PEDOT sensor nonlinear response compensation optimization system

Abstract

The invention discloses a PEDOT sensor nonlinear response compensation optimization system, and relates to the technical field of PEDOT sensors, the system comprises a cloud management center, and the cloud management center is in communication connection with a data acquisition module, a feature extraction module, a response compensation analysis module, a cloud compensation module, a compensation deviation analysis module and a feedback adjustment module. The method combines the neural network model through the cloud, precisely learns the nonlinear characteristics of the sensor, carries out precise compensation, carries out initial linear compensation at the end side, reduces the data amount transmitted to the cloud, improves the real-time performance of the system, carries out the model training and optimization at the cloud through a large amount of historical data, and improves the real-time performance of the system. According to the PEDOT sensor, the error introduced by nonlinear response is remarkably reduced, and the measurement precision and accuracy of the PEDOT sensor are effectively improved in an end-cloud cooperation mode, so that the sensor can provide high-reliability measurement data under different environmental conditions, and the requirement of high-precision measurement is met.

Description

Technical Field

[0001] The present invention relates to the technical field of PEDOT sensors, and in particular to a PEDOT sensor nonlinear response compensation optimization system. Background Art

[0002] PEDOT is a polymer of EDOT (3,4-ethylenedioxythiophene monomer). As a high-performance conductive polymer, it has been widely used in the sensor field due to its unique electrical, optical and chemical properties. PEDOT sensors can detect a variety of physical quantities, such as temperature and strain, and are highly favored for their high sensitivity, good mechanical properties and stretchability. However, in practical applications, PEDOT sensors often exhibit nonlinear response characteristics, that is, the output signal of the sensor is not strictly linear with the measured quantity.

[0003] In the prior art, the nonlinear response of PEDOT sensors introduces measurement errors, resulting in limited measurement accuracy and reduced sensor accuracy and reliability. Moreover, the influence of various environmental factors in practical applications of PEDOT sensors further introduces noise and instability. Therefore, how to transfer computing tasks to the cloud through end-cloud collaboration, and combine neural network models to learn the nonlinear characteristics of sensors, adapt to complex and changeable environmental conditions, and achieve high-precision linear compensation is the problem to be solved by the present invention. To this end, a PEDOT sensor nonlinear response compensation optimization system is proposed. Summary of the Invention

[0004] The present invention aims to provide a PEDOT sensor nonlinear response compensation optimization system to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A PEDOT sensor nonlinear response compensation optimization system includes a cloud management center, wherein the cloud management center is communicatively connected to a data acquisition module, a feature extraction module, a response compensation analysis module, a cloud compensation module, a compensation deviation analysis module, and a feedback adjustment module, wherein the modules are electrically connected;

[0007] The data acquisition module is used to collect and pre-process the raw data of the PEDOT sensor in real time, including the sensor output signal and environmental parameter data;

[0008] The feature extraction module is used to extract response compensation features from the pre-processed sensor output signal and environmental parameter data to obtain a response compensation feature set;

[0009] The response compensation analysis module is used to perform preliminary linear compensation on the sensor output signal on the terminal side and learn the nonlinear characteristics of the sensor in the cloud to train the response compensation model;

[0010] The cloud-based compensation module is used to perform nonlinear response compensation on the uploaded PEDOT sensor data based on the trained response compensation model, thereby generating a compensated output signal, significantly improving the measurement precision and accuracy of the sensor and reducing the error introduced by the nonlinear response.

[0011] The compensation deviation analysis module is used to analyze the deviation degree of the signal before and after compensation, compare it with the preset supplementary deviation threshold, and analyze whether it meets the response supplement requirements;

[0012] The feedback adjustment module is used to determine whether the response compensation model needs to be optimized and updated based on the compensation deviation analysis results to adapt to changes in sensor performance and fluctuations in environmental conditions and maintain high-precision measurement performance.

[0013] A further improvement of the technical solution of the present invention is that: the response compensation analysis module includes a terminal-side preliminary compensation unit, a cloud-side transmission unit, and a model training optimization unit;

[0014] The device-side preliminary compensation unit is used to deploy edge devices and run a lightweight compensation algorithm to perform preliminary linear compensation on the sensor's output signal.

[0015] The cloud transmission unit is used to upload the data preliminarily processed by the terminal side to the cloud management center to perform data exchange between the terminal and the cloud to ensure the integrity and security of the data;

[0016] The model training optimization unit is used to train the response compensation model and learn the nonlinear characteristics of the sensor by utilizing historical data and the basic architecture of the neural network model.

[0017] A further improvement of the technical solution of the present invention is that the data acquisition module specifically includes:

[0018] Establish a communication connection with the PEDOT sensor through a high-precision data acquisition interface, collect the original sensor output signal in real time according to the preset sampling frequency, and integrate multiple environmental sensors to synchronously obtain environmental parameter data, including temperature, humidity and electromagnetic interference, while collecting the sensor output signal;

[0019] The collected environmental parameter data and the sensor output signal are time-aligned through timestamps to ensure data consistency and relevance. The sensor is also calibrated, including checking and adjusting the sensor's zero point, range, and linearity to eliminate the sensor's own systematic errors.

[0020] The raw data of the PEDOT sensor after time alignment and calibration are preprocessed, including data cleaning and normalization.

[0021] A further improvement of the technical solution of the present invention is that the feature extraction module specifically includes:

[0022] Reading data from the preprocessed sensor output signal and environmental parameter data, and performing feature analysis on the sensor output signal and environmental parameter data to extract response compensation features required for sensor nonlinear response compensation optimization, wherein the response compensation features include signal features and environmental features;

[0023] Signal characteristics include output signal amplitude, signal drift, and signal nonlinear error rate; environmental characteristics include temperature change rate, humidity change rate, and electromagnetic interference intensity;

[0024] According to the requirements of sensor nonlinear response compensation optimization, the baseline values of each signal feature and environmental feature are set, and then the extracted signal features and environmental features are integrated to form a complete response compensation feature set.

[0025] A further improvement of the technical solution of the present invention is that the terminal-side preliminary compensation unit specifically includes:

[0026] Deploy edge devices in the target environment, perform initial configuration of hardware and software, set initial values of compensation parameters, load lightweight compensation algorithms, and establish communication connections with sensors;

[0027] The edge device continuously collects the raw output signal of the sensor and applies a lightweight compensation algorithm to perform preliminary linear compensation;

[0028] After initial compensation, the quality of the compensated data is evaluated using edge devices. The evaluation process includes checking whether the compensated signal meets the expected linearity and stability requirements. If the data quality does not meet the expected standards, the edge device automatically adjusts the compensation parameters through a real-time feedback mechanism.

[0029] A further improvement of the technical solution of the present invention is that the cloud transmission unit specifically includes:

[0030] After completing preliminary data processing, the edge device on the end side encapsulates the compensated sensor data, quality assessment indicators, and related metadata in a predetermined format. The encapsulation process uses efficient serialization protocols such as JSON or Protobuf to ensure that the data is structured and compact. The data is then encrypted using the AES-256 symmetric encryption algorithm combined with the device's unique key to generate ciphertext data. A digital signature is also added, and the message digest is signed using the device's private key to ensure the authenticity and integrity of the data source.

[0031] Based on the network environment and reliability requirements, the MQTT transport protocol is selected and a secure connection is established between the edge device on the end and the cloud management center. The legitimacy of the cloud certificate is verified through a TLS handshake, and the encryption algorithm and key are negotiated. After the connection is established, the edge device on the end registers and authenticates its identity with the cloud management center, obtaining a dedicated session token to ensure the legitimacy and uniqueness of subsequent data transmission and prevent unauthorized access.

[0032] The encrypted data is sent to the cloud management center via the MQTT transport protocol. During transmission, a fragmented transmission and confirmation mechanism is used to divide the data into fixed-size packets. Each packet is accompanied by a sequence number. The cloud management center returns a confirmation message after receiving it. If the client does not receive the confirmation within the timeout period, the packet is automatically retransmitted.

[0033] After the cloud management center receives the data uploaded by the end side, it verifies the digital signature, decrypts the signature using the device public key and compares the message digest to confirm that the data source is legal and has not been tampered with. It then performs AES-256 decryption on the ciphertext data to restore the original data. It then performs data format verification and logic verification to check the integrity and rationality of the data fields. After verification, the data is stored in the cloud database and a successful reception confirmation message is returned to the end side. If the verification fails, the cloud management center discards the data and notifies the end side to re-upload it to ensure the accuracy and security of the cloud-stored data.

[0034] A further improvement of the technical solution of the present invention is that the model training optimization unit specifically includes:

[0035] Historical data is obtained from the cloud database, including sensor output signals that have undergone preliminary compensation and quality assessment on the device side, as well as relevant environmental parameter data. After data acquisition, it is cleaned and preprocessed to remove outliers and noise, ensure data quality and consistency, and then integrated into a complete dataset. The preprocessing process includes data normalization and feature scaling to convert the data into a format suitable for neural network model training. The dataset is then divided into training, validation, and test sets in a ratio of 7:1.5:1.5 to support model training, validation, and performance evaluation.

[0036] According to the nonlinear characteristics of the sensor, a convolutional neural network model-based infrastructure is selected to train the response compensation model and perform model initialization settings;

[0037] The convolutional neural network model is trained using the training set. The model output is calculated through forward propagation, and the MSE loss is calculated by comparing it with the true value. The gradient is calculated using the backpropagation algorithm, and the weight parameters are updated using the Adam optimizer. After training for 5 epochs, the model performance is evaluated on the validation set, and the loss curve and R are monitored. 2The score and mean absolute error (MAE) indicators are used. When the validation set loss does not decrease for 10 consecutive times, the early stopping mechanism is triggered. If there are signs of overfitting (training loss is significantly lower than validation loss), the learning rate decay strategy is adjusted or a dropout layer is added. Training is continued until the preset upper limit of 100 iterations is reached to ensure the model's generalization performance on unseen data.

[0038] The trained model was subjected to structured pruning to remove redundant convolution kernels and neurons. A comprehensive evaluation was performed on the test set, and cross-validation was used to calculate measurement accuracy (RMSE ≤ 0.5%), response time (≤ 50ms), and resource consumption (model size ≤ 5MB). This resulted in a trained response compensation model that accurately compensated for the nonlinear error in the sensor output, significantly improving measurement accuracy.

[0039] A further improvement of the technical solution of the present invention is that the cloud compensation module specifically includes:

[0040] Extract PEDOT sensor data uploaded from edge devices, including sensor output signals and related environmental parameters, and load a pre-trained response compensation model. Check the model version and compatibility to ensure that the model matches the current data and processing environment.

[0041] The PEDOT sensor data is input into the loaded response compensation model. The response compensation model performs nonlinear operations on the data based on its internal structure and parameters, gradually processes the data through the forward propagation process, and outputs the compensated signal value.

[0042] After the nonlinear response compensation calculation, the cloud compensation module generates a compensated output signal and performs a quality assessment on the generated signal to check whether it meets the expected precision and accuracy requirements. The compensated signal and related quality assessment indicators are encapsulated into a data packet in a standard format, which is fed back to the edge device and stored in the cloud database.

[0043] A further improvement of the technical solution of the present invention is that the compensation deviation analysis module specifically includes:

[0044] The compensation deviation analysis module obtains sensor signal data before and after compensation from the cloud database, performs integrity checks on the acquired data to ensure that there is no missing or damaged data, and performs time alignment on the data before and after compensation to ensure one-to-one correspondence between the data points.

[0045] By comparing the signal values before and after compensation point by point, the deviation of each data point is obtained by point-by-point subtraction, and the deviation is statistically analyzed to obtain the deviation value of the signal before and after compensation;

[0046] The calculated deviation value of the signal before and after compensation is compared with the preset compensation deviation threshold to determine whether the compensated signal meets the response supplement requirements. If the deviation value exceeds the compensation deviation threshold, the data is marked as abnormal and further analysis and optimization processes are triggered.

[0047] A further improvement of the technical solution of the present invention is that the feedback adjustment module specifically includes:

[0048] Receive the deviation analysis results output by the compensation deviation analysis module, including the deviation values of the signals before and after compensation and abnormal marking information, and verify the timestamp and metadata of the data to ensure the timeliness and relevance of the data;

[0049] Based on the deviation analysis results after analysis, the performance of the response compensation model is evaluated, and the compensation accuracy indicators of the model are calculated, including the error rate and measurement accuracy after compensation. The compensation accuracy indicators are compared with the preset performance threshold to determine whether the model meets the high-precision measurement requirements. If the compensation accuracy indicators are all within the performance threshold range, it means that the current performance of the model is good and no optimization update is required. If any compensation accuracy indicator exceeds the performance threshold, it indicates that the model performance has deteriorated and optimization adjustment is required.

[0050] When it is determined that the response compensation model needs to be optimized and updated, the feedback adjustment module formulates a specific optimization strategy, analyzes the causes of the deviation, including changes in sensor performance and fluctuations in environmental conditions, determines the optimization direction, and selects the corresponding optimization method based on the analysis results, including adjusting model parameters, increasing training data, and improving model structure;

[0051] If it is decided to optimize and update the response compensation model, the feedback adjustment module will perform model update operations according to the optimization strategy, input the new training data into the response compensation model for retraining, adjust the model parameters, and improve the model's adaptability and compensation accuracy. After the update is completed, the new response compensation model will be verified for effectiveness, and the compensation deviation analysis will be performed again to evaluate the compensation accuracy index of the new response compensation model. If the performance of the new response compensation model meets the requirements, it will be deployed in actual applications. If the performance still does not meet the standards, return to the optimization strategy formulation step and readjust the optimization plan until the model performance meets the high-precision measurement requirements.

[0052] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared to the prior art:

[0053] 1. The present invention provides a PEDOT sensor nonlinear response compensation and optimization system. By combining the cloud with a neural network model, the nonlinear characteristics of the sensor are accurately learned and accurately compensated. Preliminary linear compensation is performed on the end side, reducing the amount of data transmitted to the cloud and improving the real-time performance of the system. At the same time, a large amount of historical data is used in the cloud for model training and optimization, significantly reducing the error introduced by the nonlinear response. The end-cloud collaboration method effectively improves the measurement precision and accuracy of the PEDOT sensor, enabling the sensor to provide highly reliable measurement data under different environmental conditions, meeting the needs of high-precision measurement.

[0054] 2. The present invention provides a PEDOT sensor nonlinear response compensation optimization system. Based on the compensation deviation analysis results, the response compensation model is dynamically adjusted to adapt to changes in sensor performance and fluctuations in environmental conditions. Through model optimization and updating, high-precision measurement performance is maintained, ensuring that the system can operate stably in various complex and changeable application scenarios, thereby improving the system's reliability and service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0056] Figure 1 This is a schematic diagram of the functional modules of the system of the present invention;

[0057] Figure 2 Schematic diagram of the working process of the cloud compensation module of the present invention. DETAILED DESCRIPTION

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0059] Example 1, as Figure 1 、 Figure 2 As shown, the present invention provides a PEDOT sensor nonlinear response compensation optimization system, including a cloud management center, the cloud management center is communicatively connected to a data acquisition module, a feature extraction module, a response compensation analysis module, a cloud compensation module, a compensation deviation analysis module, and a feedback adjustment module, wherein the modules are electrically connected;

[0060] The data acquisition module is used to collect and preprocess the raw data of the PEDOT sensor in real time, including the sensor output signal and environmental parameter data. It establishes a communication connection with the PEDOT sensor through a high-precision data acquisition interface, collects the raw sensor output signal in real time according to a preset sampling frequency, and integrates multiple environmental sensors to synchronously obtain environmental parameter data, including temperature, humidity, and electromagnetic interference. The collected environmental parameter data is time-aligned with the sensor output signal using timestamps to ensure data consistency and correlation. The sensor is calibrated, including checking and adjusting the sensor's zero point, range, and linearity to eliminate the sensor's own systematic errors. The raw data of the PEDOT sensor after time alignment and calibration is preprocessed, including data cleaning and normalization. A filtering algorithm is used to smooth the sensor output signal, remove high-frequency noise and random interference, identify and eliminate abnormal data points, including abnormal output caused by sensor failure or sudden environmental changes, and normalize the cleaned data to scale the data to the range of [0,1] to make different features comparable.

[0061] The feature extraction module is used to extract response compensation features from the preprocessed sensor output signal and environmental parameter data to obtain a response compensation feature set, read data from the preprocessed sensor output signal and environmental parameter data, and perform feature analysis on the sensor output signal and environmental parameter data to extract the response compensation features required for sensor nonlinear response compensation optimization. The response compensation features include signal features and environmental features. The signal features include output signal amplitude, signal drift, and signal nonlinear error rate. The output signal amplitude is the average amplitude of the sensor output signal, which reflects the response strength of the sensor to the measured physical quantity. The signal drift is the drift of the sensor output signal over time, which reflects the The stability of the sensor, the signal nonlinear error rate is the deviation ratio between the sensor output signal and the ideal linear output, quantifying the degree of nonlinearity. The environmental characteristics include temperature change rate, humidity change rate and electromagnetic interference intensity. The temperature change rate is the rate of change of the ambient temperature over time, reflecting the dynamic change of the ambient temperature. The humidity change rate is the rate of change of the ambient humidity over time, reflecting the dynamic change of the ambient humidity. The electromagnetic interference intensity is the intensity of the electromagnetic interference in the environment, reflecting the impact of the electromagnetic environment on the sensor. According to the requirements of the sensor nonlinear response compensation optimization, the baseline values of each signal feature and environmental feature are set, and then the extracted signal features and environmental features are integrated to form a complete response compensation feature set;

[0062] The response compensation analysis module performs preliminary linear compensation on the sensor output signal on the device side and learns the nonlinear characteristics of the sensor on the cloud side to train the response compensation model. The response compensation analysis module includes a preliminary compensation unit on the device side, a cloud transmission unit, and a model training and optimization unit.

[0063] Among them, the terminal side preliminary compensation unit is used to deploy edge devices and run a lightweight compensation algorithm to perform preliminary linear compensation on the output signal of the sensor to reduce the amount of data transmitted to the cloud, reduce communication costs, improve the real-time performance of the system, and reduce the pressure on cloud computing. The edge device is deployed in the target environment, the hardware and software are initialized and configured, the initial values of the compensation parameters are set, and the lightweight compensation algorithm is loaded to establish a communication connection with the sensor. Among them, the edge device selects an industrial-grade edge computing device (NVIDIA Jetson Nano / Raspberry Pi 4B), IP67 protection grade housing, wide operating temperature range (-40℃~85℃), sensor interface supports 4-20mA / 0-10V analog input, RS485 / CAN bus communication, operating system is customized Linux real-time system (RT-Preempt patch), and YAML file is used to define sensor parameters (range / resolution / calibration coefficient). The edge device continuously collects the original output signal of the sensor and applies a lightweight compensation algorithm for preliminary linear compensation. The compensation process reduces the nonlinear characteristics of the sensor output by adjusting the amplitude, drift and nonlinear error rate of the sensor output signal. The lightweight compensation algorithm uses preset compensation parameters and real-time data to quickly calculate the The compensated signal value significantly reduces the amount of data that needs to be transmitted to the cloud, lowering communication costs while improving the real-time performance of the system. After initial compensation, the quality of the compensated data is evaluated using edge devices. The evaluation process includes checking whether the compensated signal meets the expected linearity and stability requirements. If the data quality does not meet the expected standards, the edge device automatically adjusts the compensation parameters through a real-time feedback mechanism to ensure that the compensation algorithm can adapt to the dynamic changes in the sensor output and optimize the compensation effect. Among them, linearity detection is achieved by calculating the Pearson correlation coefficient, with a target of ρ ≥ 0.995. Stability detection is achieved by sliding window standard deviation, with a threshold of σ < 0.05. The dynamic response is a step response rise time < 100ms.

[0064] The cloud transmission unit is used to upload the data after preliminary processing on the end side to the cloud management center, conduct data interaction between the end and the cloud, and ensure the integrity and security of the data. After completing the preliminary data processing, the edge device on the end side will encapsulate the compensated sensor data, quality assessment indicators and related metadata in a predetermined format. The encapsulation process uses efficient serialization protocols such as JSON or Protobuf to ensure that the data is structured and compact, and encrypts the data. The AES-256 symmetric encryption algorithm is combined with the device's unique key to generate ciphertext data, and a digital signature is attached. The message digest is signed with the device's private key to ensure the authenticity and integrity of the data source. According to the network environment and reliability requirements, the MQTT transmission protocol is selected, and a secure connection is established between the edge device on the end side and the cloud management center. The legitimacy of the cloud certificate is verified through a TLS handshake, and the encryption algorithm and key are negotiated. After the connection is established, the edge device on the end side registers with the cloud management center and authenticates its identity. Obtain a dedicated session token to ensure the legitimacy and uniqueness of subsequent data transmission, prevent illegal device access, and send the encrypted data to the cloud management center via the MQTT transmission protocol. During the transmission process, a fragmented transmission and confirmation mechanism is used to divide the data into fixed-size data packets. Each packet is accompanied by a serial number. After receiving it, the cloud management center returns a confirmation message. If the end side does not receive the confirmation within the timeout period, it automatically retransmits the data packet. After receiving the data uploaded by the end side, the cloud management center verifies the digital signature, uses the device public key to decrypt the signature and compares the message digest to confirm that the data source is legal and has not been tampered with. Then, the ciphertext data is decrypted with AES-256 to restore the original data. Through data format verification and logical verification, the integrity and rationality of the data fields are checked. After verification, the data is stored in the cloud database and a successful reception confirmation message is returned to the end side. If the verification fails, the cloud management center discards the data and notifies the end side to re-upload it to ensure the accuracy and security of the cloud-stored data.

[0065] The model training optimization unit is used to use historical data and the infrastructure of the neural network model to train the response compensation model, learn the nonlinear characteristics of the sensor, and continuously optimize the model parameters to accurately compensate for the nonlinear error of the sensor output, significantly improving the measurement accuracy. Historical data is obtained from the cloud database, including the sensor output signal and related environmental parameter data that have undergone preliminary compensation and quality assessment on the end side. After the data is acquired, it is cleaned and preprocessed to remove outliers and noise, ensure the quality and consistency of the data, and integrate it into a complete data set. The preprocessing process includes data normalization and feature scaling, converting the data into a format suitable for neural network model training, and then dividing the data set into training set, validation set and test set in a ratio of 7:1.5:1.5 to support model training, validation and performance evaluation. According to the nonlinear characteristics of the sensor, the infrastructure based on the convolutional neural network model is selected to train the response compensation model, and then Initialization settings of the model, when designing the network layer structure, use one-dimensional convolution layer to capture time series features, combine with pooling layer to reduce dimension, fully connected layer to realize feature fusion, set ReLU activation function to introduce nonlinear mapping capability, use He initialization or Xavier initialization method to generate initial weights, select mean square error (MSE) as loss function to measure prediction deviation, use Adam optimizer to achieve efficient parameter update, adjust hyperparameters according to data characteristics: set initial learning rate to 0.001, batch size to 64, introduce L2 regularization parameter 0.0001 to prevent overfitting, ensure that the model has good generalization ability, use training set to train convolutional neural network model, calculate model output through forward propagation, compare with true value to calculate MSE loss, use back propagation algorithm to calculate gradient, update weight parameters through Adam optimizer, evaluate model performance on validation set after every 5 epochs, monitor loss curve and R 2 Score and mean absolute error (MAE) indicators. When the validation set loss does not decrease for 10 consecutive times, the early stopping mechanism is triggered. If there are signs of overfitting (training loss is significantly lower than validation loss), the learning rate decay strategy is adjusted or a dropout layer is added. Training is continued until the preset upper limit of 100 iterations is reached to ensure the generalization performance of the model on unseen data. Structured pruning is performed on the trained model to remove redundant convolution kernels and neurons. 8-bit quantization technology is used to compress the model volume. The original model capabilities are transferred to a lightweight model through the knowledge distillation method to reduce computing resource consumption. A comprehensive evaluation is performed on the test set. The cross-validation method is used to calculate the measurement accuracy (RMSE ≤ 0.5%), response time (≤ 50ms) and resource consumption (model size ≤ 5MB). The trained response compensation model is obtained to accurately compensate for the nonlinear error of the sensor output, significantly improving the measurement accuracy.

[0066] The cloud-based compensation module is used to perform nonlinear response compensation on uploaded PEDOT sensor data based on a trained response compensation model, thereby generating a compensated output signal. This significantly improves the sensor's measurement precision and accuracy, and reduces errors introduced by nonlinear response.

[0067] The compensation deviation analysis module is used to analyze the deviation degree of the signal before and after compensation, compare it with the preset supplementary deviation threshold, and analyze whether it meets the response supplement requirements;

[0068] The feedback adjustment module is used to combine the compensation deviation analysis results to determine whether the response compensation model needs to be optimized and updated to adapt to changes in sensor performance and fluctuations in environmental conditions, thereby maintaining high-precision measurement performance.

[0069] Example 2, as Figure 1 、 Figure 2 As shown, based on Example 1, the present invention provides a technical solution: preferably, the cloud compensation module specifically includes:

[0070] The system extracts PEDOT sensor data uploaded from the edge device, including the sensor output signal and related environmental parameters, and loads a pre-trained response compensation model. The model version and compatibility are checked to ensure that the model matches the current data and processing environment. The PEDOT sensor data is then fed into the loaded response compensation model. Based on its internal structure and parameters, the response compensation model performs nonlinear operations on the data, progressively processing the data through a forward propagation process and outputting compensated signal values. Simultaneously, the system monitors the computational status in real time, including CPU / GPU utilization and memory usage, to detect possible anomalies, including numerical overflow and vanishing gradients. Once an anomaly is detected, a pre-set fault-tolerance mechanism is immediately activated to adjust the computational accuracy and restart the computational process, ensuring the stability and reliability of the compensation computation. After the nonlinear response compensation computation, the cloud-based compensation module generates a compensated output signal and performs a quality assessment on the generated signal to check whether it meets the expected precision and accuracy requirements. The compensated signal and related quality assessment metrics are then packaged into a standard data packet, which is fed back to the edge device and stored in a cloud database.

[0071] The compensation deviation analysis module specifically includes:

[0072] The compensation deviation analysis module obtains sensor signal data before and after compensation from the cloud database and performs an integrity check on the obtained data to ensure that the data is not missing or damaged. It also time-aligns the data before and after compensation to make the data points correspond one to one. By comparing the signal values before and after compensation point by point, the deviation of each data point is obtained by point-by-point subtraction. The deviation is statistically analyzed to obtain the deviation value of the signal before and after compensation. The calculated deviation value of the signal before and after compensation is compared with the preset compensation deviation threshold to determine whether the compensated signal meets the response supplement requirements. If the deviation value exceeds the compensation deviation threshold, the data is marked as abnormal, and further analysis and optimization processes are triggered.

[0073] The feedback adjustment module specifically includes:

[0074] Receive the deviation analysis results output by the compensation deviation analysis module, including the deviation values of the signals before and after compensation and the abnormal marking information, and verify the timestamp and metadata of the data to ensure the timeliness and relevance of the data. Based on the parsed deviation analysis results, evaluate the performance of the response compensation model, calculate the compensation accuracy index of the model, including the error rate and measurement accuracy after compensation, and compare the compensation accuracy index with the preset performance threshold to determine whether the model meets the high-precision measurement requirements. If the compensation accuracy indicators are all within the performance threshold range, it means that the current performance of the model is good and no optimization update is required. If there are compensation accuracy indicators that exceed the performance threshold, it indicates that the model performance has declined and optimization adjustment is required. When it is determined that the response compensation model needs to be optimized and updated, the feedback adjustment module formulates a specific optimization strategy and analyzes the cause of the deviation. , including changes in sensor performance, fluctuations in environmental conditions, etc., determine the optimization direction, and select the corresponding optimization method based on the analysis results, including adjusting model parameters, increasing training data and improving model structure. If it is decided to optimize and update the response compensation model, the feedback adjustment module performs model update operations according to the optimization strategy, inputs the new training data into the response compensation model for retraining, adjusts the model parameters, and improves the adaptability and compensation accuracy of the model. After the update is completed, the effect of the new response compensation model is verified, and the compensation deviation analysis is performed again to evaluate the compensation accuracy index of the new response compensation model. If the performance of the new response compensation model meets the requirements, it will be deployed in actual applications. If the performance still does not meet the standards, return to the optimization strategy formulation step and readjust the optimization plan until the model performance meets the high-precision measurement requirements.

[0075] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A PEDOT sensor nonlinear response compensation optimization system, including a cloud management center, characterized by: The cloud management center is communicatively connected to a data acquisition module, a feature extraction module, a response compensation analysis module, a cloud compensation module, a compensation deviation analysis module, and a feedback adjustment module, wherein electrical signal connections are made between the modules; The data acquisition module is used to collect and pre-process the raw data of the PEDOT sensor in real time, including the sensor output signal and environmental parameter data; The feature extraction module is used to extract response compensation features from the pre-processed sensor output signal and environmental parameter data to obtain a response compensation feature set; The response compensation analysis module is used to perform preliminary linear compensation on the sensor output signal on the terminal side and learn the nonlinear characteristics of the sensor in the cloud to train the response compensation model; The cloud-based compensation module is used to perform nonlinear response compensation on the uploaded PEDOT sensor data based on the trained response compensation model, thereby generating a compensated output signal; The compensation deviation analysis module is used to analyze the deviation degree of the signal before and after compensation, compare it with the preset supplementary deviation threshold, and analyze whether it meets the response supplement requirements; The feedback adjustment module is used to determine whether the response compensation model needs to be optimized and updated based on the compensation deviation analysis results.

2. A PEDOT sensor nonlinear response compensation optimization system according to claim 1, characterized in that: The response compensation analysis module includes a terminal-side preliminary compensation unit, a cloud-side transmission unit, and a model training optimization unit; The device-side preliminary compensation unit is used to deploy edge devices and run a lightweight compensation algorithm to perform preliminary linear compensation on the sensor's output signal. The cloud transmission unit is used to upload the data preliminarily processed by the terminal side to the cloud management center to perform data exchange between the terminal and the cloud; The model training optimization unit is used to train the response compensation model and learn the nonlinear characteristics of the sensor by utilizing historical data and the basic architecture of the neural network model.

3. The PEDOT sensor nonlinear response compensation optimization system according to claim 1, characterized in that: The data acquisition module specifically includes: Establish a communication connection with the PEDOT sensor through a high-precision data acquisition interface, collect the original sensor output signal in real time according to the preset sampling frequency, and integrate multiple environmental sensors to synchronously obtain environmental parameter data, including temperature, humidity and electromagnetic interference, while collecting the sensor output signal; The collected environmental parameter data is time-aligned with the sensor output signal through the timestamp, and the sensor is calibrated, including checking and adjusting the sensor's zero point, range, and linearity; The raw data of the PEDOT sensor after time alignment and calibration are preprocessed, including data cleaning and normalization.

4. The PEDOT sensor nonlinear response compensation optimization system according to claim 1, characterized in that: The feature extraction module specifically includes: Reading data from the preprocessed sensor output signal and environmental parameter data, and performing feature analysis on the sensor output signal and environmental parameter data to extract response compensation features required for sensor nonlinear response compensation optimization, wherein the response compensation features include signal features and environmental features; Signal characteristics include output signal amplitude, signal drift, and signal nonlinear error rate; environmental characteristics include temperature change rate, humidity change rate, and electromagnetic interference intensity; According to the requirements of sensor nonlinear response compensation optimization, the baseline values of each signal feature and environmental feature are set, and then the extracted signal features and environmental features are integrated to form a complete response compensation feature set.

5. The PEDOT sensor nonlinear response compensation optimization system according to claim 2, characterized in that: The terminal-side preliminary compensation unit specifically includes: Deploy edge devices in the target environment, perform initial configuration of hardware and software, set initial values of compensation parameters, load lightweight compensation algorithms, and establish communication connections with sensors; The edge device continuously collects the raw output signal of the sensor and applies a lightweight compensation algorithm to perform preliminary linear compensation; After initial compensation, the quality of the compensated data is evaluated using edge devices. The evaluation process includes checking whether the compensated signal meets the expected linearity and stability requirements. If the data quality does not meet the expected standards, the edge device automatically adjusts the compensation parameters through a real-time feedback mechanism.

6. The PEDOT sensor nonlinear response compensation optimization system according to claim 2, characterized in that: The cloud transmission unit specifically includes: After completing preliminary data processing, the edge device on the end side encapsulates the compensated sensor data, quality assessment indicators, and related metadata in a predetermined format and encrypts the data. It uses the AES-256 symmetric encryption algorithm combined with the device's unique key to generate ciphertext data, append a digital signature, and sign the message digest using the device's private key. Based on the network environment and reliability requirements, the MQTT transport protocol is selected and a secure connection is established between the edge device on the end and the cloud management center. The legitimacy of the cloud certificate is verified through a TLS handshake, and the encryption algorithm and key are negotiated. After the connection is established, the edge device on the end registers and authenticates its identity with the cloud management center, obtaining a dedicated session token. The encrypted data is sent to the cloud management center via the MQTT transport protocol. During transmission, a fragmented transmission and confirmation mechanism is used to divide the data into fixed-size packets. Each packet is accompanied by a sequence number. The cloud management center returns a confirmation message after receiving it. If the client does not receive the confirmation within the timeout period, the packet is automatically retransmitted. After the cloud management center receives the data uploaded by the client, it verifies the digital signature, decrypts the signature using the device public key and compares the message digest. It then performs AES-256 decryption on the ciphertext data to restore the original data. After verification, the data is stored in the cloud database and a successful reception confirmation message is returned to the client. If verification fails, the cloud management center discards the data and notifies the client to re-upload it.

7. The PEDOT sensor nonlinear response compensation optimization system according to claim 2, characterized in that: The model training optimization unit specifically includes: Historical data is obtained from the cloud database, including sensor output signals that have undergone preliminary compensation and quality assessment on the device side, as well as relevant environmental parameter data. After data acquisition, it is cleaned and preprocessed, and integrated into a complete dataset. The dataset is then divided into training, validation, and test sets in a ratio of 7:1.5:1.

5. According to the nonlinear characteristics of the sensor, a convolutional neural network model-based infrastructure is selected to train the response compensation model and perform model initialization settings; The convolutional neural network model is trained using the training set. The model output is calculated through forward propagation, and the MSE loss is calculated by comparing it with the true value. The gradient is calculated using the backpropagation algorithm, and the weight parameters are updated using the Adam optimizer. After training for 5 epochs, the model performance is evaluated on the validation set, and the loss curve and R are monitored. 2 Score and mean absolute error indicators. When the validation set loss does not decrease for 10 consecutive times, the early stopping mechanism is triggered. If there are signs of overfitting, the learning rate decay strategy is adjusted or a dropout layer is added. Training continues until the preset upper limit of 100 iterations is reached. The trained model is subjected to structured pruning to remove redundant convolution kernels and neurons. A comprehensive evaluation is performed on the test set. The cross-validation method is used to calculate the measurement accuracy, response time and resource consumption. The trained response compensation model is obtained to accurately compensate for the nonlinear error of the sensor output.

8. The PEDOT sensor nonlinear response compensation optimization system according to claim 2, characterized in that: The cloud compensation module specifically includes: Extract PEDOT sensor data uploaded from edge devices, including sensor output signals and related environmental parameters, and load pre-trained response compensation models; The PEDOT sensor data is input into the loaded response compensation model. The response compensation model performs nonlinear operations on the data based on its internal structure and parameters, gradually processes the data through the forward propagation process, and outputs the compensated signal value. After the nonlinear response compensation calculation, the cloud compensation module generates a compensated output signal and performs a quality assessment on the generated signal to check whether it meets the expected precision and accuracy requirements. The compensated signal and related quality assessment indicators are encapsulated into a data packet in a standard format, which is fed back to the edge device and stored in the cloud database.

9. The PEDOT sensor nonlinear response compensation optimization system according to claim 8, characterized in that: The compensation deviation analysis module specifically includes: The compensation deviation analysis module obtains sensor signal data before and after compensation from the cloud database, performs integrity checks on the acquired data, and time-aligns the data before and after compensation to ensure one-to-one correspondence between the data points. By comparing the signal values before and after compensation point by point, the deviation of each data point is obtained by point-by-point subtraction, and the deviation is statistically analyzed to obtain the deviation value of the signal before and after compensation; The calculated deviation value of the signal before and after compensation is compared with the preset compensation deviation threshold to determine whether the compensated signal meets the response supplement requirements. If the deviation value exceeds the compensation deviation threshold, the data is marked as abnormal and further analysis and optimization processes are triggered.

10. The PEDOT sensor nonlinear response compensation optimization system according to claim 9, characterized in that: The feedback adjustment module specifically includes: Receive the deviation analysis results output by the compensation deviation analysis module, including the deviation values of the signals before and after compensation and abnormal marking information, and verify the timestamp and metadata of the data; Based on the deviation analysis results after analysis, the performance of the response compensation model is evaluated, and the compensation accuracy indicators of the model are calculated, including the error rate and measurement accuracy after compensation. The compensation accuracy indicators are compared with the preset performance threshold to determine whether the model meets the high-precision measurement requirements. If the compensation accuracy indicators are all within the performance threshold range, it means that the current performance of the model is good and no optimization update is required. If any compensation accuracy indicator exceeds the performance threshold, it indicates that the model performance has deteriorated and optimization adjustment is required. When it is determined that the response compensation model needs to be optimized and updated, the feedback adjustment module formulates a specific optimization strategy, analyzes the cause of the deviation, determines the optimization direction, and selects the corresponding optimization method based on the analysis results, including adjusting model parameters, increasing training data, and improving model structure; If it is decided to optimize and update the response compensation model, the feedback adjustment module will perform model update operations according to the optimization strategy, input the new training data into the response compensation model for retraining, and adjust the model parameters. After the update is completed, the effect of the new response compensation model will be verified, and the compensation deviation analysis will be performed again to evaluate the compensation accuracy index of the new response compensation model. If the performance of the new response compensation model meets the requirements, it will be deployed in actual applications. If the performance still does not meet the standards, return to the optimization strategy formulation step and readjust the optimization plan until the model performance meets the high-precision measurement requirements.

Images