An optimization processing method and device for a flexible strain sensor

Abstract

The application provides an optimization processing method and device for a flexible strain sensor. The method comprises the following steps: determining characteristic data of a strain sensor to be analyzed; the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silica gel elastomer according to a preset ratio; inputting the characteristic data of the strain sensor into a sensor performance prediction model to obtain an output sensor performance prediction result, so as to calibrate a performance parameter of the strain sensor based on the sensor performance prediction result and obtain a target strain sensor meeting a preset performance condition; and the sensor performance prediction model is a deep learning model trained on the basis of sample characteristic data and actual sensor performance prediction results corresponding to the sample characteristic data. The method provided by the application realizes rapid and accurate calibration of the performance parameter of the strain sensor through the sensor performance prediction model, thereby improving the performance and stability of the strain sensor.

Description

Technical Field

[0001] This invention relates to the field of computer application technology, specifically to an optimized processing method and apparatus for flexible strain sensors. Additionally, it relates to an electronic device and a processor-readable storage medium. Background Technology

[0002] A strain sensor is a detection device that senses the strain being measured and converts the sensed strain information into an electrical signal output according to a certain rule, to meet the needs of information recording and feedback control. Strain sensors are one of the key technologies in current intelligent robots. For example, they enable robotic exoskeletons or prosthetic systems to obtain tactile sensations close to those of the human body, and also provide industrial robots and bionic robots with more precise force and torque control, as well as accurate monitoring. In addition, strain sensors also play an important role in portable intelligent electronic products. For example, wearable strain sensor devices can detect the mechanical signals generated when electronic products come into direct contact with the external environment, and are a core component for future Internet of Things (IoT)-assisted human-environment interaction.

[0003] Traditional strain sensor materials typically utilize metals or piezoelectric ceramics, but these materials exhibit high rigidity and low strain, making it difficult to meet the growing demand for strain capacity and flexible monitoring. In recent years, with the rise of functional nanomaterials (such as graphene and carbon nanotubes), more and more researchers are using nanomaterials as active materials for strain sensing. These nanomaterials can be deposited as ultrathin coatings on flexible substrates, simultaneously achieving high stability, high sensitivity, high linearity, and low hysteresis. However, the fabrication process of these nanomaterials is extremely complex and cumbersome, requiring large-scale equipment and continuous power, resulting in significant carbon emissions, environmental pollution, and resource waste. Therefore, designing an optimized strain sensor that combines high stability, high sensitivity, and high linearity while achieving a simple and efficient process has become an urgent problem to be solved. Summary of the Invention

[0004] Therefore, the present invention provides an optimization processing method and apparatus for flexible strain sensors to solve the defects of existing strain sensor optimization processing schemes, which have high limitations and cannot achieve rapid and accurate calibration of the performance parameters of the strain sensor, resulting in poor stability and performance of the obtained strain sensor.

[0005] In a first aspect, the present invention provides an optimization processing method for flexible strain sensors, comprising: determining characteristic data of the strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio;

[0006] The characteristic data of the strain sensor is input into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model, so as to calibrate the performance parameters of the strain sensor based on the sensor performance prediction result and obtain a target strain sensor that meets the preset performance conditions.

[0007] The sensor performance prediction model is a deep learning model trained on the basis of sample feature data and the actual sensor performance prediction results corresponding to the sample feature data.

[0008] Furthermore, the optimization processing method for flexible strain sensors further includes: pre-determining the sample feature data;

[0009] The pre-determined sample feature data specifically includes:

[0010] Collect raw characteristic data of the carbon black sensor; the raw characteristic data includes the property characteristic data of the carbon black sensor, the data of the sensor resistance changing with the strain magnitude, the data of the sensor sensitivity changing with the number of cyclic stretching, the data of the sensor linearity changing with the number of cyclic stretching, the data of the sensor hysteresis effect changing with the number of cyclic stretching, and the data of the sensor resistance changing with the number of cyclic stretching.

[0011] The original feature data is optimized based on a preset sample preprocessing model to obtain the sample feature data.

[0012] Furthermore, the attribute feature data includes at least one of the following features of the carbon black sensor: length, width, thickness, Young's modulus, fracture strength, and electrical conductivity.

[0013] Furthermore, the performance parameters of the strain sensor are calibrated based on the sensor performance prediction results. Specifically, this includes calibrating the sensitivity parameters, linear parameters, and hysteresis effect parameters of the strain sensor using the sensor performance prediction results, so as to amplify and eliminate the sensitivity parameters, correct the linear parameters, and eliminate the data deviations corresponding to the hysteresis effect parameters.

[0014] Furthermore, the characteristic data of the strain sensor is input into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model. Specifically, this includes: inputting the characteristic data of the strain sensor into the corresponding sensor performance prediction model according to the type of different application scenarios to obtain the sensor performance prediction result output by the sensor performance prediction model.

[0015] The sensor performance prediction model is one of the following: a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network, and a deep learning model based on knowledge distillation.

[0016] Furthermore, the silicone elastomer is a liquid silicone resin elastomer; the strain sensor is a carbon black sensor obtained by mixing carbon black material into the liquid silicone resin elastomer at a ratio of 5 wt.% or 7 wt.%; the carbon black material powder is dispersed in the silicone resin elastomer after cross-linking liquid curing to form a conductive network.

[0017] Secondly, the present invention also provides an optimization processing device for flexible strain sensors, comprising: a feature data determination unit for determining feature data of the strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio;

[0018] The calibration processing unit is used to input the characteristic data of the strain sensor into a preset sensor performance prediction model, and obtain the sensor performance prediction result output by the sensor performance prediction model, so as to calibrate the performance parameters of the strain sensor based on the sensor performance prediction result and obtain a target strain sensor that meets the preset performance conditions.

[0019] The sensor performance prediction model is a deep learning model trained on the basis of sample feature data and the actual sensor performance prediction results corresponding to the sample feature data.

[0020] Furthermore, the calibration of the performance parameters of the strain sensor based on the sensor performance prediction results specifically includes: calibrating the sensitivity parameters, linear parameters, and hysteresis effect parameters of the strain sensor using the sensor performance prediction results, so as to amplify and eliminate the sensitivity parameters, correct the linear parameters, and eliminate the data deviations corresponding to the hysteresis effect parameters.

[0021] Furthermore, the optimization processing device for the flexible strain sensor further includes: a training sample determination unit, used to predetermine the sample feature data;

[0022] The training sample determination unit is specifically used for:

[0023] Collect raw characteristic data of the carbon black sensor; the raw characteristic data includes the property characteristic data of the carbon black sensor, the data of the sensor resistance changing with the strain magnitude, the data of the sensor sensitivity changing with the number of cyclic stretching, the data of the sensor linearity changing with the number of cyclic stretching, the data of the sensor hysteresis effect changing with the number of cyclic stretching, and the data of the sensor resistance changing with the number of cyclic stretching.

[0024] The original feature data is optimized based on a preset sample preprocessing model to obtain the sample feature data.

[0025] Furthermore, the attribute feature data includes at least one of the following features of the carbon black sensor: length, width, thickness, Young's modulus, fracture strength, and electrical conductivity.

[0026] Furthermore, the calibration processing unit is specifically used to: input the characteristic data of the strain sensor into the corresponding sensor performance prediction model according to the type of different application scenarios, and obtain the sensor performance prediction result output by the sensor performance prediction model;

[0027] The sensor performance prediction model is one of the following: a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network, and a deep learning model based on knowledge distillation.

[0028] Furthermore, the silicone elastomer is a liquid silicone resin elastomer; the strain sensor is a carbon black sensor obtained by mixing carbon black material into the liquid silicone resin elastomer at a ratio of 5 wt.% or 7 wt.%; the carbon black material powder is dispersed in the silicone resin elastomer after cross-linking liquid curing to form a conductive network.

[0029] Thirdly, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the optimization processing method for flexible strain sensors as described in any of the preceding claims.

[0030] Fourthly, the present invention also provides a processor-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the optimized processing method for flexible strain sensors as described in any of the preceding claims.

[0031] The optimization processing method for flexible strain sensors provided by this invention inputs the characteristic data of the strain sensor to be analyzed into a sensor performance prediction model to obtain the corresponding sensor performance prediction results. This enables rapid and accurate calibration of the performance parameters of the strain sensor based on the sensor performance prediction results, resulting in a target strain sensor that meets preset performance conditions, thereby improving the performance and stability of the strain sensor. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the optimization processing method for flexible strain sensors provided in an embodiment of the present invention.

[0034] Figure 2 This is an application diagram of the optimized processing method for flexible strain sensors provided in the embodiments of the present invention;

[0035] Figure 3 This is a schematic diagram of mixing carbon black and elastomer and then coating them into a uniform plane according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram showing the changes in the sensing performance of the strain sensor device under different carbon black material doping ratios provided in the embodiments of the present invention;

[0037] Figure 5 This is a schematic diagram comparing the carbon black sensor and the nanomaterial sensor provided in the embodiments of the present invention;

[0038] Figure 6 Figures 1-4 are, in order, schematic diagrams illustrating the inherent low sensitivity of the carbon black sensor provided in the embodiments of the present invention, the low linearity and large hysteresis effect exhibited during the application of the sensor performance prediction model, and the self-calibration of long-term stability.

[0039] Figure 7 This is a complete implementation diagram of the optimization processing method for flexible strain sensors provided in the embodiments of the present invention;

[0040] Figure 8 This is a schematic diagram of the structure of the optimized processing device for flexible strain sensors provided in an embodiment of the present invention;

[0041] Figure 9This is a schematic diagram of the physical structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] The following is a detailed description of embodiments of the optimized processing method for flexible strain sensors described in this invention. For example... Figure 1 As shown, it is a flowchart illustrating the optimization processing method for flexible strain sensors provided in an embodiment of the present invention. The specific implementation process includes the following steps:

[0044] Step 101: Determine the characteristic data of the strain sensor to be analyzed; wherein, the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio.

[0045] In this embodiment of the invention, sample feature data needs to be predetermined before performing this step, and model training is performed based on the sample feature data to obtain a sensor performance prediction model. This sensor performance prediction model is a trained deep learning model, which can adaptively learn according to the basic characteristics of the strain sensor. Its internal deep neural network can learn the correlation between parameters such as the strain sensor signal, working time, working strain, and number of stretching cycles, thereby achieving subsequent intelligent control and calibration.

[0046] During model training, the first step is to determine the model training samples. Specifically, the raw feature data of the carbon black sensor is first collected. This raw feature data includes the sensor's property characteristics, the sensor's resistance as a function of strain, the sensor's sensitivity as a function of the number of cyclic stretching cycles, the sensor's linearity as a function of the number of cyclic stretching cycles, the sensor's hysteresis effect as a function of the number of cyclic stretching cycles, and the sensor's resistance as a function of the number of cyclic stretching cycles. The raw feature data is then optimized based on a pre-defined sample preprocessing model to obtain the sample feature data. The property characteristics include at least one of the following basic information: length, width, thickness, Young's modulus, fracture strength, and conductivity of the carbon black sensor. Cyclic testing is performed by applying an increasing strain (from 0% to 100%) for 5000 cycles to obtain data on the changes in the carbon black sensor's resistance as a function of strain, the sensor's sensitivity as a function of the number of cyclic stretching cycles, the sensor's linearity as a function of the number of cyclic stretching cycles, the sensor's hysteresis effect as a function of the number of cyclic stretching cycles, and the sensor's resistance as a function of the number of cyclic stretching cycles. It should be noted that the raw sensor data suffers from problems such as uneven time interval sequences, irregular arrangement, messy multi-mode data, noise, and large data volume. For the raw sensor data (i.e., raw feature data) obtained above, a preset preprocessing algorithm (including batch cycle determination and batch accurate strain degree calculation) can be used for batch processing to obtain training datasets (i.e., sample feature data) for training different deep learning modules, thereby improving the quantity and quality of sample data in deep learning model training. Specifically, in the process of optimizing the raw feature data based on the preset sample preprocessing model to obtain the sample feature data, the following processing steps are performed after loading the raw feature data: removing erroneous data with negative acquisition time changes based on the preset sample preprocessing model; determining the time of a complete strain sensor tensile cycle and the start time of the tensile experiment; determining the cycle number of each raw feature data point; determining the percentage time interval of each data point within its cycle number, and obtaining the strain value corresponding to each data point through the percentage time interval.

[0047] In this embodiment of the invention, after training the sensor performance prediction model, the characteristic data of the strain sensor to be analyzed can be determined in this step. The strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio. The silicone elastomer is a liquid silicone resin elastomer; the strain sensor can refer to a carbon black sensor obtained by mixing carbon black material into the liquid silicone resin elastomer at a ratio of 5 wt.% or 7 wt.%, and is not specifically limited here.

[0048] Step 102: Input the feature data of the strain sensor into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model. This allows for the calibration of the strain sensor's performance parameters based on the sensor performance prediction result, resulting in a target strain sensor that meets preset performance conditions. The sensor performance prediction model is a deep learning model trained on the sample feature data and the corresponding actual sensor performance prediction results.

[0049] In this embodiment of the invention, the feature data of the strain sensor is input into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model. The specific implementation process includes: inputting the feature data of the strain sensor into a corresponding sensor performance prediction model according to different application scenarios to obtain the sensor performance prediction result output by the sensor performance prediction model. The sensor performance prediction model is one of the following: a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network (attention mechanism network), and a deep learning model based on knowledge distillation. Specifically, the performance parameters of the strain sensor are calibrated based on the sensor performance prediction result. The actual implementation process includes: calibrating the sensitivity parameter, linear parameter, and hysteresis effect parameter of the strain sensor using the sensor performance prediction result to amplify and eliminate the sensitivity parameter, correct the linear parameter, and eliminate the data deviation corresponding to the hysteresis effect parameter.

[0050] In practical implementation, it is necessary to collect basic characteristic data (i.e., property characteristic data) of the strain sensor in advance. This basic characteristic data includes Young's modulus, fracture strength, electrical conductivity, and characteristic parameters such as length, width, and thickness. These characteristic parameters affect the performance of the strain sensor, and it is necessary to effectively predict and calibrate the performance of the strain sensor based on changes in these characteristic parameters.

[0051] like Figure 7As shown, in one embodiment, different deep learning models need to be constructed during the calibration of the sensitivity parameters, linearity parameters, and hysteresis effect parameters of the strain sensor. These deep learning models specifically include deep learning models based on one-dimensional convolutional neural networks, deep learning models based on long short-term memory recurrent neural networks, deep learning models based on gated recurrent neural networks, deep learning models based on Transformer networks (attention mechanism networks), and deep learning models based on knowledge distillation. These deep learning models are trained using collected sample feature data, enabling all models to accurately predict the resistance change of the strain sensor based on the applied strain. Among these, the deep learning model based on Transformer networks performs optimally. However, the deep learning model based on one-dimensional convolutional neural networks has the fewest parameters, is more energy-efficient, and can be deployed more easily on ordinary hardware. This necessitates deploying different trained deep learning models for different application requirements in practical implementations to achieve the best results.

[0052] For example, in practical implementation, a deep learning model based on a one-dimensional convolutional neural network can be selected as the sensor performance prediction model. Based on the prediction results of the sensor performance prediction model (i.e., the sensor performance prediction results), the resistance change of the strain sensor caused by strain is transformed into a linear equation using a variant of the sigmoid function (the linear equation parameters have been verified with 5000 cycles of data, and the mean square error is less than 10). This process completes the linear calibration. Here, after converting the resistance-strain curve into a linear line, multiplying the resistance change by a factor of 1000 directly corrects the sensitivity of the strain sensor to a factor of 1000. Finally, regarding the hysteresis effect, based on the same idea, a variant of the sigmoid function is used to transform both the outward path (strain increasing from 0% to 100%) and the return path (strain decreasing from 100% to 0%) into a linear curve. After the above calibration process, the carbon black sensor can achieve high sensitivity (GF = 1000) and high linearity (R² = 1000). 2 The strain gauge impedance is approximately 0.99, and there is zero hysteresis. Furthermore, due to the poor long-term stability of strain sensor signals, a pre-defined adaptive algorithm model can be used to eliminate environmental noise and handle sudden irregular signals, ultimately achieving adaptive learning for each strain sensor. This adaptive algorithm model utilizes a Transformer as its core, trained on large datasets to automatically adapt to the corresponding material system. It can automatically eliminate the influence of environmental noise, and due to its good predictive generalization, it can also handle sudden irregular signals.

[0053] To address the stability issue of strain sensors, a deep learning model based on Transformer networks (i.e., the Transformer model) can be used. When processing very long sequences, the Transformer model can learn the subtle changes in material properties with increasing cycle count, thus enabling real-time dynamic adjustment of the strain sensor signal. However, since the Transformer model has a large number of parameters, which may be detrimental to energy conservation and unsuitable for mobile deployment, this embodiment of the invention uses knowledge distillation to condense a smaller target model with approximate performance. Specifically, a preset Transformer model with the best predictive performance can be used as the "teacher model," and a smaller "student model" learns the output of the "teacher model," thereby achieving the effect of knowledge distillation. It should be noted that different applications may place different requirements on strain sensors and corresponding intelligent systems; therefore, an optimal combination of intelligent systems can be used for different applications. This intelligent system combination can be as follows: Figure 7 As shown, thanks to the support of intelligent algorithms, this carbon black sensor can exhibit high sensitivity (GF=1000, more than a hundredfold improvement) and high linearity (R0). 2 (≈0.99), zero hysteresis, and extended service life by at least two orders of magnitude.

[0054] In addition, such as Figure 2 As shown, the embodiments of the present invention also include a multi-channel wireless transmission system for carbon black sensors, which can simultaneously detect the electrical signals of multiple strain sensors to obtain data from the real-time feedback and closed-loop control system of the machine equipment. The data is transmitted wirelessly via Bluetooth to the sensor performance prediction model of the terminal computing device to realize real-time feedback and closed-loop control of the robot.

[0055] In practical implementation, when calibrating the performance parameters of the strain sensor based on the sensor performance prediction results to obtain a target strain sensor that meets the preset performance conditions, carbon black (CB) extracted from chemical waste and liquid silicone elastomer (Smooth-On, Ecoflex 00-30) can be mixed at a weight ratio of 1:10 (CB: Ecoflex). Figure 3As shown, after thorough mixing, the mixture is evenly coated onto a flat surface with a thickness of approximately 0.3 cm using a scraping method. After being left at room temperature for 24 hours, it is cut into a predetermined size. Wires are then connected to both ends of the rectangle to obtain a strain sensor based on carbon black material. After mixing carbon black material with an elastomer according to the above process, carbon black particles are dispersed within the elastomer to form a conductive network. When external strain is applied, the distance between the carbon black particles dispersed in the elastomer is stretched, causing a change in the conductive network and thus an increase in resistance. Therefore, this device can serve as an effective strain sensor: that is, the resistance increases with increasing strain.

[0056] This invention optimized and verified the doping ratio of carbon black materials, and the results are as follows: Figure 4 As shown, when the carbon black doping ratio is 10 wt.%, the resulting device struggles to exhibit large tensile behavior and fractures under 10 wt.% strain. When the carbon black doping ratio is 2 wt.%, the insufficient carbon black content results in extremely poor conductivity, making it difficult to achieve a change in resistance under strain. After doping with 5 wt.% and 7 wt.%, the resulting devices can serve as effective strain sensors. Considering that the 5 wt.% device has a smaller error, 5 wt.% can be used as the standard for strain sensor fabrication in practical applications.

[0057] like Figure 5 As shown, nanomaterial sensors are many times more sensitive than carbon black sensors. The gauge factor (GF) can be used to evaluate the sensitivity of strain sensors. GF ​​can be calculated using the following formula:

[0058]

[0059] Where ε refers to strain, R0 and R ε These are the initial resistance and the resistance under ε strain of the sensor, respectively. In practice, the GF of a strain sensor based on carbon black material is 1.7, while the GF of a nanomaterial sensor is 100.

[0060] In the data collection process of strain sensors, this invention uses a cyclic tensile testing machine to obtain a large amount of sensor data from carbon black sensors and nanomaterial sensors. Data types include sensitivity, linearity, hysteresis data, and long-term stability data. During the automatic calibration of errors and biases using deep learning algorithms, a deep learning model is used to learn and analyze the carbon black sensor data. The results show that the trained deep learning model can process carbon black sensor data in real time, autonomously calibrating and compensating for the inherent low sensitivity, low linearity, large hysteresis effect, and weak long-term stability of carbon black sensors, thereby achieving performance exceeding that of nanomaterial sensors. Figure 6 The diagram illustrates the autonomous calibration and data compensation of a deep learning model for the inherent low sensitivity, low linearity, large hysteresis, and long-term stability of carbon black sensors during application. Furthermore, in the construction of the strain sensor network, multiple carbon black sensors can be connected by wires and attached to multiple joints of the robotic arm to form a comprehensive sensor network for detecting robotic arm movement. The sensors are integrated into a design that improves ease of installation, adapts sensor size, designs sensor circuits, and facilitates signal transmission from multiple sensors, thereby ensuring stable sensor signals and ensuring that the operational performance of the robotic arm equipped with the strain sensors of this invention is not affected.

[0061] It should be noted that, in this embodiment of the invention, carbon black material extracted from chemical waste is used to design the strain sensor. Compared with the preparation of nanomaterials, this significantly reduces carbon emissions. Furthermore, a deep learning model is introduced to calibrate the sensitivity, linearity, and hysteresis parameters of the strain sensor, eliminating the data bias and errors caused by these parameters, thereby improving the performance of the strain sensor. In addition, this strain sensor can be integrated with a robotic arm and utilize a pre-set artificial intelligence platform to process large amounts of sensor data to construct a digital and intelligent network.

[0062] The optimization processing method for flexible strain sensors described in this invention involves inputting the characteristic data of the strain sensor to be analyzed into a sensor performance prediction model to obtain the corresponding sensor performance prediction results. This enables rapid and accurate calibration of the performance parameters of the strain sensor based on the sensor performance prediction results, resulting in a target strain sensor that meets preset performance conditions, thereby improving the performance and stability of the strain sensor.

[0063] Corresponding to the optimization processing method for flexible strain sensors provided above, this invention also provides an optimization processing apparatus for flexible strain sensors. Since the embodiments of this apparatus are similar to the method embodiments described above, the description is relatively simple. For relevant details, please refer to the description in the method embodiment section above. The embodiments of the optimization processing apparatus for flexible strain sensors described below are merely illustrative. Please refer to... Figure 8 As shown, it is a structural schematic diagram of an optimized processing device for flexible strain sensors provided in an embodiment of the present invention.

[0064] The optimized processing device for flexible strain sensors described in this invention specifically includes the following parts:

[0065] The feature data determination unit 801 is used to determine the feature data of the strain sensor to be analyzed; wherein, the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio;

[0066] The calibration processing unit 802 is used to input the feature data of the strain sensor into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model, so as to calibrate the performance parameters of the strain sensor based on the sensor performance prediction result and obtain a target strain sensor that meets the preset performance conditions; wherein, the sensor performance prediction model is a deep learning model trained on the basis of sample feature data and the actual sensor performance prediction result corresponding to the sample feature data.

[0067] Furthermore, the calibration of the performance parameters of the strain sensor based on the sensor performance prediction results specifically includes: calibrating the sensitivity parameters, linear parameters, and hysteresis effect parameters of the strain sensor using the sensor performance prediction results, so as to amplify and eliminate the sensitivity parameters, correct the linear parameters, and eliminate the data deviations corresponding to the hysteresis effect parameters.

[0068] Furthermore, the optimization processing device for the flexible strain sensor further includes: a training sample determination unit, used to predetermine the sample feature data;

[0069] The training sample determination unit is specifically used for:

[0070] Collect raw characteristic data of the carbon black sensor; the raw characteristic data includes the property characteristic data of the carbon black sensor, the data of the sensor resistance changing with the strain magnitude, the data of the sensor sensitivity changing with the number of cyclic stretching, the data of the sensor linearity changing with the number of cyclic stretching, the data of the sensor hysteresis effect changing with the number of cyclic stretching, and the data of the sensor resistance changing with the number of cyclic stretching.

[0071] The original feature data is optimized based on a preset sample preprocessing model to obtain the sample feature data.

[0072] Furthermore, the attribute feature data includes at least one of the following features of the carbon black sensor: length, width, thickness, Young's modulus, fracture strength, and electrical conductivity.

[0073] Furthermore, the calibration processing unit is specifically used to: input the characteristic data of the strain sensor into the corresponding sensor performance prediction model according to the type of different application scenarios, and obtain the sensor performance prediction result output by the sensor performance prediction model;

[0074] The sensor performance prediction model is one of the following: a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network, and a deep learning model based on knowledge distillation.

[0075] Furthermore, the silicone elastomer is a liquid silicone resin elastomer; the strain sensor is a carbon black sensor obtained by mixing carbon black material into the liquid silicone resin elastomer at a ratio of 5 wt.% or 7 wt.%; the carbon black material powder is dispersed in the silicone resin elastomer after cross-linking liquid curing to form a conductive network.

[0076] The optimization processing device for flexible strain sensors described in this embodiment of the invention inputs the characteristic data of the strain sensor to be analyzed into the sensor performance prediction model to obtain the corresponding sensor performance prediction results. This enables rapid and accurate calibration of the performance parameters of the strain sensor based on the sensor performance prediction results, resulting in a target strain sensor that meets the preset performance conditions, thereby improving the performance and stability of the strain sensor.

[0077] Corresponding to the optimization processing method for flexible strain sensors provided above, this invention also provides an electronic device. Since the embodiment of this electronic device is similar to the method embodiment described above, it is described simply. For relevant details, please refer to the description in the method embodiment section above. The electronic device described below is merely illustrative. Figure 9The diagram shows a physical structure of an electronic device disclosed in an embodiment of the present invention. The electronic device may include a processor 901, a memory 902, and a communication bus 903. The processor 901 and the memory 902 communicate with each other via the communication bus 903 and communicate with external devices via a communication interface 904. The processor 901 can call logical instructions in the memory 902 to execute an optimization processing method for a flexible strain sensor. This method includes: determining the characteristic data of the strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio; inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain the sensor performance prediction result output by the sensor performance prediction model, thereby calibrating the performance parameters of the strain sensor based on the sensor performance prediction result to obtain a target strain sensor that meets preset performance conditions; wherein the sensor performance prediction model is a deep learning model trained on sample characteristic data and the actual sensor performance prediction results corresponding to the sample characteristic data.

[0078] Furthermore, the logical instructions in the aforementioned memory 902 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: memory chips, USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0079] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program stored on a processor-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer can execute the optimization processing method for flexible strain sensors provided in the above-described method embodiments. The method includes: determining the characteristic data of a strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio; inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, so as to calibrate the performance parameters of the strain sensor based on the sensor performance prediction result, and obtain a target strain sensor that meets preset performance conditions; wherein the sensor performance prediction model is a deep learning model trained on sample characteristic data and the actual sensor performance prediction results corresponding to the sample characteristic data.

[0080] In another aspect, embodiments of the present invention also provide a processor-readable storage medium storing a computer program, which, when executed by a processor, implements the optimization processing method for flexible strain sensors provided in the above embodiments. The method includes: determining characteristic data of a strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silicone elastomer in a preset ratio; inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, thereby calibrating the performance parameters of the strain sensor based on the sensor performance prediction result to obtain a target strain sensor that meets preset performance conditions; wherein the sensor performance prediction model is a deep learning model trained based on sample characteristic data and the actual sensor performance prediction results corresponding to the sample characteristic data.

[0081] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0082] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0083] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method of optimization processing for a flexible strain sensor, characterized by, The method comprises the steps of: determining characteristic data of a strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing carbon black material powder and silica gel elastomer according to a preset ratio; inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, so as to calibrate a performance parameter of the strain sensor based on the sensor performance prediction result and obtain a target strain sensor meeting a preset performance condition; wherein the sensor performance prediction model is a deep learning model trained based on sample characteristic data and actual sensor performance prediction results corresponding to the sample characteristic data; calibrating the performance parameter of the strain sensor based on the sensor performance prediction result, specifically including calibrating a sensitivity parameter, a linearity parameter and a hysteresis effect parameter of the strain sensor by using the sensor performance prediction result, so as to amplify and eliminate the data deviation corresponding to the sensitivity parameter, correct the data deviation corresponding to the linearity parameter and eliminate the data deviation corresponding to the hysteresis effect parameter; inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, specifically including inputting the characteristic data of the strain sensor into a corresponding sensor performance prediction model according to the type of different application scenarios to obtain a sensor performance prediction result output by the sensor performance prediction model; wherein the sensor performance prediction model is one of a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network and a deep learning model based on knowledge distillation processing.

2. The method for optimizing processing for a flexible strain sensor according to claim 1, wherein Further comprising: pre-determining the sample characteristic data; the pre-determining the sample characteristic data specifically includes: collecting original characteristic data of a carbon black sensor; the original characteristic data includes attribute characteristic data of the carbon black sensor, data of changes of sensor resistance with strain size, data of changes of sensor sensitivity with cycle tensile number, data of changes of sensor linearity with cycle tensile number, data of changes of sensor hysteresis effect with cycle tensile number and data of changes of sensor resistance with cycle tensile number; optimizing the original characteristic data based on a preset sample preprocessing model to obtain the sample characteristic data.

3. The method for optimization process for flexible strain sensor according to claim 2, wherein, The attribute characteristic data includes at least one of length, width, thickness, Young's modulus, breaking strength and conductivity of the carbon black sensor.

4. The method for optimization process for flexible strain sensor according to claim 1, wherein, inputting the characteristic data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, specifically including inputting the characteristic data of the strain sensor into a corresponding sensor performance prediction model according to the type of different application scenarios to obtain a sensor performance prediction result output by the sensor performance prediction model.

5. The method for optimization process for flexible strain sensor according to claim 1, wherein, The silica gel elastomer is a liquid silicone elastomer; the strain sensor is a carbon black sensor obtained by mixing a carbon black material into the liquid silicone elastomer at a proportion of 5 wt.% or 7 wt.%; and the carbon black material powder is dispersed in the silicone elastomer after cross-linking and solidification of the liquid to form a conductive network.

6. An optimized processing device for a flexible strain sensor, characterized by, Comprise: The feature data determination unit is configured to determine feature data of a strain sensor to be analyzed; wherein the strain sensor is a carbon black sensor obtained by mixing a carbon black material powder and a silica gel elastomer at a preset proportion; The calibration processing unit is configured to input the feature data of the strain sensor into a preset sensor performance prediction model to obtain a sensor performance prediction result output by the sensor performance prediction model, so as to calibrate a performance parameter of the strain sensor based on the sensor performance prediction result to obtain a target strain sensor meeting a preset performance condition; The sensor performance prediction model is a deep learning model trained based on sample feature data and actual sensor performance prediction results corresponding to the sample feature data; The calibration of the performance parameter of the strain sensor based on the sensor performance prediction result specifically includes: calibrating a sensitivity parameter, a linearity parameter and a hysteresis effect parameter of the strain sensor by using the sensor performance prediction result, so as to amplify and eliminate the data deviation corresponding to the sensitivity parameter, correct the data deviation corresponding to the linearity parameter, and eliminate the data deviation corresponding to the hysteresis effect parameter. The sensor performance prediction model is one of a deep learning model based on a one-dimensional convolutional neural network, a deep learning model based on a long short-term memory recurrent neural network, a deep learning model based on a gated recurrent neural network, a deep learning model based on a Transformer network, and a deep learning model after knowledge distillation processing.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the optimization processing method for the flexible strain sensor according to any one of claims 1 to 5.

8. A processor-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the steps of the optimization processing method for the flexible strain sensor according to any one of claims 1 to 5.

Images