Description:Complete Specification:
The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed.

Field of the invention:
[0001] The present invention relates to a control unit and method for generation of engine maps for calibration.

Background of the invention:
[0002] In current system for two wheeler vehicles, it requires significant effort in calibration due to a large number of calibration labels owing to four wheeler legacy (about 4000 and 12000 for motorcycles from commuter segment and power sports segment respectively). Moreover, it lacks modularity and has limited possibility to customize calibration effort to individual vehicle's features and components. Manual/semiautomated calibration approaches currently in practice require a high amount of testbench/dyno time, and as operating a testbench is very expensive, this incurs a high cost per calibration project. Moreover, human errors might induce a wastage of dyno time. This bottleneck of testbench also requires the calibration engineer to be in the vicinity of the vehicle to be calibrated. This in turn hinders the possibility to localize calibration projects.

[0003] The Electronic Control Unit (ECU) is an integral part of the automotive vehicle, which actuates many of its important function. The ECU software consists of many look up tables or maps which helps performing the multiple control functions of the vehicle. For any newly developed engine these maps need to be calibrated for its optimal performance. The calibration process is conducted on a vehicle dynamometer also known as the vehicle test bench. The state of art involves performing the calibration activity mostly manually, which leads to high time and cost for the entire process. Further, all through the runtime of the project, the Original Equipment Manufacturer (OEM) may request for a component change which further leads to redoing same process multiple times. Through this activity we aim at leveraging models to reduce these costs in calibration activity.

[0004] According to state of the art 201641043689, a system and method to calibrate an Engine Control Unit of a vehicle. The system is provided to calibrate the ECU of the vehicle. The system comprises a remote computer, a central server, a local computer, and setup comprising at least a dynamo meter, and at least one actuator. The dynamo meter and the actuator are interfaced and operated with the local computer. The central server is connected to the local computer by a second networking means, and a remote computer is connected to the central server by a first networking means. The remote computer, uploads instructions to the central server, executes the instructions through the local computer to operate the dynamo meter and the actuator, and calibrates the ECU of the vehicle. The instructions are downloaded to the local computer by the second networking means.

Brief description of the accompanying drawings:
[0005] An embodiment of the disclosure is described with reference to the following accompanying drawings,
[0006] Fig. 1 illustrates a block diagram of a control unit for generation of engine maps for calibration, according to an embodiment of the present invention;
[0007] Fig. 2 illustrates a block diagram of generation of engine map, according to an embodiment of the present invention, and
[0008] Fig. 3 illustrates a method for generating engine maps for calibration, according to the present invention.

Detailed description of the embodiments:
[0009] Fig. 1 illustrates a block diagram of a control unit for generation of engine maps for calibration, according to an embodiment of the present invention. The control unit 110 configured to receive input dataset 116 of an engine map 102 (or engine labels) to be calibrated, characterized in that, the input dataset 116 comprises measured output data 118 for limited operating points of an engine of a vehicle 100. The operating point is defined by at least one input variable, mostly two input variables or more. The control unit 110 processes the operating points for the measured output data 118 through a physics based engine model 112 (or physics based engine model 112 with variable parameter) and predicts an output. The engine model is 112 processed iteratively for each of the operating points until a difference between the predicted output and the corresponding measured output is within a predetermined threshold range or is below or above a threshold value. The physics based engine model 112 is pre-stored in a memory element 108 of the control unit 110. The control unit 110 optimizes, using an optimizer 120, value of parameters (also known as model parameters) of the engine model 112 using the predicted output. The control unit 110 determines, using a Machine Learning (ML) based data model 114, value of parameters for remaining operating points of the engine with unmeasured output data using the optimized value of parameters for the operating points with measured output data 118. The control unit 110 estimates unmeasured output data using the determined values of the parameters and generate complete engine map 102 for calibration. The output data is also referred to as target variable.

[0010] In accordance to an embodiment of the present invention, the parameters of the engine model 112 are selected from a group comprising a heat transfer coefficient (h) and a pressure pulsation factor (p) and not limited to the same. The parameter may also include throttle coefficient. The parameters are specific to the engine model 112 and selected based on its effect to a desired output/target. Further, the limited operating points are selected based on sensitivity study of the parameters in different operating points/regions of the engine, and those are selected where the parameters show higher non-linearity.

[0011] In accordance to an embodiment of the present invention, the control unit 110 is part of a device. The device is at least one selected from a group comprising a cloud 104, a communication device 106 and a computing device. The device is interfaced with an Electronic Control Unit (ECU) of the vehicle 100 either through wireless or wired manner as known in the art. The ECU 124 (or controller) is at least one of an Engine Management System (EMS) controller, a Tire Pressure Monitoring System (TPMS) controller, a Telematics Control Unit (TCU) controller, Anti-lock Braking System (ABS) ECU 124, a Body Control Unit (BCU), a Human-Machine Interface (HMI) cluster unit, other vehicular controllers, and a combination thereof. The communication device 106 corresponds to electronic computing devices such as smartphone, tablets, wearable electronics such as smart watch, etc. The cloud 104 corresponds to cloud computing architecture having single or network of servers, databases connected with each other for processing of inputs and providing outputs.

[0012] According to the present invention, control unit 110 is provided with necessary signal detection, acquisition, and processing circuits. The control unit 110 is a controller which comprises input/output interfaces having pins or ports, a memory element 108 such as Random Access Memory (RAM) and/or Read Only Memory (ROM), Analog-to-Digital Converter (ADC) and a Digital-to-Analog Convertor (DAC), clocks, timers, counters and at least one processor (capable of implementing machine learning) connected with each other and to other components through communication bus channels. The memory element 108 is pre-stored with logics or instructions or programs or applications or modules/models and/or threshold values/ranges, which is/are accessed by the at least one processor as per the defined routines. The internal components of the control unit 110 are not explained for being state of the art, and the same must not be understood in a limiting manner. The control unit 110 may also comprise communication units to communicate with ECU 124 of the vehicle 100 or computing devices used for calibration, through wireless or wired means such as Global System for Mobile Communications (GSM), 3G, 4G, 5G, Wi-Fi, Bluetooth, Ethernet, serial networks, and the like. The control unit 110 is implementable in the form of System-in-Package (SiP) or System-on-Chip (SOC) or any other known types.

[0013] According to the present invention, a working of the control unit 110 is described. But before describing the working of the control unit 110, the prerequisites of the engine model 112 is explained. In a first step, the necessary parameters of the engine model 112 are identified. For the task of parameter identification, figuring out the sensitivity of the parameters is of utmost importance. A state of the art engine model 112 is selected and the most sensitive parameters with respect to the target variables/output data is selected. The method used to select the sensitive parameters is Morris method. The study using Morris method gives information on the most important parameters to be optimized for accurate results and helps tailor efforts towards the sensitive parameters while ignoring the rest. The Morris method is just an example and other methods are also usable and well within the scope of the present invention. Depending on the target variable, the parameters may change. In a second step, the selection of training points (or specific operating points) or Design of Experiments (DOE) is performed. A model-free approach is used, and the operating points are selected based on a space filling and sensitivity study of the parameters in different regions of the input space. More points are chosen in the region where the parameters tend to show higher non-linearity. A third step comprises measurement of training data, i.e., input dataset 116. The input dataset 116, using which the engine model 112 is trained, is obtained directly from the customer or Original Equipment Manufacturer (OEM), or obtained internally using a dyno 122. The input dataset 116 is used to optimize and tune the parameters of the engine model 112.

[0014] Once the above three steps are completed, the control unit 110 performs the remaining steps. The control unit 110 receives the input dataset 116 and processes through the engine model 112 which is basic and general model for an engine of the vehicle 100. The control unit 110 takes values for each available operating point as per the input dataset 116 and predicts an output. The control unit 110 refers the input dataset 116 and obtains the measured output data 118 of the engine from the memory element 108. The control unit 110 calculates the difference and revises the values of the selected parameters and predicts the output again. The control unit 110 iterates output prediction, until the difference is within acceptable threshold range or is below or above a threshold value. When the predicted output is satisfactory, the control unit 110 saves the optimized values of the parameters in the memory element 108 and repeats the step for other operating points of the input dataset 116.

[0015] As an example, the physics-based engine model 112 is written in the Low-Level Language (LLL) such as C and optimizer 120 is made using High-Level Language (HLL) such as Python. The languages mentioned for LLL and HLL is for ease of explanation and the present invention is not limited to the same. Further, the control unit 110 uses a Nelder-Mead method for optimization of the parameters at every training point or operating point of the input dataset 116, which are used in the next step to predict the parameter values at the remaining operating points with unmeasured data. Though the Nelder-Mead is a local optimization scheme but works for a non-linear system-based engine model 112 mainly because engine model 112 is well-known, and a good starting point is available with the initial values of parameters for the optimization. Again, Nelder-Mead is not the only method for optimization, but other state of the art methods is equally implementable such as basin hopping method. The specific method named above are for explanation and the present invention is not limited by the same.

[0016] Once all the operating points with measured output variable is exhausted by the control unit 110, i.e., the values of the parameters of the engine model 112 for the input dataset 116 is tuned/optimized, the control unit 110 proceeds with implementation of data model 114. The data model 114 uses interpolation or regression methodology to determine the tuned parameters for the remaining operating points with unmeasured output variables. The control unit 110 uses the optimized parameters from the previous step to predict the values at the remaining points. For example, the data model 114 uses a Gaussian Process Regression (GPR) method for prediction or estimation of values of parameters for remaining operating points with unknown output data. Once the parameter values at all the points of interest are known/estimated, the engine model 112 is said to be tuned by the control unit 110. The control unit 110 then proceeds with generation of complete engine map 102. Using the tuned model, the control unit 110 predicts the target variable/unmeasured output data at the remaining operating points (Total operating points of interest minus the training points). Basically, the control unit 110 runs the tuned engine model 112 at the appropriate input values and the output of the engine model 112 is used to predict the values of interest (i.e., unmeasured output data) at the remaining operating points.

[0017] Fig. 2 illustrates a block diagram of generation of engine map, according to an embodiment of the present invention. A first sub-figure 200 represents the engine map 102 with partial data and a second sub-figure 210 shows the engine map 102 with complete data. In the first sub-figure 200, the engine map 102 with partial data comprises input dataset 116 with limited number of measured output data 118 for selected operating points. For example, consider the engine map 102 of concern is air charge map, where there are two input variables comprising throttle position in percentage and engine speed in Rotations Per Minute (RPM). For selected combination of two input variables, the air-charge is measured and saved provided emission is within limits. The measured output data 118 is indicated by white stars 202. It is to be noted that the white stars 202 are distributed across the engine map 102 because the output data is measured for specific operating point based on the sensitivity of the parameters of the engine model 112. The control unit 110 optimizes the values of the parameters at each operating point indicated by the white stars 202 and saves the optimized/tuned values of the parameters in the memory element 108. Some examples of engine maps 102 are but not limited to throttle based air charge map, pressure based air charge map, ignition angle calibration map, exhaust temperature map, etc.

[0018] Once the parameters for the input dataset 116 (i.e., for the white stars 202) is tuned, the control unit 110 computes the value of parameters for remaining operating points using the tuned parameters. The control unit 110 uses interpolation or regression or both to arrive to predict/estimate the values of the parameter at the remaining operating points. The value of parameter for remaining operating points is represented by black stars 204 in the second sub-figure 210. Once the parameter for all the operating points of the concerned engine map 102 is available, the engine model 112 is said to be tuned and is ready to aid in generation of complete engine map 102.

[0019] In accordance to an embodiment of the present invention, different implementation of the control unit 110 is envisaged. In one form of implementation, the control unit 110 is part of the cloud 104. The input dataset 116 is provided to the cloud 104 through the communication device 106 or the computing device, through the stored/saved input datasets 116. The cloud 104 receives the input dataset 116, processes the engine model 112, optimizes the parameters, and generates the complete engine map 102. The generated engine map 102 is transmitted back to the vehicle 100 directly or to the computing device or the communication device 106, which then flashes the complete engine map 102 to the ECU 124 of the vehicle 100. In an alternate embodiment, the control unit 110 is part of the communication device 106. The input dataset 116 is processed by an application installed in the communication device 106. The application generates the complete engine map 102 after processing the engine model 112 and optimizing the parameters. The application then transmits back the complete engine map 102 either directly to the vehicle 100 or to the computing device, followed by flashing the complete engine map 102 to the ECU 124 of the vehicle 100. Alternatively, the control unit 110 is part of the computing device such as a computer or laptop, which is used located in proximity to the vehicle 100. The control unit 110 then generates the complete engine maps 102 as described above and flashes the same to the ECU 124 of the vehicle 100.

[0020] In yet another embodiment of the present invention, the control unit 110 is part of the cloud 104 and communicates with a computing device in the dyno 122. The vehicle 100 is interfaced with the computing device with all the necessary actuators having input control and sensors to measure outputs. The control unit 110 is configured to operate the dyno 122 based on automation scripts in the computing device and obtain the input datasets 116 directly. Alternatively, the input datasets 116 is obtained from other sources (such stored in external memory elements) are used. The cloud 104 receives the input dataset 116, processes the engine model 112, optimizes the parameters, and generates the complete engine map 102. The generated engine maps are flashed to the ECU 124. Alternatively, the control unit 110 is part of the computing device which receives the input dataset 116 from the cloud 104 or from remote location such as another computing device. The computing device processes the engine model 112, optimizes the parameters, and generates the complete engine map 102, and the uploads back to the cloud 104 or the remote computer for flashing to the ECU 124 of the vehicle 100. In still another alternative, the cloud 104 receives/provided with the input dataset 116, processes through the engine model 112, optimizes the parameters and generates the complete engine maps 102. The generated engine maps 102 are available for downloading by a requester/user to his/her computing device or communication device 106. Once downloaded, the engine maps 102 are flashed into the ECU 124 of the vehicle 100. The input datasets 116 is either generated by the cloud 104 using the automation scripts or manual operation. The cloud 104 based solution is provided as Software as a Service (SAAS), where users obtain complete ready-to-use calibrated engine maps 102 after uploading minimal input datasets 116.

[0021] According to an embodiment of the present invention, a software based calibration framework is provided. The calibration framework is programmed and stored in the memory element 108 of the control unit 110 and executed by the control unit 110 as per requirement. The calibration framework is configured to receive input dataset 116 of the engine map 102 to be calibrated, characterized in that, the input dataset 116 comprises measured output data 118 for limited operating points of the engine of the vehicle 100. The operating point is defined by at least one input variable, specifically two input variables or more. The control unit 110 processes the operating points for the measured output data 118 through the physics based engine model 112 and predicts an output. The engine model is 112 processed iteratively for each of the operating points until the difference between the predicted output and the corresponding measured output is within the predetermined threshold range or is below or above a threshold value. The physics based engine model 112 is pre-stored in the memory element 108 of the control unit 110. The calibration framework optimizes, using an optimizer 120, value of parameters of the engine model 112 using the measured output. The calibration framework determines, using a Machine Learning (ML) based data model 114, value of parameters for remaining operating points of the engine with unmeasured output data using the optimized value of parameters for the operating points with measured output data 118. The control unit 110 estimates unmeasured output data using the determined values of the parameters and generate complete engine map 102 for calibration.

[0022] Fig. 3 illustrates a method for generating engine maps for calibration, according to the present invention. The method, performed/executed by the control unit 110, comprising plurality of steps of which a first step 302 comprises receiving, by the control unit 110, input dataset 116 of the engine map 102. The method is characterized by the measured output data 118 is for limited operating points of the engine of the vehicle 100. The operating point is defined by at least one input variable, mostly two input variables or more. A step 304 which comprises processing, by the control unit 110, the operating points for measured output data 118 through the physics based engine model 112 and predicting the output. The physics based engine model 112 is pre-stored in the memory element 108 of the control unit 110. A step 306 comprises optimizing, using an optimizer 120 within the control unit 110, value of parameters of the engine model 112 using the predicted output. A step 308 comprises determining, by the control unit 110 using a Machine Learning (ML) based data model 114, value of parameters for remaining operating points of the engine with unmeasured output data using the optimized value of parameters for the operating points with measured output data 118. A step 310 comprises estimating unmeasured output data using the determined values of the parameters and generating complete engine map 102 for calibration.

[0023] The parameters are selected from a group comprising the heat transfer coefficient (h) and the pressure pulsation factor (p) and not limited to the same. The step 304 further comprises iteratively processing, by the control unit 110, the engine model 112 for each of the operating points until the difference between the predicted output and the measured output is within the predetermined threshold range/value. Further, the limited operating points are selected based on sensitivity study of the parameters in different operating points/regions of the engine and where the parameters tend to show higher non-linearity. The method is performed by the control unit 110 of the device. The device is at least one selected from the group comprising the cloud 104, the communication device 106 and the computing device.

[0024] The working of the present invention is already explained above. However, the method is explained in brief manner in order to avoid repetition. In this approach, the vehicle 100 is run on the dyno 122 for a given input cycle. The input cycle is specially designed to extract the maximum information of the vehicle 100 in its operating regime to build the input datasets 116. Then the collected input dataset 116 is used to tune the engine model 112 with the help of data model 114, so that the engine model 112 is able to replicate the actual behavior of the engine/vehicle 100 to a high level of accuracy. Then the tuned engine model 112 is utilized to generate the complete engine maps 102 and calibrate the ECU 124 of the vehicle 100. By utilizing the testbench/dyno 122 mainly for collecting measurements and optimizing the engine model 112 and the subsequent calibration in the device connected to the test bench/dyno 122, the calibration effort significantly eliminated. An effort reduction is reduced to the tune of at least 25% from the state of art. The control unit 110 and method improves the efficiency of calibration of the Electronic Control Units (ECUs) 124 in two-wheelers and other types of vehicles 100. The vehicle 100 is any of the motorcycle, scooter, auto-rickshaws, cars, buses, trucks, power sports, water sports and the like. By implementation of the present invention, significant amount of process time, manual effort and dynamometer usage time can be reduced. Through the present invention, the objective/aim is to leverage tuned engine models 112 to reduce these costs in calibration activity. The present invention specifically targets the base calibration and transient calibration packages for two-wheeler vehicles 100 and other vehicles 100 which forms a significant part of the entire calibration effort requiring dyno 122.

[0025] According to an embodiment of the present invention, a smart calibration framework for vehicle 100 is provided. It is well known that the costs of engine/vehicle 100 calibration at the dynamometer is staggering. The smart calibration framework is able to reduce the time and cost of the overall calibration effort by at least 25%. This coupled with the fact that every change in the engine requires re-calibration means the present invention enables saving at least 25 % of the cost for every single each calibration effort. This translate to significant cost savings. The smart calibration framework is also possible to be served/provided as a Software As A Service (SAAS) product. The present invention provides a future of model based automated smart calibration for vehicles 100, thereby minimizing the usage of hardware and human intervention. The present invention provides the framework for smart auto-calibration of the vehicle 100 on the dynamometer (or dyno 122).

[0026] In comparison to manual calibration where a calibration engineer must tune the engine/vehicle 100 at every point of the engine map 102, in the present invention, calibrated measurements at only a minimal set of operating points are required. The measurements are then used to tune the engine model 112 using the data model 114, which after tuning replicates the dynamics of interest of the remaining operating points. Next this tuned engine model 112 is used to generate the output/target values at the remaining points of the engine map 102. The technical benefit and advancement of the smart calibration lies with the fact that the physics-based engine model 112 is computationally cheap, i.e., runs faster on a computer while retaining the required accuracy as compared to other possible software in the market. The combination of machine learning based data model 114 and physics-based engine model 112 is another new feature, which results in accurate calibrated map generation from the calibration framework. Overall, the complete framework, once implemented through a control unit 110 (or computer) in the calibration setup, it enhances the speed of calibration by assisting manual calibration with the computer based smart calibration. This has significant cost savings for a company/organization. There is always a trade-off between model accuracy and computation cost, but the present invention strikes a balance between the two, all the while being capable of handling the target dynamics of the engine model 112. Moreover, the present invention has potential of scalability for other application areas as well not limiting to calibration of vehicle 100, but other areas involving complex physical systems.

[0027] According to the present invention, the control unit 110 and the method significantly reduces cost and effort of calibration. As the engine and vehicle calibration involves a high cost and effort, the concept of leveraging models for reducing cost and effort for calibration is optimal towards calibration of vehicles 100, and has a good potential to be scaled up for four-wheel vehicles 100 and electric vehicles 100. Further, the control unit 110 and method aids in case of component changes in the vehicle 100. In practice a calibration project is seldom completed at one go. Multiple requests are made by the customers for change in components during the runtime of the project. In many such cases many calibration activities need to be repeated. The model based approach benefits in effort and cost savings due to such multiple runs of the activities. The entire process of calibration is possible to be automated if the present invention is combined with existing calibration tools (such as but not limited to INCA™) and automation scripts for collecting measurements from vehicle 100. This lead to better operational efficiency. In addition, the present invention has a potential to be developed into a software as a service for OEMs who have recently taken up the responsibility of calibration themselves.

[0028] It should be understood that the embodiments explained in the description above are only illustrative and do not limit the scope of this invention. Many such embodiments and other modifications and changes in the embodiment explained in the description are envisaged. The scope of the invention is only limited by the scope of the claims.
, C , Claims:
We claim:
1. A control unit (110) for generation of engine maps (102) for calibration, said control unit (110) configured to,
receive input dataset (116) of an engine map (102) to be calibrated characterized in that, said input dataset (116) comprises measured output data (118) for limited operating points of an engine of a vehicle (100), said operating point is defined by at least one input variable,
process said operating points for said measured output data (118) through a physics based engine model (112) and predict an output, said physics based engine model (112) is pre-stored in a memory element (108) of said control unit (110);
optimize, using an optimizer (120), value of parameters of said engine model (112) using said predicted output;
determine, using a Machine Learning (ML) based data model (114), value of parameters for remaining operating points of said engine with unmeasured output data using optimized value of parameters for said operating points with measured output data (118), and
estimate unmeasured output data using said determined values of said parameters and generate complete engine map (102) for calibration.

2. The control unit (110) as claimed in claim 1, wherein said parameters are selected from a group comprising a heat transfer coefficient (h) and a pressure pulsation factor (p).

3. The control unit (110) as claimed in claim 2, wherein said limited operating points are selected based on sensitivity study of said parameters in different operating points/regions of said engine where said parameters show higher non-linearity.

4. The control unit (110) as claimed in claim 1, wherein said engine model (112) is processed iteratively for each of said operating points until a difference between said predicted output and said measured output is within a predetermined threshold range.

5. The control unit (110) as claimed in claim 1 is part of a device, said device is at least one selected from a group comprising a cloud (104), a communication device (106) and a computing device.

6. A method for generating engine maps (102) for calibration, said method comprising the steps of,
receiving input dataset (116) of an engine map (102), characterized by, said input dataset (116) comprises measured output data (118) for limited operating points of an engine of a vehicle (100), said operating point is defined by at least one input variable,
processing said operating points for measured output data (118) through a physics based engine model (112) and predicting an output, said physics based engine model (112) is pre-stored in a memory element (108) of a control unit (110);
optimizing, using an optimizer (120), value of parameters of said engine model (112) using said predicted output;
determining, using a Machine Learning (ML) based data model (114), value of parameters for remaining operating points of said engine with unmeasured output data using optimized value of parameters for said operating points with measured output data (118), and
estimating unmeasured output data using said determined values of said parameters and generating complete engine map (102) for calibration.

7. The method as claimed in claim 6, wherein said parameters are selected from a group comprising a heat transfer coefficient (h) and a pressure pulsation factor (p).

8. The method as claimed in claim 6, comprises iteratively processing said engine model (112) for each of said operating points until a difference between said predicted output and said measured output is within a predetermined threshold range.

9. The method as claimed in claim 6, wherein said limited operating points are selected based on sensitivity study of the parameters in different operating points/regions of said engine and where said parameters show higher non-linearity.

10. The method as claimed in claim 6 is performed by a device, said device is at least one selected from a group comprising a cloud (104), a communication device (106) and a computing device.