Machine learning based dynamic compression-shear testing machine adaptive control method and system

Abstract

The application relates to the technical field of compression-shear testing machines, in particular to a dynamic compression-shear testing machine adaptive control method and system based on machine learning, which comprises the following steps: constructing a material parameter prediction model to predict the detection parameters and mechanical properties of the material; correcting the control parameters of the compression-shear machine through an adaptive control algorithm according to the real-time response data of the material in the testing process; collecting the surface image and internal ultrasonic signal of the material, extracting failure characteristic parameters such as crack area proportion, number of broken sections and maximum fracture length, and weighting and fusing to construct a critical failure criterion of the material; triggering safety protection when the material failure characteristic parameters reach a preset threshold value, automatically stopping the test and recording relevant data; and using the recorded data under the failure condition to incrementally learn and continuously optimize the material parameter prediction model. The method realizes intelligent adaptive control of the dynamic compression-shear testing process, and improves the testing efficiency, precision and safety.

Description

Technical Field

[0001] This application relates to the field of compression-shear testing machine technology, specifically to a dynamic compression-shear testing machine adaptive control method and system based on machine learning. Background Technology

[0002] Dynamic compression-shear testing is an important method for evaluating the mechanical properties of materials under high strain rate and large deformation conditions. Traditional dynamic compression-shear testing machines apply loads to materials through preset control parameters, obtain the material's mechanical response, and then calculate the material's mechanical property indicators based on the test data. However, the optimal testing parameters differ for different materials, and using uniform control parameters makes it difficult to account for various materials, leading to reduced accuracy and reliability of test results. Furthermore, the mechanical behavior of materials during dynamic compression-shear processes is complex and variable. If the control parameters are not adjusted in time to adapt to the material's real-time response, it may cause premature failure or damage to the material, affecting the safety and effectiveness of the test.

[0003] In dynamic compression-shear testing, determining whether a material has reached its failure limit state is a crucial issue. Traditional methods primarily rely on abrupt changes in the load-displacement curve or macroscopic cracks on the material surface to predict failure. However, these indicators often lag behind the generation and evolution of internal damage, making it difficult to accurately capture the onset of material failure. The initiation and propagation of internal defects are the root cause of ultimate failure. However, because the internal damage process is difficult to observe directly, it is impossible to monitor the material's integrity in real time, resulting in insufficient early warning capabilities for material failure.

[0004] Existing dynamic compression-shear testing machines lack effective safety protection mechanisms, failing to stop loading and record data promptly when materials reach critical failure states. Furthermore, the lack of experimental data under material failure states hinders in-depth research into material failure mechanisms and the summarization of failure patterns, making it difficult to further improve the accuracy and reliability of material mechanical property assessments.

[0005] In view of this, this application proposes an adaptive control method for dynamic compression-shear testing machines based on machine learning. Summary of the Invention

[0006] To achieve the above objectives, this application provides an adaptive control method and system for a dynamic compression-shear testing machine based on machine learning, including:

[0007] Acquire data on the material to be tested and construct a material parameter prediction model based on machine learning technology;

[0008] Based on the real-time response data of the material during the dynamic compression-shear test, the control parameters of the dynamic compression-shear testing machine are corrected by the control algorithm to control the material deformation and failure process;

[0009] The system collects surface images and internal ultrasonic signals of materials, extracts the crack area ratio, number of fractures and maximum fracture length, and weights and fuses them to construct a critical failure criterion for materials, thereby monitoring and providing early warning of material failure risks.

[0010] The safety protection mechanism of the dynamic compression-shear testing machine is set. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, the test is automatically stopped, and the material test parameters and mechanical properties are recorded.

[0011] Based on the material testing parameters and mechanical property data recorded under failure conditions, the material parameter prediction model is incrementally learned and continuously optimized.

[0012] Preferably, data on tested materials and standard materials are collected, including material name, material type, geometric parameters, test parameters, and mechanical performance indicators. The test parameters include loading rate, compression-shear angle, and maximum load. The mechanical performance indicators include shear strength, fracture toughness, and impact toughness.

[0013] Based on the collected data of tested materials and standard materials, a material parameter prediction model is constructed using the support vector regression (SVR) algorithm.

[0014] The material parameter prediction model outputs the optimal testing parameters and expected mechanical properties of the material to be tested.

[0015] Preferably, real-time response data of the material is collected, including load, displacement, and strain;

[0016] The control parameters of the dynamic compression-shear testing machine, including loading rate, compression-shear angle, and maximum load, are corrected by a control algorithm; the proportional-integral-derivative (PID) control algorithm is used to dynamically adjust the control parameters.

[0017] Preferably, an adaptive control strategy is constructed to adjust the parameters of the PID controller online. An adaptive law is introduced to dynamically adjust the parameters of the PID controller based on the deviation between the material response and the expected response and the rate of change of the deviation.

[0018] Preferably, the material surface image sequence and internal ultrasonic signals are acquired;

[0019] Preprocess the image sequence of the material surface to extract the crack region and calculate the crack area ratio;

[0020] The internal ultrasonic signals are preprocessed and image-processed to extract the number of fractures and the maximum fracture length.

[0021] By weighted fusion of crack area ratio, fracture number, and maximum fracture length, a comprehensive criterion for material critical failure is obtained.

[0022] Preferably, a threshold for the crack area ratio, a threshold for the number of fractures, a threshold for the maximum fracture length, and a threshold for a comprehensive criterion are set;

[0023] When any one of the following data—crack area ratio, number of fractures, and maximum fracture length—exceeds the corresponding threshold for crack area ratio, number of fractures, and maximum fracture length, it is determined whether the material's exterior and interior have entered a critical failure state.

[0024] When the comprehensive criterion exceeds the comprehensive criterion threshold, it is determined that the material as a whole has reached a critical failure state, the test data is recorded, and a decision is made on whether to continue testing.

[0025] Preferably, after the material as a whole reaches the critical failure state, the surface images and internal ultrasonic signal data of the material are continuously collected to calculate the growth rate of crack area ratio, number of fractures and maximum fracture length per unit time.

[0026] Set limits for the crack area ratio, number of fractures, and maximum fracture length growth rate; continuously monitor the growth rate of material failure characteristic parameters; when the growth rate of any failure characteristic parameter exceeds the corresponding limit multiple times consecutively, trigger the safety protection mechanism, stop the dynamic compression-shear testing machine, and automatically record the test data.

[0027] Preferably, the detection data automatically recorded when the safety protection mechanism is triggered and the test is stopped includes: real-time response data of the material, including load, displacement and strain; mechanical performance indicators, including shear strength, fracture toughness and impact toughness at the time of shutdown; failure characteristic parameters, including crack area ratio, number of fractures and maximum fracture length at the time of shutdown.

[0028] Preferably, material testing parameters and mechanical property data recorded when the safety protection mechanism is triggered during the dynamic compression-shear test are collected to construct a material failure dataset, and the material failure dataset is merged with the original training dataset;

[0029] For the merged dataset, the material parameter prediction model is incrementally trained;

[0030] The incrementally learned material parameter prediction model is saved and deployed, and new material failure data is continuously collected. The incremental learning process is triggered periodically to update and continuously optimize the material parameter prediction model.

[0031] The adaptive control system for a dynamic compression-shear testing machine based on machine learning is used to implement the aforementioned adaptive control method for a dynamic compression-shear testing machine based on machine learning. It includes: a material parameter prediction module, a dynamic compression-shear control optimization module, a material critical failure detection module, a test safety protection module, and a material parameter model optimization module.

[0032] The material parameter prediction module acquires data of the material to be tested and constructs a material parameter prediction model based on machine learning technology.

[0033] The dynamic compression-shear control optimization module corrects the control parameters of the dynamic compression-shear testing machine through a control algorithm based on the real-time response data of the material during the dynamic compression-shear test.

[0034] The material critical failure detection module is used to acquire material surface images and internal ultrasonic signals, extract crack area ratio, fracture number and maximum fracture length, and weighted fuse them to construct material critical failure criteria, monitor and warn of material failure risks;

[0035] The test safety protection module is used to set the safety protection mechanism of the dynamic compression-shear tester. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, the test is automatically stopped, and the material test parameters and mechanical properties are recorded.

[0036] The material parameter model optimization module performs incremental learning and continuous optimization of the material parameter prediction model based on the material testing parameters and mechanical property data recorded under failure conditions.

[0037] The beneficial effects of this application are: This application constructs a material parameter prediction model through machine learning, which can accurately predict the optimal testing parameters and mechanical properties of the material to be tested, thereby improving the efficiency of the test and the reliability of the results.

[0038] This application adaptively adjusts the control parameters of the dynamic compression-shear testing machine based on real-time material response data, thereby achieving precise control and optimization of the testing process and improving the accuracy of the test results.

[0039] This application utilizes machine learning to construct a critical failure criterion for materials, and monitors the integrity of materials in real time through material surface images and internal ultrasonic signal data, accurately predicting the risk of failure.

[0040] This application establishes a safety protection mechanism for the dynamic compression-shear testing machine, which automatically stops the machine and records data when the material reaches a critical failure state, ensuring test safety and preventing the expansion of material damage.

[0041] This application utilizes detection parameters and mechanical property data under material failure conditions to continuously optimize the prediction model, thereby constantly improving the accuracy and reliability of material performance evaluation.

[0042] This application achieves accurate evaluation of material mechanical properties through measures such as material performance prediction, adaptive control of the test process, intelligent detection of material failure, safety protection mechanisms, and continuous model optimization. It improves the efficiency, accuracy, safety, and reliability of dynamic compression-shear tests, and promotes the development and innovation of dynamic mechanical property testing technology for materials. Attached Figure Description

[0043] Figure 1 Flowchart of the adaptive control method for dynamic compression-shear testing machine based on machine learning provided in this application;

[0044] Figure 2 Flowchart for constructing the material parameter prediction model provided in this application;

[0045] Figure 3 The flowchart for adaptive control parameter correction provided in this application;

[0046] Figure 4 The material critical failure detection flowchart provided for this application;

[0047] Figure 5 The flowchart for the safety protection of dynamic compression-shear test provided in this application;

[0048] Figure 6 The structure diagram of the adaptive control system of the dynamic compression-shear testing machine based on machine learning provided in this application. Detailed Implementation

[0049] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that excludes other embodiments.

[0052] Example 1:

[0053] Reference Figures 1 to 5 The first embodiment of this application provides an adaptive control method for a dynamic compression-shear testing machine based on machine learning.

[0054] Step 1: Obtain the data of the material to be tested, and build a material parameter prediction model based on machine learning technology. (See...) Figure 2 This is a flowchart of the material parameter prediction model construction process in this step.

[0055] Collect basic information on the tested and standard materials, including material name M and material type C; obtain the geometric dimensional parameters G of the tested and standard materials, including length l, width w, and thickness h, denoted as G = {l, w, h}; obtain the test parameters of the tested and standard materials in the dynamic compression-shear test, including loading rate v, compression-shear angle θ, and maximum load F. max And the corresponding mechanical properties, including shear strength τ and fracture toughness K. IC and impact toughness a k .

[0056] Based on the collected data of tested and standard materials, a material parameter prediction model is constructed using the Support Vector Regression (SVR) algorithm. The material name M, material type C, and geometric dimension parameter G are used as the input features of the model, denoted as x = [M, CG]. T The measured parameters v, θ, and F of the material in the dynamic compression-shear test. max and mechanical performance indicators τ, K IC a k As the output parameters of the model, denoted as y = [v, θ, F] max ,τ,K IC a k ] T .

[0057] Construct a training dataset D = {(x1, y1), (x2, y2), ..., (x...} N y N ), where N is the sample size, x N Let y be the input feature of the Nth sample. N The output parameters are for the Nth sample; the SVR model is trained using the training dataset D to obtain the material parameter prediction model for output detection parameters and mechanical properties.

[0058] The performance of the SVR model was evaluated on the test set, and the accuracy of the predictions was measured using the coefficient of determination and root mean square error. The model hyperparameters, such as the kernel type and penalty coefficient, were optimized through cross-validation and grid search to improve the model's generalization ability.

[0059] For the material to be tested, its material name is M. new Material type T new Geometric dimensional parameters G new Composition of feature vector x new =[M new T new G new ] T The parameters are input into the trained material parameter prediction model, which then outputs the optimal detection parameters for the material to be detected. and expected mechanical properties

[0060] After the material parameter prediction model outputs the optimal testing parameters and expected mechanical properties of the material to be tested, the user can use a dynamic compression-shear testing machine to test and verify the testing parameters and mechanical properties of the material: when the expected mechanical properties of the material are selected as the testing target, the dynamic compression-shear testing machine is used to test whether the material can complete the corresponding testing parameters; when the optimal testing parameters of the material are selected as the testing target, the dynamic compression-shear testing machine is used to test whether the material has achieved the expected mechanical properties.

[0061] The material parameter prediction model constructed in this step can quickly and accurately predict the optimal testing parameters and expected mechanical properties of the material to be tested. Compared with the traditional trial and error method, the material parameter prediction model constructed in this step improves the efficiency and accuracy of material testing.

[0062] Step 2: Based on the real-time response data of the material during the dynamic compression-shear test, the control parameters of the dynamic compression-shear testing machine are corrected through a control algorithm to control the material deformation and failure process. (See [link to relevant documentation]). Figure 3 This is a flowchart of the adaptive control parameter correction process in this step.

[0063] During the dynamic compression-shear test, real-time response data of the material is collected by sensors, including load F(t), displacement d(t), and strain ε(t). A force sensor measures the load F(t), a displacement sensor measures the displacement d(t), and a strain gauge measures the strain ε(t), denoted as s(t) = [F(t), d(t), ε(t)]. T , where t is time.

[0064] Based on the real-time response data s(t) of the material, the control parameters of the dynamic compression-shear testing machine are corrected through a control algorithm, including the loading rate v(t), compression-shear angle θ(t), and maximum load F. max (t); A proportional-integral-derivative (PID) control algorithm is adopted to dynamically adjust the control parameters according to the deviation between the material response and the expected response; the input of the PID controller is the deviation between the material response and the expected response, and the output is the correction amount of the control parameters Δu(t)=[Δv(t), Δθ(t), ΔF max (t)] T Where Δv(t) is the correction for the loading rate, Δθ(t) is the correction for the compression-shear angle, and ΔF max (t) represents the correction amount for the maximum load;

[0065] The mathematical expression for a PID controller is:

[0066]

[0067] Among them, Kp K i K d These are the proportional, integral, and derivative coefficients, respectively, which are adjusted using an adaptive law. e(t) is the control error. This indicates that the error e(t) is integrated over time, reflecting the accumulation of the error over time. This represents the derivative of the error e(t) with respect to time t.

[0068] The control parameters of the dynamic compression-shear testing machine are corrected based on the output of the PID controller.

[0069] v(t)=v0+Δv(t), θ(t)=θ0+Δθ(t), F max (t)=F max，0 +ΔF max (t);

[0070] Where, v0, θ0, F max，0 The initial control parameters are the initial loading rate, initial compression-shear angle, and initial maximum load, which are given by the material parameter prediction model.

[0071] To adapt to changes in material properties and disturbances in experimental conditions, an adaptive control strategy is adopted to adjust the parameters of the PID controller online. An adaptive law is introduced, based on the deviation e(t) between the material response and the expected response and the rate of change of the deviation. Dynamically adjust the parameter K of the PID controller p K i and K d The mathematical expression for the adaptive law is:

[0072]

[0073] Where, γ p γ i γ d The learning rate for the adaptive law needs to be selected based on the material properties and experimental requirements.

[0074] For example, the learning rate γ of the adaptive law p γ i and γ d The selection steps are as follows:

[0075] Based on the nonlinear characteristics of the material, a smaller γ is selected. p Values, such as γ p =0.1; because the response of a nonlinear material is not linear with respect to the load, an excessively large γ p The value may cause the control parameters to change too quickly, resulting in fluctuations in the load curve;

[0076] Based on the anisotropy of the material, a larger γ value is selected. i Values, such as γ i =0.5; because anisotropic materials have significantly different mechanical properties in different directions, a larger γ is required. i The values ​​are adapted to accommodate this difference, ensuring precise control in all directions;

[0077] Based on the strain rate sensitivity of the material, a larger γ is selected. d Values, such as γ d =0.3; because the mechanical properties of strain rate-sensitive materials change with strain rate, a larger γ d Values ​​can help control systems respond quickly to changes in strain rate and prevent fluctuations in the load curve.

[0078] Conduct dynamic compression-shear tests and observe whether the material response meets the test requirements; if the load curve shows significant fluctuations, the γ value can be appropriately reduced. p and γ d The value of γ; if the test is not stopped in time before the material fails, the value of γ can be appropriately increased. i The value of γ is determined; based on the experimental results, the learning rate of the adaptive law is fine-tuned until the optimal value suitable for the material and experimental conditions is found; the final learning rate γ is recorded. p γ i γ d , which is the optimal learning rate for the material under the experimental conditions.

[0079] Through adaptive control strategies, the parameters of the PID controller can be adjusted in real time to adapt to changes in material properties and disturbances in test conditions, thereby improving the control accuracy and robustness of dynamic compression-shear tests.

[0080] The PID control algorithm and adaptive law are integrated into the control system of the dynamic compression-shear testing machine to achieve real-time data acquisition, control parameter correction and adaptive control.

[0081] This step enables the dynamic compression-shear testing machine to adaptively adjust control parameters based on the real-time response data of the material, thereby achieving precise control over the material deformation and failure process. At the same time, the introduction of the adaptive control strategy allows the dynamic compression-shear testing machine to adapt to different materials and test conditions, exhibiting better robustness and versatility.

[0082] Step 3: Acquire surface images and internal ultrasonic signals of the material, extract crack area ratio, number of fractures, and maximum fracture length, and weightedly fuse these to construct a critical failure criterion for the material, monitoring and providing early warning of material failure risks. (See [link to relevant documentation]). Figure 4 This is a flowchart of the material critical failure detection process in this step.

[0083] During the dynamic compression-shear test, surface image data and internal ultrasonic signal data of the material were acquired simultaneously; a high-speed camera was used to capture the deformation and damage process of the material surface, and the recorded material surface image sequence was... Where N1 is the number of image frames. This represents the surface image of the material in frame N1; acoustic signals from inside the material were acquired using an ultrasonic probe array; the recorded internal ultrasonic signals are... Where N2 is the number of sampling points. This represents the internal ultrasonic signal at the N2th sampling point.

[0084] The acquired material surface image sequence I is preprocessed, including denoising, enhancement, and segmentation operations, to extract the image region of the crack on the material surface. An edge detection image segmentation algorithm is then used to extract the crack region, resulting in a binary mask image M of the crack. I Based on the binary mask image M of the crack I Calculate the pixel area A of the crack. I And compare it with the current cross-sectional area A0 of the material to obtain the crack area ratio B. I :

[0085] Set the crack area ratio threshold λ I When B I =λ I When the material reaches a critical failure state, the threshold λ is used to determine that the material's external structure has reached that state. I The parameters are preset based on factors such as the type and geometry of the material. For example, for a certain brittle material, λ can be set in advance. I =5%, while for tough materials, λ can be appropriately increased. I value.

[0086] The acquired ultrasonic signals R from within the material are preprocessed, including filtering, amplification, and calibration, to extract the ultrasonic echo signals from the internal fractures. Ultrasonic imaging algorithms are then used to process the echo signals to obtain the ultrasonic images S from within the material. R ; for ultrasound images S R Image segmentation and feature extraction are performed to obtain the number N of fractures inside the material. R and maximum fracture length L R .

[0087] Set the fracture number threshold λ N and the maximum fracture length threshold λ L When N R ≥λ N or L R ≥λ L When the material reaches a critical failure state, the threshold λ is used to determine that the material has reached that state. N and λ LThe parameters are preset based on factors such as the type and geometry of the material. For example, for composite materials, λ can be set. N =10, λ L =5mm.

[0088] The proportion of surface crack area B of the material I and internal fracture characteristics N R L R A fusion method was used to construct a criterion for judging the overall critical failure of the material. A weighted fusion method was employed to sum the judgment results of surface cracks and internal fractures, resulting in a comprehensive criterion J for the overall critical failure of the material.

[0089]

[0090] in, This is an indicator function; it takes the value 1 if the condition is true, and 0 otherwise. I w N w L The weighting coefficients for the surface crack area ratio, the number of internal fractures, and the maximum fracture length, satisfying w I +w N +w L =1, which can be set according to the failure mechanism of the material and experimental experience; for example, for brittle materials, since the damage of brittle materials mainly occurs on the surface, w should be set. I ≥w N +w L For tough materials, since damage mainly occurs internally, a w value should be set. N ≥w L >w I .

[0091] Set the comprehensive criterion threshold λ J When J≥λ J When the material as a whole reaches a critical failure state, the test is automatically stopped and the test data is recorded; the comprehensive judgment threshold λ is used. J The value range of λ is (0, 1], and can be set according to the failure mechanism of the material and experimental experience. For example, λ can be chosen as the value of λ. J =0.6.

[0092] When it is determined that the material as a whole has reached a critical failure state, the test data is recorded and the test personnel decide whether to continue the test.

[0093] This step constructs a critical failure criterion for the material as a whole by using material surface images and internal ultrasonic signals, enabling real-time monitoring and early warning of the critical failure state during the dynamic compression and shearing process of the material.

[0094] This approach comprehensively considers two failure mechanisms: surface crack propagation and internal fracture evolution. It extracts key features of material failure through image segmentation and ultrasonic imaging algorithms, and uses a weighted fusion method to integrate and judge the collected multimodal information. This overcomes the limitations of a single failure criterion and improves the accuracy and reliability of material critical failure detection.

[0095] This step also sets reasonable thresholds and weighting coefficients, and combines the failure mechanism of materials and experimental experience to make the solution highly applicable and practical. It can provide effective failure monitoring and control methods for dynamic compression-shear tests, and improve the safety and efficiency of the test.

[0096] Step 4: Set the safety protection mechanism of the dynamic compression-shear testing machine. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, automatically stopping the test and recording the material test parameters and mechanical properties. See [link / reference] Figure 5 This is a flowchart of the safety protection process for the dynamic compression-shear test in this step.

[0097] After the material reaches a critical failure state, surface images and internal ultrasonic signal data are continuously acquired, and the growth rate of each failure characteristic parameter per unit time is calculated:

[0098] Crack area growth rate v B (t i ): B c (t i ) indicates at time point t i Crack area at time B; c (t i-1 ) represents time point t i Crack area at time -1;

[0099] Fracture quantity growth rate v R (t i ): N R (t i ) indicates at time point t i The number of cracks at time N R (t i-1 ) represents time point t i The number of cracks at time -1;

[0100] Maximum fracture length growth rate v L (t i ): L max (t i ) indicates at time point t i The maximum crack length at time L max (ti-1 ) indicates at time point t i The maximum crack length at time -1; t i Let t be the time of the i-th data acquisition, and Δt be the time interval between two consecutive data acquisitions.

[0101] Limits are set for the crack area ratio, the number of fractures, and the maximum fracture length growth rate, respectively, V. B，lim v R，lim and V L，lim .

[0102] Continuously monitor the growth rate of material failure characteristic parameters to determine whether the corresponding limit has been exceeded multiple times consecutively: if the crack area ratio growth rate is continuously n B This exceeds the limit once, that is:

[0103] v B (t i )>V B，lim i = k, k+1, ..., k+n B -1;

[0104] This triggers the security protection mechanism, where k is the starting sequence number, indicating that starting from the k-th measurement time, the subsequent n measurements are continuously counted. B Second-rate.

[0105] If the rate of increase in the number of fractures is continuous n R This exceeds the limit once, that is:

[0106] v R (t i )>V R，lim i = k, k+1, ..., k+n R -1;

[0107] This will trigger the security protection mechanism.

[0108] If the maximum fracture length growth rate is continuously n L This exceeds the limit once, that is:

[0109] v L (t i )>V L，lim i = k, k+1, ..., k+n L -1;

[0110] This will trigger the security protection mechanism.

[0111] Number of consecutive exceedances n B n R and n L It can be set according to the material properties and test requirements, and is generally between 3 and 10.

[0112] When any safety protection mechanism is met, the shutdown procedure of the dynamic compression-shear testing machine is immediately initiated to cut off the load and stop the test, preventing damage to the test site and injury to the test personnel.

[0113] When the safety protection mechanism is triggered and the test is stopped, the following data are automatically recorded: real-time response data of the material, including load, displacement and strain; mechanical performance indicators, including shear strength, fracture toughness and impact toughness at the time of shutdown; failure characteristic parameters, including crack area ratio, number of fractures and maximum fracture length at the time of shutdown; and macroscopic failure mode of the material, which is captured by a high-definition camera at the state of the material at the time of shutdown.

[0114] The safety protection mechanism established in this step enables the dynamic compression-shear testing machine to promptly determine the integrity and load-bearing capacity of the material based on the fact that the material failure characteristic parameters exceed the limit multiple times in a row, and to initiate the shutdown procedure, effectively preventing catastrophic damage to the material and ensuring the safety and reliability of the testing process.

[0115] The technical solution in this step allows testers to explore the upper limit of material testing while ensuring safety. By recording in detail the various parameters and performance indicators when the material fails, it provides data support for subsequent analysis of the material's failure mechanism and improvement of the material's dynamic mechanical properties.

[0116] Step 5: Based on the material testing parameters and mechanical property data recorded under failure conditions, perform incremental learning and continuous optimization of the material parameter prediction model.

[0117] We collected and organized the material testing parameters and corresponding material mechanical property data recorded when the safety protection mechanism was triggered during the dynamic compression-shear test, and constructed a material failure dataset.

[0118] Material testing parameters include load, displacement, strain, crack area ratio, number of fractures, and maximum fracture length. Mechanical performance indicators include shear strength, fracture toughness, and impact toughness. The material failure dataset is preprocessed to remove outliers and ensure the accuracy and consistency of the data.

[0119] The established material failure dataset is merged with the original dataset used to train the material parameter prediction model. The merged dataset is randomly shuffled to ensure that the old and new data are fully mixed and to avoid the impact of abrupt changes in data distribution on model performance. Depending on the size of the dataset and the limitations of computing resources, a batch incremental learning strategy can be adopted, merging only a portion of the new failure data each time.

[0120] For the merged dataset, the hyperparameters of the material parameter prediction model are fine-tuned. For the support vector regression model, the type and parameters of the kernel function can be adjusted, such as the bandwidth of the Gaussian kernel and the degree of the polynomial kernel. The regularization parameters of the support vector regression model are optimized to balance the model's fitting ability and generalization ability. Grid search and random search methods are used to find the optimal parameter combination within the range of hyperparameter values.

[0121] Using the merged dataset, the material parameter prediction model is incrementally trained, and optimization algorithms such as mini-batch gradient descent are used to adjust the model parameters so that it achieves optimal fit on the new data distribution.

[0122] The optimal material parameter prediction model after incremental learning is saved as a new model version. This new model version is then deployed to the control system of the dynamic compression-shear testing machine, replacing the original model. In subsequent dynamic compression-shear tests, the updated material parameter prediction model is used to predict the material's test parameters and mechanical properties in real time. New material failure data is continuously collected, and the incremental learning process is triggered periodically to achieve adaptive updates and continuous optimization of the model.

[0123] This step utilizes newly collected material failure data from dynamic compression-shear tests to incrementally learn and optimize the material parameter prediction model. By merging the new data into the original training dataset, the model's hyperparameters are fine-tuned, and incremental training is performed on the merged dataset. This allows the model to adapt to constantly changing data distributions and improves prediction performance. This step also ensures that the material parameter prediction model can continuously learn and evolve through regular incremental learning and model updates, providing more accurate and reliable support for material performance evaluation and process optimization.

[0124] Example 2:

[0125] Reference Figure 6 The second embodiment of this application provides an adaptive control system for a dynamic compression-shear testing machine based on machine learning.

[0126] The system includes: a material parameter prediction module, a dynamic compression-shear control optimization module, a material critical failure detection module, an experimental safety protection module, and a material parameter model optimization module.

[0127] The material parameter prediction module acquires data on the material to be tested and constructs a material parameter prediction model based on machine learning technology.

[0128] The dynamic compression-shear control optimization module corrects the control parameters of the dynamic compression-shear testing machine through a control algorithm based on the real-time response data of the material during the dynamic compression-shear test.

[0129] The material critical failure detection module is used to acquire material surface images and internal ultrasonic signals, extract crack area ratio, fracture number and maximum fracture length, and weighted fuse them to construct material critical failure criteria, thereby monitoring and warning of material failure risks.

[0130] The test safety protection module is used to set the safety protection mechanism of the dynamic compression-shear tester. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, the test is automatically stopped, and the material test parameters and mechanical properties are recorded.

[0131] The material parameter model optimization module performs incremental learning and continuous optimization of the material parameter prediction model based on the material testing parameters and mechanical property data recorded under failure conditions.

[0132] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings or direct couplings or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0133] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of this application without departing from the spirit and scope of protection of the claims. All of these variations are within the protection scope of this application.

Claims

1. An adaptive control method for a dynamic compression-shear testing machine based on machine learning, characterized in that, include: Acquire data on the material to be tested and construct a material parameter prediction model based on machine learning technology; Based on the real-time response data of the material during the dynamic compression-shear test, the control parameters of the dynamic compression-shear testing machine are corrected by the control algorithm to control the material deformation and failure process; The system collects surface images and internal ultrasonic signals of materials, extracts the crack area ratio, number of fractures and maximum fracture length, and weights and fuses them to construct a critical failure criterion for materials, thereby monitoring and providing early warning of material failure risks. Acquire material surface image sequences and internal ultrasonic signals, preprocess the material surface image sequences to extract crack regions, and calculate the crack area ratio; The internal ultrasonic signals are preprocessed and image-processed to extract the number of fractures and the maximum fracture length. A comprehensive criterion for material critical failure is obtained by weighted fusion of crack area ratio, number of fractures, and maximum fracture length. The safety protection mechanism of the dynamic compression-shear testing machine is set. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, the test is automatically stopped, and the material test parameters and mechanical properties are recorded. Based on the material testing parameters and mechanical property data recorded under failure conditions, the material parameter prediction model is incrementally learned and continuously optimized. Material testing parameters and mechanical property data recorded during dynamic compression-shear tests when the safety protection mechanism is triggered are collected to construct a material failure dataset, which is then merged with the original training dataset. The material parameter prediction model is then incrementally trained on the merged dataset. The incrementally learned material parameter prediction model is saved and deployed, and new material failure data is continuously collected. The incremental learning process is triggered periodically to update and continuously optimize the material parameter prediction model.

2. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 1, characterized in that, Collect data on tested materials and standard materials, including material name, material type, geometric parameters, test parameters, and mechanical property indicators. The test parameters include loading rate, compression-shear angle, and maximum load. The mechanical property indicators include shear strength, fracture toughness, and impact toughness. Based on the collected data of tested materials and standard materials, a material parameter prediction model is constructed using the support vector regression (SVR) algorithm. The material parameter prediction model outputs the optimal testing parameters and expected mechanical properties of the material to be tested.

3. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 2, characterized in that, Collect real-time response data of the material, including load, displacement, and strain; The control parameters of the dynamic compression-shear testing machine, including loading rate, compression-shear angle, and maximum load, are corrected by a control algorithm; the proportional-integral-derivative (PID) control algorithm is used to dynamically adjust the control parameters.

4. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 3, characterized in that, An adaptive control strategy is constructed to adjust the parameters of the PID controller online. An adaptive law is introduced to dynamically adjust the parameters of the PID controller based on the deviation between the material response and the expected response and the rate of change of the deviation.

5. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 4, characterized in that, Set thresholds for crack area ratio, number of fractures, and maximum fracture length, as well as a comprehensive criterion threshold; When any one of the following data—crack area ratio, number of fractures, and maximum fracture length—exceeds the corresponding threshold for crack area ratio, number of fractures, and maximum fracture length, it is determined whether the material's exterior and interior have entered a critical failure state. When the comprehensive criterion exceeds the comprehensive criterion threshold, it is determined that the material as a whole has reached a critical failure state, the test data is recorded, and a decision is made on whether to continue testing.

6. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 5, characterized in that, After the material as a whole reaches the critical failure state, the surface images and internal ultrasonic signal data of the material are continuously collected to calculate the growth rate of crack area ratio, number of fractures and maximum fracture length per unit time. Set limits for the crack area ratio, number of fractures, and maximum fracture length growth rate; continuously monitor the growth rate of material failure characteristic parameters; when the growth rate of any failure characteristic parameter exceeds the corresponding limit multiple times consecutively, trigger the safety protection mechanism, stop the dynamic compression-shear testing machine, and automatically record the test data.

7. The adaptive control method for a dynamic compression-shear testing machine based on machine learning according to claim 6, characterized in that, The detection data automatically recorded when the safety protection mechanism is triggered and the test is stopped includes: real-time response data of the material, including load, displacement and strain; mechanical performance indicators, including shear strength, fracture toughness and impact toughness at the time of shutdown; and failure characteristic parameters, including the crack area ratio, number of fractures and maximum fracture length at the time of shutdown.

8. A machine learning-based adaptive control system for a dynamic compression-shear testing machine, used to implement the machine learning-based adaptive control method for a dynamic compression-shear testing machine as described in any one of claims 1 to 7, characterized in that, include: The module includes a material parameter prediction module, a dynamic compression-shear control optimization module, a material critical failure detection module, an experimental safety protection module, and a material parameter model optimization module. The material parameter prediction module acquires data of the material to be tested and constructs a material parameter prediction model based on machine learning technology. The dynamic compression-shear control optimization module corrects the control parameters of the dynamic compression-shear testing machine through a control algorithm based on the real-time response data of the material during the dynamic compression-shear test. The material critical failure detection module is used to acquire material surface images and internal ultrasonic signals, extract crack area ratio, fracture number and maximum fracture length, and weighted fuse them to construct material critical failure criteria, monitor and warn of material failure risks; The test safety protection module is used to set the safety protection mechanism of the dynamic compression-shear tester. When the material failure characteristic parameters reach the preset threshold, the safety protection mechanism is triggered, the test is automatically stopped, and the material test parameters and mechanical properties are recorded. The material parameter model optimization module performs incremental learning and continuous optimization of the material parameter prediction model based on the material testing parameters and mechanical property data recorded under failure conditions.

Images