmware, resident software, microcode, etc. The functions performed by various components described herein may be implemented in other components or combinations of other components. For the purposes of this description, a computer- usable or computer readable medium can be any apparatus that can comprise, store, communicate, propagate, or transport the program for use by or in connection with the
15 instruction execution system, apparatus, or device.
[101] The illustrated steps are set out to explain the exemplary embodiments shown, and it should be anticipated that ongoing technological development will change the manner in which particular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries
20 of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries can be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, etc., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings
25 contained herein. Such alternatives fall within the scope of the disclosed embodiments. Also, the words “comprising,” “having,” “containing,” and “including,” and other similar forms are intended to be equivalent in meaning and be open ended in that an item or items following any one of these words is not meant to be an exhaustive listing

of such item or items, or meant to be limited to only the listed item or items. It must also be noted that as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
[102] Furthermore, one or more computer-readable storage media may be
5 utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent
10 with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include random access memory (RAM), read-only memory (ROM), volatile memory, nonvolatile memory, hard drives, CD ROMs, DVDs, flash drives, disks, and any other known physical storage media.
15 [103] It is intended that the disclosure and examples be considered as exemplary only, with a true scope of disclosed embodiments being indicated by the following claims.