Method and system for automated artificial intelligence (AI) testing machine

Abstract

To provide devices and methods for materials testing with reduced error due to reduced variation in testing parameters.SOLUTION: The disclosure is directed at a testing machine for material samples. The testing machine includes a loading station and a testing station along with a pick and place apparatus that moves the material sample being tested between the loading station and the testing station. A control system controls movement of the material sample. The control system also generates testing machine parameters along with testing parameters.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] Cross - Reference to Related Applications This disclosure claims priority to U.S. Provisional Application No. 62 / 703,985, filed Jul. 27, 2018, which is incorporated herein by reference.

[0002] Field of the Disclosure This disclosure relates generally to the manufacture and testing of machines, and more particularly, to methods and systems for automated artificial intelligence testing machines.

Background Art

[0003] Conventional material testing is typically performed by a user who manually loads a material sample into a testing apparatus and then tests the material sample. Examples of material testing include tensile testing, compression testing, dynamic mechanical testing, hardness testing, and wear testing. The parameters used during each test can affect the test results. Depending on the nature of the test, a material sample can be fixed within the testing apparatus by applying pressure to the sample such that the applied pressure appears as a test parameter. When the pressure applied to the sample varies, the measurement results of the material test vary, causing errors in the test. There is a need in the art for apparatuses and methods for material testing that reduce errors due to a reduction in variations in test parameters.

[0004] Accordingly, new methods and systems for automated artificial intelligence testing machines are provided.

Summary of the Invention

Means for Solving the Problems

[0005] In one aspect of the present disclosure, an automated artificial intelligence (AI)-driven testing machine for testing at least one material sample is provided, including a loading station for receiving at least one material sample, a testing station for testing test characteristics of at least one material sample, a pick-and-place (PP) device for transferring at least one material sample between the loading station and the testing station, and a control system for controlling the testing station and the PP device to collect data associated with the testing station.

[0006] In another aspect, the system further includes a measurement station for measuring measurement characteristics of at least one material sample. In another aspect, the loading station includes a loading tray or a magazine loading system. In a further aspect, the testing station includes a pair of AI grippers.

[0007] In yet another aspect, the pair of AI grippers includes a fixed AI gripper and a movable AI gripper. In a further aspect, the movable AI gripper moves with respect to the fixed AI gripper to test at least one material sample. In still a further aspect, the strain and stress of at least one material sample are tested. In one aspect, each of the pair of AI grippers includes an actuator for enabling the AI gripper to grasp at least one material sample. In another aspect, the actuator is a stepper motor.

[0008] In one aspect, the pair of AI grippers further includes a set of sensors. In another aspect, the set of sensors detects slip. In still a further aspect, the control system processes the measurement characteristics to generate parameters for the testing station. In yet another aspect, the parameters are associated with AI gripper characteristics. In yet another aspect, the AI gripper characteristics include grip strength.

[0009] In another aspect of the present disclosure, a method for automated testing of at least one material sample is provided, including receiving at least one material sample, determining test parameters for the at least one material sample, and testing the at least one material sample with the determined test parameters.

[0010] In yet another aspect, determining the test parameters includes determining at least one measured property of the at least one material sample and processing the at least one measured property to determine the test parameters. In another aspect, the test parameters include grip strength or grip force. In still further aspects, testing the at least one material sample includes performing a tensile test on the at least one material sample. In a further aspect, the method includes measuring the stress force applied to the at least one material sample. In another aspect, the method includes measuring the strain force applied to the at least one material sample.

[0011] Embodiments of the present disclosure are now described, by way of example only, with reference to the accompanying drawings.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16A

Figure 16B

Figure 17

Modes for Carrying Out the Invention

[0013] The present disclosure is directed to systems and methods for automated materials testing that use artificial intelligence (AI) to improve sample loading and / or determine test parameters, reduce errors, and automatically perform materials testing.

[0014] FIG. 1 is a front view of an automated artificial intelligence (AI) - driven testing machine 100 with a housing 105. FIG. 2 is a schematic diagram of an embodiment of the automated AI - driven testing machine 100. In one embodiment, the machine 100 includes a loading, or tray - loading section 210 for receiving a sample tray, a first measurement station 220, a second measurement station 221, a pick - and - place (PP) station 230, a test station 240, a controller 250, and a marking system 260. The controller 250 includes a processor 251 and a memory 252 that may include processor - readable persistent data storage. In the drawings, some connections between components are shown, but it will be understood that not all connections are shown.

[0015] A material sample scheduled to be tested by the testing machine 100 can be loaded into the loading section, such as by a sample tray. In other words, the testing machine 100 can receive a material sample by loading the material sample into the sample tray and loading the sample tray into the tray loading section 210. In one embodiment, the sample tray 210 is manually filled and then inserted into the loading section. In another embodiment, the sample tray can be a permanent component within the housing 105, and the samples can be individually inserted into the sample tray. This insertion can be performed manually or in an automated manner. The PP system 230 is used to transfer the material sample within the testing machine 100. For example, the PP system can transfer the material sample in an automated manner between different stations within the machine 100, such as between the sample tray or loading station 210, the first measurement station 220, the second measurement station 221, the marking station 260, and the testing station 240. In one embodiment, the processor 251 can access a program stored in the memory 252 to control the movement of the PP system 230 or control the movement of the sample based on an input from the user. The first measurement station 220 can measure the first measurement characteristics of the sample, such as the hardness, surface roughness, and / or density of the sample. The hardness can be determined, for example, by a Rockwell hardness test, a Vickers hardness test, a Knoop hardness test, and / or a Brinell hardness test. The second measurement station 221 can measure the second measurement characteristics of the sample, such as the thickness and width of the sample. The thickness and width of the sample can be determined, for example, by a dial gauge, a dial thickness gauge, a high-resolution camera, a line scan system, a laser rangefinder, and / or edge detection. In a preferred embodiment, the second measurement station can be calibrated with the known thickness and width of a standard sample. The measurements obtained by the measurement stations 220 and 221 can be stored in the memory 252. It will be understood that the system can include other measurement stations for determining the measurement characteristics of the material sample.

[0016] Measurements can be used as data for the test station 240 and for post-test analysis, and to modify test parameters. In a preferred embodiment, each of the measurement stations 220 and 221 is an integrated part or component of the machine 100, although the stations 220 and 221 can be peripheral components that are added to and / or removed from the machine 100 as needed.

[0017] The marking system 260 can apply visible marks to the material sample in an automated manner. For example, the marking system 260 can apply two marks to the material sample for testing, analysis, or information gathering purposes. The marking system 260 can include a marker, an inkjet printer, a laser, or any other means of marking the sample. Although not shown, the test station 240 preferably includes a set of AI grippers, as will be described in more detail below.

[0018] The processor 251 can load data from the memory 252 and compare the parameters of the sample and the test station 240 with the parameters from previous samples and tests. The processor can also send commands to the controller 250 to change the characteristics of the AI grippers.

[0019] The test station 240 can test the material sample in an automated manner, for example, by using AI grippers to perform tests on the sample. Non-exclusive examples of tests that can be performed include, but are not limited to, tensile, tear, fatigue, compression, flexure, and bend tests.

[0020] Regarding a tensile test, the sample is typically gripped at both ends by an AI gripper, and the gripping force and the gripping position are determined by a processor, for example, by input from a user or by data from a measurement station. Then a tensile force is applied to the sample, typically until the sample breaks, using the AI gripper, with the force (i.e., stress) required to pull the sample and the elongation of the sample (i.e., strain) resulting from the measured tensile force. The stress-strain relationship provides information about the properties of the material sample and can include the strength, toughness, modulus, signs of plastic deformation, etc. of the sample. The gripping force can be determined by the user or retrieved from memory and can vary depending on the material. If the gripping force is too low, the sample can slip during the tensile test, causing sudden changes in the measured stress and the measured strain, and thus errors in the measurement. If the gripping force is too high, the sample can be damaged and prematurely broken, also causing errors in the measurement. In the present disclosure, the gripping strength can be determined by measurement to reduce the likelihood of errors during the test. The systems, devices, and methods of the present disclosure are discussed with respect to tensile tests for clarity, but those skilled in the art who benefit from the present disclosure will understand that the present disclosure can be applied to a wide variety of material tests, such as compression tests, dynamic mechanical tests, wear tests, and the like.

[0021] In one embodiment, the test station 240 can perform a tensile test on a sample by pulling the sample at a strain rate of 8.33 mm / s. In one embodiment, the test station 240 can perform a tensile test on a sample by pulling the sample at a maximum strain rate of 100 mm / s. The test station 240 can also perform a tensile test on a sample by pulling the sample with a tensile force of up to 1,000 newtons, or up to 10,000 newtons. To maintain a constant strain rate, the tensile force can be dynamically adjusted during the test. The test station 240 can stop the test when sample breakage occurs, for example, by detecting when the tensile force required to maintain a constant strain rate drops to at least approximately zero.

[0022] In one embodiment, the test station 240 includes a computer vision system such as a high-resolution camera. The computer vision system is positioned and oriented to generate a video of the sample when the sample is being tested and is communicatively coupled to the controller 250. The video is stored in the memory 252 and can be analyzed by a computer vision program that is executed to monitor the positions of the marks created by the marking system. The positions of the marks can be used by the processor to determine the distance between the marks as determined by the computer vision system, thereby determining the distortion of the sample when the sample is pulled by the test station. The position of the marks and / or the distance between the marks can be calibrated with a calibration sample. In addition to determining the mark positions, the computer vision system can determine the sample loading position and compare the sample loading position to a preferred sample loading position. The sample loading position can be determined by the computer vision system by overlaying an image of the sample obtained by the computer vision system on a preferred image stored in the memory 252 to determine any difference between the actual position of the sample and the preferred position of the sample in the reference image. The position of the sample can be determined by the computer vision system by comparing the position of the sample to a physical reference position whose position can be verified by the computer vision system. The preferred sample loading position can be a sample loading position correlated with successful test performance by an AI algorithm. The computer vision can determine the elongation of the sample with an error of 1% or less. The computer vision system can include two synchronized cameras to determine the distortion of the sample when the sample is being tested.

[0023] The computer vision system can also determine the shape of the sample, compare the sample shape to a known sample shape, and automatically select a test for a matching sample shape. The computer vision system can also determine the distortion of the sample by directly analyzing changes in the shape of the sample as determined by the computer vision, i.e., without using marks.

[0024] Immediately prior to testing, the AI gripper can adjust the grip strength and distance based on feedback from previous tests. The feedback can include measured parameters such as the hardness, thickness, width, density, and surface roughness of the sample, and / or data from similar samples that have already been tested in the past. This past data, along with the sample data for each sample, and its AI analysis are used to determine the preferred grip strength and are used during testing within the test station 240 to perform the test in a repeatable manner. In this regard, the AI gripper can learn from each test performed, and after each test, the accuracy of the optimal or preferred grip strength determination can be improved.

[0025] Figure 3 shows a schematic diagram of a system 300 for determining test parameters using AI. The system 300 includes an input component that provides an input 320 to a processor 310 that processes the input 320. The processor 310, which can be the same as the processor 251, preferably includes an algorithm 310 that processes the input 320 to determine test parameter values 330 for improving the grip strength or parameters of the AI gripper. Non-exclusive examples of the input 320 include material sample composition, hardness, thickness, width, and density. Non-exclusive examples of the test parameter values 330 are grip force, grip closing distance, and dynamic closing rate. The dynamic closing rate is the ratio of sample strain to sample thickness at that strain, in other words, the amount by which the grip closing distance of the AI gripper can be decreased to compensate for the thinning of the sample as it stretches. Improving the gripping ability of the AI gripper can include improving the ability of the gripper to grasp various materials. Improving the gripping ability of the gripper can include grasping the sample with test parameters that correlate with successful tests. Additionally, in some embodiments, the PP system can include a movable gripper, and the grip strength of the movable gripper can be the same as the grip strength of the AI gripper.

[0026] Figure 4 shows a flowchart for method 400 for automated AI testing of materials. First, a material sample is loaded into or received by an AI-driven testing machine (410). Loading a material sample into an AI-driven testing machine can include loading the material sample into a single sample holder and loading the sample holder into the AI-driven testing machine. Another example of loading a material sample into a machine can include loading a plurality of material samples into a plurality of slots in a loading tray.

[0027] A set of material sample parameters is then determined or measured (420). The sample parameters can be determined by measuring the properties of the material sample at at least one measurement station to generate measurement data. The material sample parameters can also be determined by accessing data associated with the material sample in a database and / or in memory. The measurement data can include physical dimensions (length, thickness, shape), composition (chemical composition, crosslink density, filler size and volume fraction, processing history), viscoelastic properties, hardness, toughness, strength, and coefficients.

[0028] The material sample parameters are then analyzed to provide a set of AI test parameters (430). In one embodiment, the set of material sample parameters can be analyzed by a processor using an AI algorithm trained with respect to training data stored in memory. The training data can include test parameters such as, but not limited to, grip strength and grip position. Prior to testing, the AI algorithm can be trained with respect to the training data, which can include analyzing test parameters for successful (e.g., no slippage) and unsuccessful (e.g., slippage occurs) tests to correlate the set of AI test parameters with successful tests.

[0029] Analyzing a set of sample parameters with an AI algorithm to provide a set of AI test parameters may also include analyzing multiple sets of sample parameters with the AI algorithm to provide multiple sets of AI test parameters, for example, by sequentially analyzing each set of sample parameters.

[0030] In one embodiment, the AI test parameters may include a fixed AI grip position, a movable AI grip position, and an AI grip strength. The fixed grip position may be determined by moving the sample relative to the fixed AI grip with a PP system. The movable grip position may be determined by moving the sample relative to the movable AI grip with a PP system or by moving the movable AI grip relative to the sample. The grip strength may be above a threshold for sample slippage, below a threshold for sample damage, or both.

[0031] The material sample is then transferred to a test station (440), such as via a PP system. The material sample is then tested according to the AI test parameters to generate test data (450). For example, the tensile strength of the material sample can be tested. In this example, the processor sends the AI test parameters (such as grip position and strength) to the AI grip and grips the sample with the determined AI test parameters. The sample can then be tested by pulling the sample apart with two AI grips (as described above with respect to stress and strain). The AI grip strength can be monitored with a pressure sensor. In another embodiment, testing the material sample in an automated manner according to the AI test parameters may include pulling the material sample by separating the movable grip from the fixed grip, measuring the strain of the material sample while pulling the material sample to generate strain data, and measuring the stress of the material sample while pulling the material sample to generate stress data.

[0032] Testing a material sample in an automated manner according to AI test parameters can include marking the material sample with at least two strain gauge marks. Measuring the strain of the material sample can include recording a video of the material sample while it is being pulled and analyzing that video with a computer vision algorithm. Recording the strain data can include recording the video, for example, in a memory (252). Recording the video can enable the video to be played back later, for example, after a failed test, to identify the cause of the test failure. Pulling the material sample can include monitoring for slippage of the material sample and flagging the test data with a slip flag if slippage occurs. Slippage can be monitored with a slip sensor or by changes in stress and / or strain rate. A test flagged with a slip flag can be reexamined to identify the root cause of the slippage, for example, by reviewing the video of the test as described above.

[0033] After the test is completed, the torn material sample can be removed by a gripper, for example, to a sample holder or tray, etc. Removing the material sample from the AI-driven testing machine can include transferring at least two sample pieces to a second part of the sample holder in an automated manner and removing the sample holder from the AI-driven testing machine. If multiple samples are tested, loading, determination, analysis, transfer, testing, and removal can be repeated for the next sample.

[0034] In another embodiment, continuous material testing can enable generating updated training data by associating a set of AI test parameters, a set of sample parameters, and test data with the training data, and training an AI algorithm with the updated training data to, for example, improve the accuracy of the AI algorithm.

[0035] FIG. 5 shows a more detailed front view of a testing machine 100 without a housing. FIG. 6 shows a perspective view of the testing machine of FIG. 5, and FIG. 7 shows a perspective view of a part of the testing machine.

[0036] The testing machine 100 includes a frame 110, a base 115, a pick and place (PP) system 120, rails 125, a tension or testing system 130 including two AI grippers 135, and a loading system 140. The first AI gripper is movably coupled to the rails 125 by a linear movement system and can be regarded as a movable gripper, and the second AI gripper 130 is fixedly coupled to the base 115 and can be called a fixed gripper. The loading system 140 is coupled to the base 115. The linear movement system can be a ball screw linear actuator driven by a servo motor, or a pulley and belt system driven by a servo motor, a DC motor or an AC motor.

[0037] The housing 105 surrounds all the components inside the testing machine 100 and has a plurality of positions for access and maintenance. The loading system 140 includes all the components necessary to insert or receive a sample into the testing machine 100. The PP system 120 transports the sample, for example, from the loading system 140 to the AI grippers 135 through the machine. The testing system 130 includes the AI grippers 135, a load cell, sensors and a linear movement system to ensure that the test is completed by the machine.

[0038] Samples are loaded into the testing machine 100 in a woven manner through an opening in the housing 105. The embodiments shown in FIGS. 1 and 5-7 use a tray 142 as shown in FIG. 8, but other embodiments may use other loading systems such as a magazine loading system or a system where samples are stacked and set inside the machine. The tray 142 includes 12 slots 144, and each slot can hold a sample. In alternative embodiments, the tray 142 can include a different number of slots 144, such as 6, 12, or any number of slots 144. The tray 142 includes a compartment 146 for holding fragments of the tested samples.

[0039] The loading system 140 can place samples in an arranged manner at positions where they can be picked up by the PP system 120. For example, each sample held in each slot 144 can be picked up in turn by the PP system 120. The order can be any desired order. Conveniently, the identification of each sample held in each slot 144 can be correlated by the testing machine 100 with the data generated by the testing of each sample. The tray 142 can move horizontally in a straight line to align each slot 144 with the PP system 120.

[0040] The tray 142 can include at least one sensor to provide sample loading information. Non-exclusive examples of sample loading information include alignment information (e.g., whether the tray 142 is properly loaded into the testing machine 100, calibration information for determining the position of each slot 144 relative to the PP system 120) as well as sample quantity and position information (e.g., which slots 144 contain samples, whether each sample is placed within each slot to enable automated sample testing). The testing machine 100 and / or the tray 142 can include sensors to detect whether the tray 142 is inserted into the testing machine 100, and the testing machine 100 can be configured to start sample testing only if the tray 142 is detected as being inserted into the testing machine 100.

[0041] Once the sample is loaded into the machine, the PP system 120 can move the sample to multiple positions within the testing machine 100.

[0042] The PP system 120 includes a movable gripper 122 for gripping a material sample held within one of the slots 144. In a preferred embodiment, the PP system 120 is vertically movable and can move a sample gripped by the movable gripper 122 in that direction. The upward vertical movement of the sample can position the sample within the AI grip. The sample can be transferred from the movable gripper 122 to the AI grip such that the AI grip can grip the sample and the movable gripper can then release that sample. The sample, now only gripped by the AI grip, can then be tested. After the test, the sample (or a fragment of the sample) can be gripped by the movable gripper 122 such that the AI grip 135 can release the sample piece, and the fragment can be vertically moved downward to return the sample to the tray 142.

[0043] Performing the test includes pulling the sample by releasing the movable gripper (movably coupled to the rail) from the fixed gripper. The AI grip can pull the sample by gripping the sample while the linear movement system releases the movable gripper from the fixed AI grip. Once the system completes the test, the sample is removed from the AI grip by the PP system 120, the fragments of the sample are returned to the tray 142, and the next sample is tested until all available or required samples have completed all tests. Returning the sample to the tray 142 can include returning the sample to compartment 146 of the tray 142 if the testing of the sample includes destroying the sample.

[0044] The PP system 120 can also position the material sample within the AI grip 135 at a plurality of positions, each position including a different height, lateral position, and / or angle of the sample relative to the AI grip.

[0045] For a test, for example, a rubber sample can be gripped with an AI grip strength determined by an AI test parameter of the grip strength used for a successful tensile test of the rubber sample, and a successful test is defined as a test in which neither slippage nor sample breakage due to excessive grip strength occurred. As another example, a nylon 6,6 sample can be gripped with an AI grip strength determined by a test parameter of the grip strength used for a successful tensile test of the nylon sample.

[0046] Figure 9 shows a perspective view of an AI grip. The AI grip 900 can be substantially similar to the AI grip 135. The AI grip 900 includes a grip housing 910, an actuator 920, a coupler 930, and a pressure pad 940. During operation, the actuator 920 generates a compressive pressure on a sample held between two pressure pads 940 by applying a linear force to the coupler 930. The linear force on the coupler 930 is transmitted through the coupler 930 to the second pressure pad 940. The second pressure pad 940 disperses the linear force across the surface of the sample in contact with the second pressure pad 940 to generate a compressive pressure.

[0047] The actuator 920 can be a stepper motor (such as shown in FIG. 9), a DC motor (such as shown in FIG. 14), a pneumatic actuator, or any type of mechanism that can be used to generate a linear pressure. The pressure pad 940 is preferably designed so that the sample does not slip during the test, but also so that the gripped portion of the sample is not damaged during the test. In one embodiment, the surface of the pressure pad is made of a plurality of coatings that can improve the grip for all materials during the test. An example of a pressure pad design is a fish scale design, shown in FIG. 16A. Another example of a pressure pad design is a fish scale design combined with a sandpaper design, shown in FIG. 16B.

[0048] Figure 10 shows a front view of another embodiment of the AI gripper 900. Along with the gripper housing 910, the actuator 920, the coupler 930, and the set of pressure pads 940, the gripper 900 further includes a pressure sensor 950, preferably internal, for measuring pressure. In this embodiment, the pressure sensor 950 is coupled to the housing 910. As described above, the actuator 920 generates a compressive pressure on the sample held between the pressure pads 940 via the coupler 930, and the pressure sensor 950 measures the intensity or force of the compressive pressure generated by the actuator 920.

[0049] The sensor 950 can be a small load cell, a brake load cell, a force sensing resistor (FSR), a quantum tunnel composite (QTC), or any other sensor for measuring pressure / force. The pressure sensor 950 can provide feedback to a processor to ensure that the sample is gripped at a pressure that reduces the likelihood of slippage. Figure 11 shows a front view of an embodiment of the AI gripper 900 with a pressure sensor in an alternative shape, where the sensor 952 is disposed outside the housing 910. In this embodiment, the pressure sensor can measure the pressure transmitted from the actuator 920 through the pressure pad 940, the sample, and the housing 910.

[0050] Figure 12 is an exploded view of the AI gripper 900. Figure 13 shows a front view of an embodiment of the AI gripper 900 with two actuators. The first actuator 920 and the second actuator 921 generate compressive pressures from both sides of the AI gripper 900. The AI gripper 900 includes a housing 910 coupled to the first actuator 920 and the second actuator 921. The coupler 930 is coupled to the first actuator 920. The first pressure pad 940 is coupled to the coupler 930. The second pressure pad 941 is coupled to the second actuator 921.

[0051] Figure 14 shows a front view of another embodiment of the AI gripper 900. In this embodiment, the actuator 921 is a DC motor.

[0052] FIG. 15 shows a horizontal cross-sectional view of one embodiment of another embodiment of the AI gripper 900. In this embodiment, the gripper 900 includes a slip sensor 960 for detecting slip. The gripper housing 910 is coupled to a slip sensor 960 that detects whether the sample slips during the test. The slip sensor 960 can be a laser measurement system, an electromechanical switch in physical contact with the sample, or any other sensor that detects movement. The slip sensor 960 can provide feedback so that the test can be flagged if slip occurs during the test. The AI gripper 900 can also dynamically move and / or increase the grip pressure to prevent slip and ensure that the results for that sample are not lost. Additionally, the AI gripper 900 can include both a slip sensor 960 and a pressure sensor 950. During the test, the gripper can detect slip through the pressure sensor and / or the slip sensor and can automatically adjust the grip pressure to stop the slip. If it is not possible to stop the slip, the machine can flag the test and / or analyze the results to confirm whether the slip affected the results.

[0053] FIG. 17 is a flowchart outlining a method for generating a material with an AI predicted composition.

[0054] First, a set of material property requirements is received (1710). Non-exclusive examples of material property requirements include hardness, toughness, Young's modulus, storage modulus, loss modulus, wear resistance, maximum fracture strain, strain at the onset of plastic deformation, and creep rate. The material property requirements can be viewed as a set of values to be satisfied by the material generated by method 1700.

[0055] The AI algorithm is then trained on a dataset (1720). The dataset can include test data from material samples having properties similar to the set of material property requirements. The AI algorithm can be a linear iterative algorithm. Training the AI algorithm can include correlating the material sample composition with the material sample properties as compared to the material sample properties generated by the material sample composition.

[0056] The set of material property requirements is then modeled by an AI algorithm to generate an AI predicted composition (1730). The AI predicted composition can include chemical composition (polymer chain length and distribution for a polymer sample, type and volume fraction of filler, presence and density of crosslinks, type and volume fraction of plasticizer), and processing conditions (maximum temperature, heating and cooling rates, pressure). The AI predicted composition can be the composition with the highest probability of meeting or exceeding the set of material property requirements as identified by the AI algorithm.

[0057] A material sample with the AI predicted composition is then manufactured (1740). Manufacturing the material sample enables testing of that material sample. The material sample is then tested on an AI-driven testing machine to determine a set of material sample properties (1750). The material sample properties determined by the AI-driven testing machine can be the same properties as the set of material property requirements.

[0058] The set of material sample properties is compared with the set of material property requirements to determine an accuracy level (1760). The accuracy level can be the ratio of important material properties, e.g., material sample hardness divided by required material hardness × 100%. The accuracy level can be a weighted average of the ratios of multiple material properties. The accuracy level is a binary (yes / no) value, where yes corresponds to all material sample properties that meet or exceed the material property requirements, and no corresponds to at least one material sample property that neither meets nor exceeds the material property requirements.

[0059] When the accuracy level exceeds the accuracy level threshold, a material with an AI-predicted composition is generated. For example, the accuracy level threshold can be 100% for important material properties, 100% for a weighted average of multiple material properties, or no for a binary accuracy level (yes when exceeding the threshold). When the accuracy level does not exceed the accuracy level threshold, the material composition, the set of material sample properties, and the accuracy level are added to the dataset to update the dataset portion of the method. The method can be repeated until a material sample is generated at an accuracy level exceeding the accuracy level threshold. The portion of the method can be repeated until the accuracy level is not significantly higher than the accuracy level of previously generated samples, where significantly higher can be 1% higher, 0.1% higher, or less than 0.1% higher.

[0060] An AI model, such as a multiple linear iteration method, can predict a material composition to achieve material properties such as strength or hardness. When a material with the predicted composition is fabricated, an automated test can be performed using the testing machine 100, and the data obtained from the automated test can then be fed back to the AI model to improve the AI model's accuracy. In one embodiment, a sample is received by the system. The sample is then placed within a gripping device (such as an AI gripper). The composition of the sample is then determined, for example, by comparing the characteristics of the sample to records stored in a database. These characteristics can be obtained by sensors within the system that detect the characteristics. Non-exclusive examples of characteristics include hardness, thickness, width, surface finish, and surface friction. The grip strength of the AI gripper can then be adjusted in response to the determination of the sample composition.

[0061] In one embodiment, the present disclosure describes a self - learning AI gripper. Thus, as more tests are performed on samples of various characteristics, the gripper features can be updated to correspond to the material being tested. This can, over time, reduce the likelihood of sample slippage during sample testing, thereby improving the effectiveness of sample gripping. The AI learning component can improve the capabilities of an automated AI - driven testing machine to test a wide variety of samples and materials with improved gripper strength accuracy.

[0062] In the foregoing description, for the purposes of explanation, numerous details have been set forth in order to provide a thorough understanding of the embodiments. However, it will be apparent to those skilled in the art that these specific details may not be necessary. In other instances, well - known structures may be shown in block diagram form in order not to obscure the understanding. For example, specific details are not provided regarding whether the elements of the embodiments described herein are implemented as software routines, hardware circuits, firmware, or combinations thereof.

[0063] Embodiments of the present disclosure, or components thereof, can be provided as, or represented by, a computer program product stored in a machine-readable medium (also referred to as a computer-readable medium having computer-readable program code embodied therein, a processor-readable medium, or a computer-usable medium). The machine-readable medium can be any suitable tangible, persistent medium, including magnetic, optical, or electrical storage media, such as floppy disks, compact disk read-only memory (CD-ROM), memory devices (volatile or non-volatile), or similar storage mechanisms. The machine-readable medium can contain various sets of instructions, code sequences, configuration information, or other data, which, when executed, cause a processor or controller to execute steps within a method according to an embodiment of the present disclosure. Those skilled in the art will understand that other instructions and operations necessary to implement the described embodiments can also be stored on the machine-readable medium. Instructions stored on the machine-readable medium can be executed by a processor, controller, or other suitable processing device and can interface with circuitry for performing the described tasks.

[0064] The foregoing embodiments are for illustrative purposes only. Changes, modifications, and variations can be made by those skilled in the art to a particular embodiment without departing from the scope, which is defined only by the claims appended hereto.

Claims

[Claim 1] The invention as described in the description and / or drawings.

Images