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 disclosure relates to the field of material science and data analysis. In particular the present invention discloses an AI based system for material identification and a method of the training an AI model thereof.

Background of the invention
[0002] Experimental investigations have been carried out on materials since centuries to classify material and identify them based on their experimental properties. Even today Material characterization techniques that are widely used to characterize the physical and chemical properties of materials at the nanoscale often require significant human effort to interpret and extract meaningful physicochemical insights. The advent of data science in the field of Materials Engineering also meant the beginning of new field of Materials Informatics wherein the experimental results are being analyzed and interpreted using machine learning and artificial intelligence techniques. The AI/ML based techniques use large amounts of data to identify the hidden correlations and patterns in the data. In material characterization this is reflected in terms of properties of material that can be discovered using these hidden correlation in the data which is missed in manual identification. Artificial intelligence (AI) techniques such as machine learning (ML) have the potential to improve the efficiency and accuracy of material characterization by automating data analysis and interpretation.

[0003] Patent Application US2022318658AA titled AI-accelerated characterization of materials discloses devices, systems, and methods for material characterization can include detecting definitional data from material samples that are positionally encoded according to know attributes as operational data, characterizing at least some of the samples as training data, and processing the training data via a machine learning model to train the model and/or to characterize the remaining samples based on the training data.

[0004] Patent application US2022051141 AA titled “Material characterization system and method” discloses method, apparatus, system, and computer program product for estimating material properties. Training data comprising results of testing samples for a set of materials over a range of loads applied to the samples is identified by a computer system. A machine learning model is trained by the computer system to output the material properties for materials in structures using the training data. While both these patent applications hint at usage of AI/ML based techniques for material characterization but they don’t lay emphasis on harnessing the application of material properties and their related combinations. Especially in the case of polymeric materials the substantial correlations between the properties are important parameters for the training data of AI/ML models.

Brief description of the accompanying drawings
[0005] An embodiment of the invention is described with reference to the following accompanying drawings:
[0006] Figure 1 depicts an AI based system (10) to identify a material ;
[0007] Figure 2 illustrates method steps to train an AI model (M) for material identification.

Detailed description of the drawings
[0008] Figure 1 depicts an AI based system (10) to identify a material. The AI based system (10) comprises an AI model (M), a control unit (11), a plurality of instrumentation apparatus (12) and optionally a visual output interface (14). The AI based system (10) has access to an extensive Training Set (TS) which is combination/compilation of Materials Database (MD) and the Properties Database (PD) of the respective materials, X1, X2, . . . Xn.

[0009] An AI model (M) with reference to this disclosure can be explained as a component which runs on specific data-driven algorithm for set application. A model can be defined as reference or an inference set of data, which uses different forms of correlation matrices. Using the data from these models, correlations can be established between amongst data to arrive at some logical understanding. A person skilled in the art would be aware of the different types of AI model (M)s such as linear regression, naïve bayes classifier, support vector machine, artificial neural networks, and the like.

[0010] Some of the typical tasks performed by AI model (M)s are classification, clustering, regression etc. Majority of classification tasks depend upon labeled datasets; that is, the datasets are labelled manually in order for neural network to learn the correlation between labels and data. This is known as supervised learning. Some of the typical applications of classifications are: face recognition, object identification, gesture recognition, voice recognition etc. Clustering or grouping is the detection of similarities in the inputs. The cluster learning techniques do not require labels to detect similarities. Learning without labels is called unsupervised learning. In the present disclosure the AI model (M) is trained on Materials Properties Database (PD) with respect to the Materials Database (MD) (i.e. to be identified) using a supervised training mechanism. The AI model (M) works coherently with the control unit to generate a particular correlation which will help to classify and identify the materials as per any given property.

[0011] The control unit (11) is logic circuitry and software programs that respond to and processes logical instructions to get a meaningful result. It may be implemented in the system as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, one or more microchips or integrated circuits interconnected using a parent board, hardwired logic, software stored by a memory device and executed by a microprocessor, firmware, an application specific integrated circuit (ASIC), and/or a field programmable gate array (FPGA), and/or any component that operates on signals based on operational instructions.

[0012] The control unit (11) is configured to receive values of properties of the material measured by the one or more instrumentation apparatus (12). The control unit (11) memory further stores a pre-determined identification information for a range of values of the one or more properties of the material. This information is used to train the AI model (M) based on the received properties of material and stored pre-determined identification information. The identification information stored in the control unit (11) is updated based on the output of the trained AI model (M) and displayed on the output visual interface (14).

[0013] The control unit (11) can further comprises a specific component termed a server computer for executing machine learning algorithms to classify material properties based on pre-fed data from a materials database. This component of the control unit (11) can either reside within the control unit or be remotely connected to it. It is equipped with dedicated GPUs, scalable architecture, and high-speed networking capabilities, which enable the server computer to accelerate model training and inference while maintaining reliability through redundancy measures and robust security protocols. It is further optimized with software libraries and frameworks for machine learning tasks, alongside monitoring and management tools. In particular the server computer enables accurate classification, by leveraging the wealth of information contained within the materials database.

[0014] Instrumentation apparatus (12) are configured to measure one or more properties of the given material . The one or more properties are thermal properties, chemical properties, mechanical properties, morphological properties, rheological properties. The table below illustrates the properties and the methodology used to measure the properties for Elastomers and Plastics.

Apparatus and
Technique Properties Measured Type of Property
DSC - Differential Scanning Calorimetry Tg Thermal Properties
Tc
Tm
Cp
TGA- Thermogravimetric analysis Degradation Temp. Thermal Properties
Filler Percentage (%)
FTIR - Fourier Transform Infrared Spectroscopy Functional Groups Chemical Properties
SEM/EDS - Scanning Electron Microscopy /
Energy Dispersive Spectroscopy Filler Size Microstructure
Distribution
Microtome
- Morphology – Crystalline or Amorphous Morphology
DMA - Dynamic mechanical analysis G’ – Stirage Modulus Rheological Properties
G’’ – Loss Modulus
Tan d - Damping Factor
UTM- Universal Testing Machines Mechanical Properties – Tensile Strength, Modulus, etc. Mechanical Properties

Table 1.1

[0015] It should be understood at the outset that, although exemplary embodiments are illustrated in the figures and described below, the present disclosure should in no way be limited to the exemplary implementations and techniques illustrated in the drawings and described below.

[0016] Figure 2 illustrates method steps to train an AI model (M) for material identification. The AI model (M) is deployed in the AI system (10) explained in accordance with figure 1. The AI model (M) is configured to receive values of one or more properties of the material measured by instrumentation. The information is provided to the AI model (M) by the control unit (11).

[0017] Method step 201 comprises feeding the value of one or more properties of the material as input to the AI model (M). This information is referred to as the Training set (Ts) which comprises the Materials Database (MD) and the Properties Database (PD). The one or more properties comprise at least one or more from a group of thermal properties, chemical properties, mechanical properties, morphological properties, rheological properties.

[0018] Method step 202 comprises receiving a pre-determined identification information for the values of one or more properties received. The pre-determined identification information for a range of values of the one or more properties of the material is stored in the control unit (11). This pre-determined identification information correlates the values of properties with a specific material in the Training set (Ts).

[0019] Method step 203 comprises executing the AI model (M) with said input to receive a crude output. This is called a crude output because as of now the AI model (M) has not learnt the correlation between values of the properties of materials and the specific type of material it corresponds to.

[0020] Method step 204 comprises defining a loss function based on the pre-determined identification information and the crude output. Loss function here refers to the difference between the crude output and the pre-determined identification information. This loss function is close to zero when the AI model (M) is trained.

[0021] Method step 205 comprises optimizing the loss function for different sets of values of the one or more properties of the material and the corresponding pre-determined identification information. Optimizing the loss function for a neural network type AI model (M) means we basically tune or configure the hyperparameters and network parameters until we reach a minimal loss function. For other AI model (M) types, we manually label the crude output and the corresponding values of plurality of properties to a particular material type. Multiple iterations are repeated for the above method steps for different materials having different sets of values of the properties.

[0022] The AI model (M) after optimizing the loss function identifies the material when fed with value of one or more properties of the material. Identifying the material comprises classifying the material into a known material type or new material type. This information is displayed on the visual output interface (11). If the values of the properties don’t correspond to a known material type these properties are clustered and the material is classified as a new material. The new material type (Xu) along with the values of it’s properties is stored in the control unit (11) as pre-determined identification information in the Training Set (Ts).

[0023] The proposed method and AI based system (10) dispenses with the need for physical experimentation and techniques to find the properties of materials. This is very pertinent in case of manufacturing industry. Further a new material can identified through output of properties of materials. This can reduce the development cost in testing phase up to around 50%. Further the usage of the proposed system in real-time reduces chances of failures or defects in the product in long term.

[0024] A person skilled in the art will appreciate that while these method steps describes only a series of steps to accomplish the objectives, these methodologies may be implemented with modification and customizations to the AI based system (10). It must be understood that the embodiments explained in the above detailed description are only illustrative and do not limit the scope of this invention. Any ancillary modification to the method of training AI model (M) for material identification and the AI system (10) thereof is envisaged. The scope of this invention is limited only by the claims.

, Claims:We Claim:
1. An AI based system (10) to identify a material, the system comprising an AI model (M) in communication with a control unit (11), the control unit (11) configured to receive information from a plurality of instrumentation apparatus (12), characterized in that system :

the control unit (11) configured to:
receive a value of one or more properties of the material from the instrumentation apparatus (12);
store a pre-determine identification information for a range of values of the one or more properties of the material;
train the AI model (M) based on the received properties of material and stored pre-determined identification information;
update the identification information stored based on the output of the trained AI model (M).

2. The AI based system (10) to identify a material as claimed in claim 1, wherein the one or more properties comprise at least one or more from a group of thermal properties, chemical properties, mechanical properties, morphological properties, rheological properties.

3. The AI based system (10) to identify a material as claimed in claim 1, wherein identifying the material comprises classifying the material into a known material type or new material type.

4. The AI based system (10) to identify a material as claimed in claim 1, wherein updating the identification information comprises storing the new material type and the corresponding value of the one or more properties in the control unit (11).

5. A method to train an AI model (M) for material identification, said AI model (M) configured to receive values of one or more properties of the material measured by instrumentation, the method comprising:
Feeding the value of one or more properties of the material as input to the AI model (M);
Receiving a pre-determined identification information for the one or more properties received;
Executing the AI model (M) with said input to receive a crude output;
Defining a loss function based on the pre-determined identification information and the crude output;
Optimizing the loss function for different sets of values of the one or more properties of the material and the corresponding pre-determined identification information.

6. The method to train an AI model (M) to identify materials as claimed in claim 5, wherein the one or more properties comprise at least one or more from a group of thermal properties, chemical properties, mechanical properties, morphological properties, rheological properties.

7. The method to train an AI model (M) to identify materials as claimed in claim 5, wherein the pre-determined identification information for a range of values the one or more properties of the material is stored in a control unit (11).

8. The method to train an AI model (M) to identify materials as claimed in claim 5, wherein the AI model (M) after optimizing the loss function identifies the material when fed with value of one or more properties of the material.

9. The method to train an AI model (M) to identify materials as claimed in claim 5, wherein identifying the material comprises classifying the material into a known material type or new material type.