Self-sensing piezoresistive signal prediction and mix proportion design method and system based on multi-scale mix proportion feature embedding

Abstract

The invention discloses a self-sensing piezoresistive signal prediction and mix proportion design method and system based on multi-scale mix proportion feature embedding, and belongs to the technical field of self-sensing concrete sensor monitoring. The method comprises the following steps: 1, constructing a stress-resistance change rate data set of the self-sensing concrete sensor; 2, obtaining a self-sensing piezoresistive signal prediction model based on mix proportion multi-scale feature embedding by using the data set in the step 1; and 3, based on the prediction model in the step 2, realizing the multi-task optimization self-sensing concrete sensor mix proportion design method based on the non-dominated sorting genetic algorithm. The method is used for solving the problems that the scale of a data set is small, system modeling of a piezoresistive performance dynamic response rule under the action of an external load is lacked, and a network architecture specially designed for SSCC mix proportion characteristics is lacked.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of self-sensing concrete sensor monitoring, and particularly relates to a self-sensing piezoresistive signal prediction and mix proportion design method and system based on multi-scale mix proportion feature embedding. BACKGROUND

[0002] Self-sensing cementitious composites (SSCC) is a kind of intelligent material which has stress-resistance response characteristics by introducing conductive materials into cement-based materials, so that the resistance changes under the action of stress. SSCC not only has good piezoresistive effect, but also can take into account certain mechanical properties, so it can be used as an embedded sensor in structural health monitoring (SHM) to realize real-time monitoring of structural stress, strain and other state parameters. Its working principle is mainly based on the piezoresistive effect, that is, by monitoring the change of resistivity of the material under external load, the mapping relationship between mechanical response and electrical signal is realized.

[0003] At present, the research on SSCC mainly focuses on the influence of different types of conductive materials (such as carbon black, carbon nanotubes, steel fibers, etc.) and their dosage on the piezoresistive performance. A large number of experiments show that the introduction of an appropriate amount of conductive material can significantly improve the self-sensing performance of SSCC, but it may also have certain adverse effects on the mechanical properties of the material, and the increase of the dosage will significantly increase the cost of the material. Therefore, under the premise of ensuring good piezoresistive performance, how to balance the mechanical properties and economy of the material is an important problem that needs to be solved in the design of SSCC mix proportion. In order to reveal the relationship between the dosage of conductive material and the performance, some studies have established a SSCC piezoresistive performance prediction model using traditional methods such as polynomial regression. However, such models have the problems of insufficient generality, limited scope of application, and limited prediction accuracy. Due to the influence of factors such as material dispersion and interface effect, the performance test results under the same mix proportion conditions often have large dispersion. In addition, existing researches mostly optimize parameters based on small sample experiments, and the sample size is usually only dozens to hundreds, which is difficult to fully cover the complex and nonlinear relationship between different types of conductive materials, dosage ratio and mechanical response, seriously restricting the application ability of the model in material design and performance prediction.

[0004] With the rapid development of machine learning technology, data-driven material performance prediction methods have shown good application potential in the field of material science. Through the construction of complex nonlinear mapping relationship, machine learning can provide a new solution for SSCC performance prediction. However, the existing SSCC performance modeling methods based on machine learning still have the following shortcomings: (1) The existing data set is small in size and cannot comprehensively cover the complex relationship between various conductive material types, mixing proportion and piezoresistive performance, resulting in limited model generalization ability and prediction accuracy; (2) Most studies only focus on the prediction of the static performance (such as compressive strength, resistivity, etc.) of SSCC, and lack of systematic modeling of the dynamic response law of piezoresistive performance under external load; (3) The existing models mostly use general machine learning algorithms, lack of network architecture specially designed for SSCC mix proportion characteristics, cannot effectively process redundant information in input features and realize efficient feature fusion, limiting the learning ability and prediction effect of complex nonlinear relationship.

[0005] In summary, the existing SSCC piezoresistive performance prediction and mix proportion design method generally has the following problems: lack of modeling ability of multi-conductive material synergistic piezoresistive effect, limited data set size, lack of learning of dynamic piezoresistive response law, and insufficient consideration of the stability of SSCC mix proportion characteristics and the modeling demand of complex nonlinear relationship. Therefore, it is urgent to develop an efficient machine learning method to solve the above problems, and combine multi-objective optimization algorithm to realize the balanced optimization design of self-sensing concrete piezoresistive performance, mechanical performance and economic cost, to meet the engineering practical application requirements. SUMMARY

[0006] The present application provides a self-sensing piezoresistive signal prediction and mix proportion design method based on multi-scale mix proportion feature embedding, to solve the problems in the prior art.

[0007] The present application provides a self-sensing piezoresistive signal prediction and mix proportion design system based on multi-scale mix proportion feature embedding, to realize a self-sensing piezoresistive signal prediction and mix proportion design method based on multi-scale mix proportion feature embedding.

[0008] The present application is realized by the following technical solutions: A self-sensing piezoresistive signal prediction and mix proportion design method based on multi-scale mix proportion feature embedding, the method comprising the following steps: Step 1: Constructing a self-sensing concrete sensor stress-resistance change rate data set; Step 2: Using the data set of step 1, obtaining a self-sensing piezoresistive signal prediction model based on mix proportion multi-scale feature embedding; Step 3: Based on the prediction model of step 2, realizing a multi-task optimization self-sensing concrete sensor mix proportion design method based on non-dominated sorting genetic algorithm.

[0009] Furthermore, step one specifically involves introducing measured piezoresistive signals based on 16 groups of self-sensing cement-based sensors with different additions of carbon black, carbon nanotubes, and carbon fibers. All the ratios involved are based on a fixed matrix material composition, cured for 28 days at 20±2°C and 90% RH, and the specimen size is 40×40×40 mm. The data set contains a total of 5,300 groups of stress-resistance change rate data pairs.

[0010] Furthermore, the prediction model in step 2 includes a neural network that embeds multi-scale mix ratio features in multiple hidden layers; the piezoresistive pattern is learned in a high-dimensional feature space through a supervised learning method to achieve universal modeling of the piezoresistive response law of the self-sensing sensor, that is, a pre-trained multi-scale feature embedded piezoresistive signal prediction network model; when the network model training is completed, it can be directly applied to the self-sensing sensor embedded in the structure to achieve real-time monitoring of the structural stress state.

[0011] Furthermore, the step 2 specifically includes the following steps: Step 21: Build a multi-scale feature embedding neural network, and take the content of conductive material, water reducer content and resistance change rate as input, and output the corresponding stress; Step 22: Based on the network in step 21, set the optimal hyperparameters obtained using the Bayesian optimization method; Step 23: Use the optimal hyperparameters obtained in step 22 to train the self-sensing piezoresistive signal prediction model in a supervised learning framework.

[0012] Furthermore, the step three is specifically to take strength and economy as dual fitness functions, set constraints, comprehensively consider performance and cost factors, and the optimization goal is to maximize strength and minimize economy at the same time, wherein the strength is calculated using a pre-trained multi-scale feature embedding piezoresistive signal prediction model, and the economy index is 1m3 per cubic meter. 3 The cost of self-sensing composite materials is used as the measurement standard; through a multi-task optimization algorithm, a balanced design between piezoresistive performance, mechanical properties and cost is achieved, and the optimal self-sensing concrete sensor mix ratio is output.

[0013] Furthermore, the step three specifically includes the following steps: Step 31: Construct fitness function; Step 32: Use the fitness function of step 31 to introduce the non-dominated sorting genetic algorithm.

[0014] Furthermore, the non-dominated sorting genetic algorithm realizes multi-objective optimization and reversely deduce the initial concrete mix ratio based on the existing objectives.

[0015] A self-sensing piezoresistive signal prediction and mix proportion design system based on multi-scale mix proportion feature embedding, which uses the self-sensing piezoresistive signal prediction and mix proportion design method based on multi-scale mix proportion feature embedding as described above, and the system comprises: A data set construction module: construct a self-sensing concrete sensor stress-resistance rate of change data set; A prediction model construction module: use the data set of step one to obtain a self-sensing piezoresistive signal prediction model based on multi-scale mix proportion feature embedding; A prediction model using module: based on the prediction model of step two, realize a multi-task optimization self-sensing concrete sensor mix proportion design method based on non-dominated sorting genetic algorithm.

[0016] A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method as described above when executing the computer program.

[0017] A computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method as described above.

[0018] The beneficial effects of the present application are: The present application first proposes a multi-scale embedding method of mix proportion features in the piezoresistive signal prediction model, which can learn the feature mapping relationship of the stress-resistance rate of change signal in a supervised learning manner, significantly enhances the adaptability of the model to the piezoresistive signal under different mix proportions, realizes the universal prediction of the self-sensing concrete sensor under various conductive materials and different mix proportions, and solves the problem that the existing method cannot effectively process multi-mix proportion data.

[0019] Compared with the traditional fully connected network model, the multi-scale splicing neural network architecture proposed in the present application can repeatedly introduce mix proportion features in the hidden layer, fully excavate the nonlinear relationship between the conductive material information and the piezoresistive signal, and significantly improve the prediction accuracy and generalization ability of the model.

[0020] Through the piezoresistive signal prediction and mix proportion design integrated framework proposed in the present application, theoretical and technical support can be provided at two levels of sensor piezoresistive performance prediction and material design, and a more efficient and intelligent solution is provided for stress state monitoring and safety evaluation in complex engineering structures.

[0021] The self-sensing piezoresistive signal prediction and mix proportion design method based on multi-scale mix proportion feature embedding provides a high-precision, high-efficiency and strong-adaptability technical means for the field of structural health monitoring, and promotes the application and development of self-sensing intelligent materials in the fields of civil engineering, transportation engineering and high-end equipment. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is a method flowchart of the present application.

[0023] Figure 2 is a stress-resistance rate data pair example of the present application.

[0024] Figure 3 is a self-perception piezoresistive signal prediction training framework based on matching ratio multi-scale feature embedding of the present application.

[0025] Figure 4 is a non-dominated sorting genetic algorithm schematic diagram of the present application.

[0026] Figure 5 is a general piezoresistive signal prediction result of the self-perception concrete sensor. DETAILED DESCRIPTION

[0027] In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular sequences of steps, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, and circuits are omitted so as not to obscure the description of the present application with unnecessary detail.

[0028] It should be understood that the term "comprises" when used in this specification and the appended claims, specifies the presence of stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0029] It should also be understood that the terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0030] The technical solutions in the embodiments of the present application are clearly and completely described below with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0031] Many specific details are set forth in the following description in order to provide a thorough understanding of the present application. However, the present application can be practiced according to other embodiments that can not be described in detail herein, and the present application is not limited to the embodiments described herein. It will be appreciated that the present application can be practiced in a variety of ways, and that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting as to the scope of the present application.

[0032] Embodiment one A self-sensing piezoresistive signal prediction and mix design method based on multi-scale mix proportion feature embedding, as shown in Figure 1 The method comprises the following steps: Step one: self-sensing concrete sensor stress-resistance change rate dataset construction.

[0033] At present, the piezoresistive performance signal of the self-sensing sensor can be collected in large quantities in the test process, but due to the differences in the types, contents, physical characteristics, preparation methods and other factors of the conductive materials in the mix proportion, these data cannot play an effective role in the task of establishing the piezoresistive signal prediction model. In the present application, a self-sensing concrete sensor stress-resistance change rate dataset of the synergistic effect of multiple conductive materials is constructed. The present application introduces the measured piezoresistive signal of the self-sensing cement-based sensor based on 16 groups of different content of carbon black, carbon nanotubes and carbon fibers, all the mix proportions involved are fixed without changing the composition of the matrix material, cured under the condition of 20±2°C and 90%RH for 28 days, and the size of the test piece is 40×40×40mm. The dataset has 5300 groups of stress-resistance change rate data pairs. Effect: This way of constructing the dataset avoids the behavior of consuming a large amount of expert knowledge and time for self-sensing mix proportion classification, only processes the information contained in the data at the data level, and can quickly and effectively obtain the supervised learning dataset.

[0034] Step two: self-sensing piezoresistive signal prediction method based on multi-scale feature embedding of mix proportion.

[0035] Based on the sensor stress-resistance change rate dataset constructed in step one, the present application proposes a pre-trained self-sensing sensor general piezoresistive signal prediction method based on multi-scale feature embedding of mix proportion. The method contains a neural network embedding multi-scale mix proportion features in multiple hidden layers, aiming to highlight the importance of conductive material information in the mix proportion to the piezoresistive signal, and realize the effective fusion and representation of multi-scale features. Through the supervised learning method, the piezoresistive mode is learned in the high-dimensional feature space, realizing the general modeling of the piezoresistive response law of the self-sensing sensor. After the model training is completed, it can be directly applied to the self-sensing sensor embedded in the structure, realizing the real-time response of the stress state of the structure.

[0036] Step three: multi-task optimization self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm.

[0037] Based on the pre-training multi-scale feature embedding piezoresistive signal prediction network proposed in steps one and two, this step sets the content of conductive material, the maximum piezoresistive resistance change rate and other constraint conditions as the dual fitness function, comprehensively considers the performance and cost factors, and heuristically optimizes to generate a mix proportion scheme that meets the engineering requirements. The optimization goal of the algorithm is to maximize the strength and minimize the economy at the same time, wherein the strength is calculated by the pre-training multi-scale feature embedding piezoresistive signal prediction model which has been proved to have excellent generalization ability in step two, and the economy index takes the self-perception composite material cost of 1m 3

[0038] The flowchart of the self-perception piezoresistive signal prediction and mix proportion design architecture based on the mix proportion multi-scale feature embedding proposed by the application is shown in Figure 1

[0039] Step one: self-perception concrete sensor stress-resistance change rate data set construction. The application proposes a data set construction method of sampling and aligning stress and resistance change rate signals under different sampling frequencies to solve the problem of inconsistent sampling frequencies and signals that cannot be directly aligned in the experimental data of self-perception concrete sensors, and to ensure the effectiveness and consistency of the data required for subsequent model training. Specifically, the following steps are included: Step one: resistance change rate conversion of measured resistance signal. In the measurement process of structural embedded sensors, a multimeter or data acquisition instrument is usually used to measure the resistance of the sensor. However, due to the large difference in resistance value under different ranges and test conditions, it is difficult to directly characterize the structural stress state (such as stress, strain). Therefore, the application uses resistance change rate as the key feature of piezoresistive effect to reflect the structural health state, as shown in equation 1.

[0040] (1) Wherein, R t is the resistance value at the current time, R 0 is the initial resistance value. By this way, the influence of the absolute resistance value difference can be effectively eliminated, making the signals under different experimental conditions comparable, and providing standardized feature input for subsequent model training.

[0041] Step two: sampling processing of resistance change rate data. Since the data collection frequency of the actual test sensor is inconsistent with the loading sampling frequency and strain collection frequency, time series alignment processing is needed, as shown in equations 2 and 3.

[0042] ​​ (2) (3) To realize the synchronization of resistance change rate and stress sequence, the application adopts an interpolation function to perform time alignment of resistance change rate data. Due to the short sampling interval of experimental data, linear interpolation can effectively avoid errors caused by complex fitting, and ensure the rationality of data mapping. Finally, the data pair of stress-resistance change rate is obtained, and the data format is shown in Figure 2

[0043] Step two: self-perception piezoresistive signal prediction method based on multi-scale feature embedding of mix proportion. In order to establish a general piezoresistive signal prediction model of self-perception concrete sensor, the application proposes a neural network architecture which can accurately extract stress-resistance change rate signal features and fuse mix proportion information. The input features of stress-resistance change rate signal mapping are special. Since the content of conductive materials and water reducing agent under each mix proportion is fixed, inputting ordinary fully connected network will cause information redundancy, which will weaken the learning ability of the model to some extent. Based on the above consideration, the present study proposes a multi-scale feature embedding neural network architecture, which realizes efficient use of mix proportion information by embedding mix proportion features in each hidden layer, thereby enhancing the learning ability and generalization performance of the model. The architecture diagram is shown in Figure 3 .

[0044] Specifically includes the following steps: Step two one: building of multi-scale feature embedding neural network. The application selects the most important related features of self-perception concrete sensor in the piezoresistive monitoring process as the input of the network, which are the content of conductive materials (the content of carbon black, carbon nanotube and carbon fiber in 1m 3 ), the content of water reducing agent and the resistance change rate, and the output is the corresponding stress. Among them, the mix proportion information is embedded in each hidden layer in the form of auxiliary features, rather than being directly transmitted into the network as ordinary input, so as to avoid the interference of redundant information and improve the learning ability of the network to the stress-resistance change rate relationship. This feature fusion method can dynamically emphasize the influence of mix proportion features on the prediction results in the high-dimensional space of hidden layer, as shown in equation 4.

[0045] (4) Step two two: optimal hyperparameter setting of Bayesian optimization method. In order to ensure that the network proposed in the application can achieve the best training effect, the Bayesian optimization method is used to optimize the hyperparameters such as the number of layers, the number of hidden layer nodes and the learning rate of the network. Through iterative search, the network hyperparameter combination suitable for stress-resistance change rate signal prediction is finally determined.

[0046] ​Step two three: supervised learning framework training. In this invention, a supervised learning training framework is used to train the general piezoresistive signal prediction model of the self-sensing concrete sensor. In the training process, the distance between the predicted stress and the measured stress in high-dimensional space is constrained by the loss function, realizing the accurate matching of the piezoresistive signal and the structural stress state. The loss function is shown in equation 5.

[0047] (5) K-fold cross-validation is used for training, the network iteration number in each fold is 100, the training batch size is 64, the learning rate is 0.005. The hidden layer is fixed to 3 layers, the number of hidden layer nodes is 128, and the optimizer is Adam. The data is scaled by maximum and minimum normalization.

[0048] The network in this invention is realized by back propagation algorithm and gradient descent method. The specific process is: first, calculate the predicted value by forward propagation, then calculate the gradient of the loss function with respect to each parameter by back propagation, and then adjust the weights and biases in the network according to the gradient information using the gradient descent method to minimize the loss function. The parameter update method of gradient descent is shown in equation 6 and equation 7.

[0049] (6) (7) Step three: multi-task optimization self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm. In this step, based on the stress-resistance rate data set constructed in step one and the piezoresistive signal general prediction model trained in step two, a multi-task optimization self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm is proposed, and the method framework is shown in Figure 4 As shown in the following steps: Step three one: construction of fitness function. The general piezoresistive signal prediction model pre-trained in step two is used to calculate the strength performance corresponding to the mix ratio as the strength evaluation index in the optimization objective. Then according to the cost of each cubic meter of self-sensing composite material, it is calculated respectively as shown in equation 8 and equation 9. The optimization objective of the algorithm is to maximize the strength and minimize the economy at the same time.

[0050]

[0051]

[0052] The optimization objective is to maximize the strength and minimize the economy at the same time under the premise of meeting the constraint conditions (including the range of each component allocation ratio and the upper and lower limits of the resistance change rate). The optimization variable is the content of each component (carbon black, carbon nanotube, carbon fiber, water reducing agent, etc.) in the mix ratio.

[0053] Step three two: non-dominated sorting genetic algorithm. To achieve multi-objective optimization, the present application introduces a non-dominated sorting genetic algorithm (NSGA-II). The algorithm stratifies solutions in the population by non-dominated sorting, assigning solutions that are not dominated by any individual to the highest rank. The diversity of solutions is maintained by calculating the crowding distance, avoiding the algorithm from falling into local optimum. In each generation, NSGA-II generates new solutions through crossover and mutation operations, merges the parent and child generations, and sorts them to select individuals with higher fitness to enter the next generation, gradually approaching the Pareto front solution set. The algorithm schematic is shown in Figure 4 .

[0054] The non-dominated sorting genetic algorithm realizes multi-objective optimization, which is to deduce the optimized self-sensing concrete mix proportion on the basis of existing targets (strength, cost).

[0055] Specifically, The following is an example of a set of self-sensing concrete sensors with 0.5% carbon black, carbon nanotubes and carbon fibers in a structural health monitoring case, providing a specific implementation method of the present application.

[0056] Constructing a self-sensing concrete sensor stress-resistance change rate dataset: a digital multimeter is used to collect sensor resistance signals in real time, and the loading process is controlled by a stress actuator. To solve the problem of inconsistent loading sampling frequency and resistance measurement sampling frequency, the resistance change rate calculation method and stress-signal sampling alignment algorithm proposed by the present application are used to preprocess the original data, so that the stress signal at any time corresponds to the corresponding resistance change rate signal, and a stress-resistance change rate data pair is constructed to form a complete dataset, meeting the subsequent piezoresistive signal prediction model training requirements.

[0057] General piezoresistive signal prediction of self-sensing concrete sensor based on multi-scale feature embedding network: the stress-resistance change rate dataset constructed above is input into the multi-scale feature embedding network architecture proposed by the present application for supervised learning to capture the internal pattern of piezoresistive signal. During training, the K-fold cross-validation method is used to evaluate the model performance, with 100 iterations per fold and a training batch size of 64 and a learning rate of 0.005. The network structure is fixed at three hidden layers with 256 nodes each, and the Adam optimizer (learning rate = 0.002) is used for parameter optimization. The data is normalized before training to improve training stability. The model finally realizes high-precision prediction of the piezoresistive signal of the self-sensing concrete sensor, and the prediction results are shown in β . Figure 5

[0058] ​Multi-task optimization of self-sensing concrete sensor mix design based on non-dominated sorting genetic algorithm: the pre-trained general piezoresistive signal prediction model obtained in step two is used as the core tool for strength fitness evaluation, and its network parameters are fixed. Based on the model, combined with 1m 3 The material economic cost is used as another fitness index, and the non-dominated sorting genetic algorithm (NSGA-II) is used for multi-objective optimization design. The genetic algorithm parameters are set as follows: population size 100, parent 50, child 100, crossover probability 0.7, mutation probability 0.2, and mutation operation subject to Gaussian distribution. The genetic iteration number is set to 1000 generations. Through algorithm running, the high-quality self-sensing concrete sensor mix ratio scheme meeting the dual objectives of strength and economy is obtained.

[0059] Embodiment two The embodiment provides a self-sensing piezoresistive signal prediction and mix design system based on multi-scale mix feature embedding, which uses the self-sensing piezoresistive signal prediction and mix design method based on multi-scale mix feature embedding as described in embodiment one, and comprises: A data set construction module: a self-sensing concrete sensor stress-resistance change rate data set is constructed; A prediction model construction module: using the data set of step one, a self-sensing piezoresistive signal prediction model based on mix multi-scale feature embedding is obtained; A prediction model use module: based on the prediction model of step two, a multi-task optimization of self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm is realized.

[0060] The data set construction module of the present application solves the problem of inconsistent sampling frequency and the inability to directly align the signal in the experimental data of the self-sensing concrete sensor, and ensures the effectiveness and consistency of the data required for subsequent model training.

[0061] Embodiment three The electronic device provided by the embodiment of the present application includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory is used to store software programs and modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory and the processor are connected through a bus. Specifically, the processor realizes any step in the above embodiment one by running the above computer program stored in the memory.

[0062] It should be understood that, in the embodiments of the present application, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0063] The memory can include read-only memory, flash memory, and random access memory, and provide instructions and data to the processor. Part or all of the memory can also include non-volatile random access memory.

[0064] As can be seen from the above, the electronic device provided by the embodiments of the present application can realize the self-perception pressure resistance signal prediction and mix design method with multi-scale cooperation feature embedding as described in Embodiment One by running a computer program, and by designing a neural network architecture with multi-scale embedding structure, combining supervised learning and multi-objective optimization strategy, the accurate prediction of the self-perception concrete sensor pressure resistance signal and the efficient mix design are realized, and a new and efficient solution for structural health monitoring is provided.

[0065] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other manners. For example, the embodiments of the apparatus / device described above are merely illustrative. For example, the division of the above modules or units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0066] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A self-sensing piezoresistive signal prediction and mix ratio design method based on multi-scale mix ratio feature embedding, characterized in that: The method comprises the following steps: Step 1: Construct a dataset of stress-resistance change rate of self-sensing concrete sensors; Step 2: Using the data set from step 1, a self-sensing piezoresistive signal prediction model based on the embedding of multi-scale features of the mix ratio is obtained; Step 3: Based on the prediction model of step 2, a multi-task optimization self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm is implemented.

2. The method according to claim 1, characterized in that Specifically, step one introduces the measured piezoresistive signals of 16 sets of self-sensing cement-based sensors with different additions of carbon black, carbon nanotubes, and carbon fibers. All the ratios involved are based on a fixed matrix material composition, cured for 28 days at 20±2°C and 90% RH, and the specimen size is 40×40×40 mm. The dataset contains a total of 5,300 sets of stress-resistance change rate data pairs.

3. The method according to claim 1, characterized in that The prediction model in step 2 includes a neural network that embeds multi-scale ratio features in multiple hidden layers; the piezoresistive pattern is learned in a high-dimensional feature space through a supervised learning method to achieve universal modeling of the piezoresistive response law of the self-sensing sensor, that is, a pre-trained multi-scale feature embedded piezoresistive signal prediction network model; after the network model training is completed, it can be directly applied to the self-sensing sensor embedded in the structure.

4. The method according to claim 3, characterized in that The step 2 specifically includes the following steps: Step 21: Build a multi-scale feature embedding neural network, and take the content of conductive material, water reducer content and resistance change rate as input, and output the corresponding stress; Step 22: Based on the network in step 21, set the optimal hyperparameters obtained using the Bayesian optimization method; Step 23: Use the optimal hyperparameters obtained in step 22 to train the self-sensing piezoresistive signal prediction model in a supervised learning framework.

5. The method according to claim 1, characterized in that: The specific step three is to take strength and economy as dual fitness functions, set constraints, comprehensively consider performance and cost factors, and optimize the goal to maximize strength and minimize economy at the same time, wherein the strength is calculated using a pre-trained multi-scale feature embedding piezoresistive signal prediction model, and the economy index is 1m3 per cubic meter. 3 The cost of self-sensing composite materials is used as the measurement standard; through a multi-task optimization algorithm, a balanced design between piezoresistive performance, mechanical properties and cost is achieved, and the optimal self-sensing concrete sensor mix ratio is output.

6. The method according to claim 5, characterized in that The step three specifically includes the following steps: Step 31: Construct fitness function; Step 32: Use the fitness function of step 31 to introduce the non-dominated sorting genetic algorithm.

7. The method according to claim 6, characterized in that The non-dominated sorting genetic algorithm realizes multi-objective optimization and, based on existing objectives, inversely derives an optimized self-sensing concrete mix ratio.

8. A self-sensing piezoresistive signal prediction and mix ratio design system based on multi-scale mix ratio feature embedding, characterized in that: The system uses the self-sensing piezoresistive signal prediction and mix ratio design method based on multi-scale mix ratio feature embedding as described in any one of claims 1 to 7, and the system includes: Dataset construction module: Construct a self-sensing concrete sensor stress-resistance change rate dataset; Prediction model building module: Using the data set from step 1, a self-sensing piezoresistive signal prediction model based on the embedding of multi-scale features of the mix ratio is obtained; Prediction model usage module: Based on the prediction model in step 2, a multi-task optimization self-sensing concrete sensor mix design method based on non-dominated sorting genetic algorithm is implemented.

9. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images