Intelligent self-adaptive environment perception sensor

Abstract

The invention relates to an intelligent self-adaptive environment perception sensor. The method comprises the following steps: detecting and reading initial environment parameters of a current sensor; analyzing the working state of the sensor according to the preliminary environmental parameters; and according to the working state, sensor detection parameters are adjusted, and corrected environment parameters are obtained. The method solves the problem of low detection precision caused by incapability of adapting to an extreme environment in the prior art. A Bayesian network model of environmental parameters and working states of the sensor is constructed to quantify the relation between the extreme environment and the anomaly, so that accurate analysis of detection anomaly in the extreme environment is realized, and a foundation is laid for correcting a detection result. And a generalized additive model between the probability that the working state is abnormal and the detection deviation is established, so that the accuracy of correction of the detection deviation is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensors and relates to an intelligent self-adaptive environment perception sensor. Background Art

[0002] Sensor technology stems from people's relentless pursuit of accurately sensing and measuring their surroundings. Early on, in simple industrial production and scientific research, people realized the need for equipment to acquire basic data such as temperature and pressure to ensure stable production and accurate experiments. With technological advancements, traditional mechanical or simple electronic methods are no longer able to meet the precision and functionality requirements. The development of electronic components such as vacuum tubes and transistors provided the hardware foundation for sensor technology, while the rise of integrated circuit technology enabled the miniaturization and intelligence of sensors. In daily life, from automatic temperature-controlled home appliances to automotive safety systems, the demand for sensors has increased dramatically across various fields, prompting continuous innovation in sensor technology to adapt to diverse scenarios.

[0003] However, existing sensors operate in direct contact with the environment being measured and are therefore also affected by it. When exposed to extreme environmental conditions, sensors often fail to adapt, resulting in low accuracy. Therefore, existing technologies require sensors that can sense their operating status in real time and automatically adjust detection strategies and parameters based on environmental changes. This allows them to maintain high detection accuracy in various extreme environments, meeting the diverse demands for reliable sensor operation in complex and harsh environments. Summary of the Invention

[0004] In order to solve the above-mentioned problems existing in the prior art, the present invention provides an intelligent adaptive environmental perception sensor, which aims to analyze the detection deviation of environmental parameters under extreme environmental conditions through historical sensors, so as to solve the problem of low accuracy caused by the inability of prior art to adapt to extreme environmental conditions.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The present application provides an intelligent adaptive environment perception sensor, comprising an environment detection module, an environment analysis module, and a parameter adjustment module, wherein the environment detection module, the environment analysis module, and the parameter adjustment module are communicatively connected, wherein:

[0007] The environmental detection module is used to detect and read the preliminary environmental parameters of the current sensor;

[0008] The environmental analysis module is used to analyze the working status of the sensor according to preliminary environmental parameters;

[0009] The parameter adjustment module is used to adjust the sensor detection parameters according to the working state to obtain the correction environment parameters.

[0010] Furthermore, the environment detection module includes a main detection unit and several redundant detection units, and specifically works as follows:

[0011] When the preliminary environmental parameters detected do not exceed the set environmental tolerance threshold, the detection results of the main detection unit are adopted;

[0012] When the preliminary environmental parameters detected exceed the set environmental tolerance threshold, all redundant detection units are used for measurement, and the optimal detection results are analyzed through the multi-layer perceptron model.

[0013] Furthermore, analyzing the optimal detection result by the multi-layer perceptron model includes the following steps:

[0014] S1. Determine initial weights: Determine initial weights for each redundant detection unit by comprehensively considering the hardware characteristics of the redundant detection unit, the accuracy of past detection data, and stability under different environmental conditions;

[0015] S2. Feature extraction and data preprocessing: Fourier transform and wavelet transform methods are used to extract key features from the multi-parameter environmental detection data obtained by redundant detection units; the data is normalized to unify the data of different parameter ranges to the same scale;

[0016] S3. Build a neural network model: Build a multi-layer perceptron model so that the number of nodes in the input layer corresponds to the number of extracted features. Adjust the number of layers and nodes in the hidden layer based on data complexity and computing resources, and use nonlinear activation functions to transform the input.

[0017] S4. Model training and optimization: The pre-processed data features of each redundant detection unit are used as input, and the known accurate environmental parameters are used as labels, which are input into the neural network model for training. The error between the model prediction results and the labels is calculated through the back-propagation algorithm, and the weights of each layer of the network are adjusted according to the error. Optimization algorithms such as stochastic gradient descent are used, and the training process is repeated until the error converges to the preset value.

[0018] S5. Determine the optimal detection result: input the detection data features of the current redundant detection unit into the trained and optimized neural network model to obtain the result output by the model.

[0019] Furthermore, analyzing the working state of the sensor according to the preliminary environmental parameters includes the following steps:

[0020] Obtaining samples of environmental parameter detection datasets in historical periods and the working status of corresponding sensors, where the working status is divided into normal and abnormal;

[0021] The k-means clustering method was used to divide the data set into several environmental levels;

[0022] Taking the environmental level as the explanatory variable and the corresponding working status as the response variable, an abnormal probability analysis model is constructed;

[0023] Obtain the environmental level of the preliminary environmental parameters, input the abnormal probability analysis model, and analyze the probability that the working state of the sensor is abnormal.

[0024] Furthermore, the k-means clustering method comprises the following steps:

[0025] K1. Determine the number of clusters: Based on the characteristics of the data, use empirical rules or the elbow method to assist in determining the number of clusters k;

[0026] K2, randomly initialize k cluster centers: randomly select k data points from the environmental parameter detection data set as the initial cluster centers as the clustering starting points;

[0027] K3, assign data points to the nearest cluster center: calculate the distance between each data point in the data set and k cluster centers, assign the data points to the corresponding cluster according to the principle of the closest distance, and form a preliminary clustering result;

[0028] K4, update cluster center: calculate the mean of all data points in each cluster and use it as the new cluster center;

[0029] K5. Repeat steps K3 and K4 until convergence: Repeat the data point allocation and cluster center update steps until the cluster center change is less than the threshold or the maximum number of iterations is reached, completing the clustering of the data set.

[0030] Furthermore, the abnormal probability analysis model is configured as a Bayesian network model, comprising the following steps:

[0031] Define nodes and variables: Define the nodes of the Bayesian network model, namely the environmental level and sensor working status, and assign corresponding variables to each node to determine the model factors;

[0032] Determine the network structure: Based on the causal relationship between the environmental level and the sensor working status, construct the network structure, using directed edges to indicate the logical connection between nodes;

[0033] Estimated conditional probability table: For each node, the probability distribution is estimated based on historical data. The environmental level node determines the prior probability, and the sensor working status node determines the conditional probability under a given environmental level.

[0034] Model validation and optimization: Validate the model using some historical data, compare the prediction with the actual working status, and adjust the probability value or network structure based on the error index;

[0035] Input and reasoning: The current environmental level is input into the optimized model, and the Bayesian inference algorithm is combined with the conditional probability table to calculate the probability that the sensor's working state is abnormal.

[0036] Furthermore, the adjusting of the sensor detection parameters according to the working state to obtain the correction environment parameters includes the following steps:

[0037] Obtain the probability of abnormal working status in the historical period and the corresponding detection deviation data samples;

[0038] Establish a relationship transformation model between the probability of abnormal working status and detection deviation;

[0039] The probability that the current sensor working state is abnormal is input into the relational transformation model, and the detection deviation is output;

[0040] According to the output detection deviation, the detection parameters of the sensor are adjusted to obtain the correction environment parameters.

[0041] Furthermore, the relationship transformation model is configured as a generalized additive model, comprising the following steps:

[0042] Data preparation: Collect historical data on anomaly probability and detection deviation, clean it to remove noise, anomalies, and missing values, and standardize or normalize it;

[0043] Model structure setting: Use the generalized additive model to construct the relationship between abnormal probability and detection deviation, use the spline function as the smoothing function, and determine the spline type, number of nodes, and location;

[0044] Model fitting: Use an iterative algorithm to fit the model, adjust the smoothing function parameters to minimize the prediction error, select appropriate penalty parameters to balance fitting and complexity, and determine the optimal value through cross-validation;

[0045] Model evaluation: Use an independent validation set to evaluate the model and calculate the coefficient of determination to determine the model's prediction and generalization capabilities.

[0046] Furthermore, the generalized additive model is calculated as follows:

[0047]

[0048] In the formula, g(u y ) is an identity connection function with additive properties; u y is the response variable, i.e., the detection deviation; s(x i , k i ) is a smooth function, the smooth function is configured as a cubic spline function, x i is the explanatory variable, i.e., the abnormal probability, k iis the number of nodes of the cubic spline function, which controls the number of intervals within the numerical range of the explanatory variable. Each interval has a curve fitted by the cubic spline function, and n is the number of explanatory variables.

[0049] Beneficial effects of the present invention:

[0050] (1) By detecting and reading preliminary environmental parameters of the current sensor; analyzing the operating status of the sensor based on the preliminary environmental parameters; and adjusting the sensor detection parameters based on the operating status to obtain corrected environmental parameters. This invention solves the problem that the existing technology cannot adapt to extreme environments, resulting in low detection accuracy.

[0051] (2) A Bayesian network model of sensor environmental parameters and working status is constructed to quantify the relationship between extreme environment and anomalies, thereby achieving accurate analysis of anomalies detected in extreme environments and laying the foundation for correcting detection results.

[0052] (3) Establish a generalized additive model between the probability of abnormal working status and detection bias to improve the accuracy of inspection bias correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0054] Figure 1 This is a structural diagram of an intelligent adaptive environmental perception sensor in the present invention.

[0055] Figure 2 This is a flow chart of analyzing the working status of a sensor according to preliminary environmental parameters in one embodiment of the present invention.

[0056] Figure 3 This is a flow chart of adjusting sensor detection parameters according to working status to obtain corrected environmental parameters in one embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0058] See also Figure 1 The present application provides an intelligent adaptive environment perception sensor, comprising an environment detection module, an environment analysis module, and a parameter adjustment module, wherein the environment detection module, the environment analysis module, and the parameter adjustment module are communicatively connected, wherein:

[0059] The environmental detection module is used to detect and read the preliminary environmental parameters of the current sensor;

[0060] The environmental analysis module is used to analyze the working status of the sensor according to preliminary environmental parameters;

[0061] The parameter adjustment module is used to adjust the sensor detection parameters according to the working state to obtain the correction environment parameters.

[0062] Furthermore, the environment detection module includes a main detection unit and several redundant detection units, and specifically works as follows:

[0063] When the preliminary environmental parameters detected do not exceed the set environmental tolerance threshold, the detection results of the main detection unit are adopted;

[0064] When the preliminary environmental parameters detected exceed the set environmental tolerance threshold, all redundant detection units are used for measurement, and the optimal detection results are analyzed through the multi-layer perceptron model.

[0065] In this embodiment, in the intelligent adaptive environmental perception sensor, the environmental detection module is a key component for obtaining environmental information. It consists of a main detection unit and several redundant detection units. This design aims to comprehensively ensure the accuracy and reliability of the detection results.

[0066] When environmental conditions are relatively stable—that is, when the initial detected environmental parameters do not exceed the set tolerance threshold—the main detection unit fully utilizes its efficient and accurate detection advantages, and the sensor directly adopts its detection results. This is because the main detection unit has been carefully calibrated and optimized under normal conditions, and can output reliable detection data with extremely high efficiency, ensuring that the sensor can quickly and accurately perceive environmental information in most situations.

[0067] However, if the initial environmental parameters detected exceed the set environmental tolerance threshold, indicating that the sensor is experiencing extreme or complex environmental conditions, the detection accuracy of the primary detection unit may be affected. To address this challenge, all redundant detection units are immediately activated and put into measurement. These redundant detection units are not simply duplicates; they may use different detection principles, technical methods, or have unique performance characteristics. They collect environmental data from multiple dimensions in extreme environments, providing a rich and diverse source of information for subsequent analysis.

[0068] On this basis, a multi-layer perceptron (MLP) model is used to analyze the large amount of data generated by the redundant detection units to determine the optimal detection results. The MLP model is a powerful artificial neural network model consisting of an input layer, multiple hidden layers, and an output layer. The input layer receives detection data from the redundant detection units, while the hidden layers perform deep feature extraction and processing on the data through a series of nonlinear transformations, exploring the complex patterns and relationships hidden in the data. In this process, the model automatically learns the associations between different environmental parameters and detection results. Based on the processing results of the hidden layers, the output layer outputs an optimal detection result that comprehensively considers the information from all redundant detection units. Through analysis by the MLP model, data from multiple redundant detection units can be effectively integrated, overcoming the limitations of a single detection unit in extreme environments. This improves the accuracy and reliability of detection results in harsh environments and provides more accurate environmental parameter information for intelligent adaptive environmental perception sensors.

[0069] Furthermore, analyzing the optimal detection result by the multi-layer perceptron model includes the following steps:

[0070] S1. Determine initial weights: When using a neural network model to analyze optimal detection results, the first step is to determine initial weights for each redundant detection unit. This process requires comprehensive consideration of the redundant detection unit's hardware characteristics, the accuracy of past detection data, and its stability under different environmental conditions. For example, a redundant detection unit that consistently maintains high-precision detection in high-temperature environments is assigned a relatively high initial weight; while units with large fluctuations in detection accuracy are given a lower initial weight. These weights will reflect the contribution of each redundant detection unit to the final result in subsequent calculations.

[0071] S2. Feature extraction and data preprocessing: After each redundant detection unit obtains the detection data, it is necessary to extract features and preprocess the data. For environmental detection data, multiple parameters such as temperature, humidity, and gas concentration may be involved. Through mathematical methods such as Fourier transform and wavelet transform, key features in the data, such as change trends and periodicity, are extracted. At the same time, the data is normalized to unify data in different parameter ranges to the same scale. For example, data in different ranges such as temperature ranging from 0-100℃ and humidity ranging from 0-100% are mapped to the interval of 0-1 to eliminate the impact of data dimension differences on model calculations and ensure that the neural network can process data more effectively.

[0072] S3. Build a neural network model: Build a neural network model suitable for this task. In this embodiment, a multi-layer perceptron (MLP) is used. The number of nodes in the input layer of the model corresponds to the number of extracted features, and the number of layers and nodes in the hidden layer can be adjusted according to the complexity of the data and computing resources. The hidden layer transforms the input through a nonlinear activation function (such as the ReLU function) to learn complex patterns in the data. The output layer outputs a result representing the optimal estimate of the environmental parameters according to the task requirements. During the construction process, the hyperparameters of the network, such as the learning rate and the number of iterations, should be reasonably set to ensure that the model can converge quickly and accurately.

[0073] S4. Model Training and Optimization: The neural network model is trained using the collected, preprocessed data. During training, the data features of each redundant detection unit are used as input, along with known, accurate environmental parameters as labels. A backpropagation algorithm is used to calculate the error between the model's predictions and the labels. Based on this error, the weights of each layer in the network are adjusted to gradually reduce the error. This process is iterated until the model error converges to an acceptable range. During training, optimization algorithms such as stochastic gradient descent can be used to improve training efficiency and prevent the model from becoming trapped in a local optimum.

[0074] S5. Determine the optimal detection result: The trained and optimized neural network model is fed with the detection data features of the current redundant detection unit. The model outputs the optimal detection result after comprehensively considering all redundant detection units. This result incorporates the information of each redundant detection unit and, through neural network learning and processing, can more accurately reflect the true parameters of the current environment. Compared with traditional methods, neural network models can automatically learn complex relationships in data and demonstrate higher adaptability and accuracy when processing multi-source detection data in extreme environments, providing more reliable environmental parameter detection results for intelligent adaptive environmental perception sensors.

[0075] Furthermore, analyzing the working state of the sensor according to the preliminary environmental parameters includes the following steps:

[0076] Obtaining samples of environmental parameter detection datasets in historical periods and the working status of corresponding sensors, where the working status is divided into normal and abnormal;

[0077] This step lays the data foundation for subsequent analysis. By collecting environmental parameter data detected by sensors over a period of time, this data covers measurements under various environmental conditions, such as temperature, humidity, and light intensity. The sensor's operating status for each test is also recorded, categorized as either "normal" or "abnormal." These historical data samples serve as training material for the analysis model, helping us identify potential connections between environmental parameters and sensor operating conditions.

[0078] The k-means clustering method was used to divide the data set into several environmental levels;

[0079] K-means clustering is an unsupervised learning algorithm used to divide data points into distinct groups (clusters). Here, a historical environmental parameter detection dataset is used as input. The k-means algorithm then divides the entire dataset into several distinct groups based on the similarities between the data points, with each group representing an environmental level. For example, environmental data with similar parameters such as temperature and humidity might be grouped together into a single level. This allows complex and diverse environmental data to be categorized and organized, enabling subsequent analysis at different environmental level levels and facilitating the discovery of patterns in sensor operating conditions under varying environmental conditions.

[0080] Taking the environmental level as the explanatory variable and the corresponding working status as the response variable, an abnormal probability analysis model is constructed;

[0081] After the environmental levels are categorized, each environmental level is used as the explanatory variable (i.e., independent variable), and the corresponding sensor operating status (normal or abnormal) for that environmental level is used as the response variable (i.e., dependent variable). A model is constructed using statistical or machine learning methods. The goal of this model is to predict the probability of a sensor in an abnormal operating state given a given environmental level. For example, by analyzing information such as the frequency of abnormal sensor operating states at different environmental levels, a model can be constructed that reflects the relationship between environmental level and the probability of sensor abnormality. This model will become an important tool for determining the current sensor operating status.

[0082] Obtain the environmental level of the preliminary environmental parameters, input the abnormal probability analysis model, and analyze the probability that the working state of the sensor is abnormal.

[0083] For the preliminary environmental parameters detected by the sensor, the environmental class is first determined based on the previously established k-means clustering rules. This environmental class information is then fed into the established abnormality probability analysis model. Based on the previously learned relationship between environmental class and abnormal operating probability, the model calculates and outputs the probability that the sensor is operating abnormally under the current environmental conditions. This probability value provides a quantitative assessment of the sensor's current operating status, providing a basis for determining whether subsequent sensor adjustments or maintenance are necessary.

[0084] Furthermore, the k-means clustering method comprises the following steps:

[0085] K1. Determining the number of clusters k: When using the k-means clustering method to process an environmental parameter detection dataset, the first task is to determine the number of clusters k—that is, the number of environmental levels into which the dataset should be divided. This decision typically requires consideration of specific business requirements and data characteristics. Rules of thumb can be employed, such as estimating the number of typical environmental states based on prior knowledge of the environmental parameter. Algorithms, such as the elbow method, can also be used to assist in determining this number. The elbow method calculates the sum of the distances from data points to their cluster centers (i.e., the sum of squared errors) for different values of k. As k increases, the sum of squared errors gradually decreases. When k reaches a certain value, the decrease in the sum of squared errors becomes minimal. This point, like the bend of an elbow, represents the k value corresponding to the appropriate number of clusters.

[0086] K2. Randomly initialize k cluster centers: After determining the number of clusters, k, k data points are randomly selected from the environmental parameter detection dataset as initial cluster centers. These cluster centers are continuously updated during subsequent iterations, ultimately representing the typical characteristics of each environmental level. The purpose of random initialization is to provide a starting point for the clustering process. However, due to the random selection, different initial cluster centers may produce different clustering results. However, through subsequent iterative optimization, the impact of this initial selection will gradually decrease.

[0087] K3. Assign data points to the nearest cluster center: For each environmental parameter data point in the dataset, calculate the distance between it and the k cluster centers. This distance calculation can use common distance metrics such as Euclidean distance. Based on the calculated distance, the data point is assigned to the cluster represented by the nearest cluster center. Through this step, each data point is classified into a specific environmental level, forming a preliminary clustering result.

[0088] K4. Update Cluster Centers: After all data points have been assigned, the centers of each cluster need to be updated. For each cluster, the mean of all data points in that cluster is calculated (or the mean for each dimension if multidimensional data is used), and this mean is used as the new cluster center. The purpose of updating cluster centers is to make them more representative of the characteristics of all data points in that cluster, thereby improving clustering accuracy.

[0089] K5. Repeat steps K3 and K4 until convergence: Repeat steps K3 (assigning data points to the nearest cluster center) and K4 (updating the cluster center) until convergence conditions are met. Convergence conditions can typically be that the cluster center no longer changes significantly, that is, the distance between the new cluster center and the cluster center of the previous iteration is less than a preset threshold; or that the preset maximum number of iterations has been reached. Through multiple iterations, the clustering results will gradually stabilize, ultimately resulting in the classification of the environmental parameter detection dataset into k environmental levels, providing an effective data classification foundation for the subsequent construction of anomaly probability analysis models.

[0090] Furthermore, the abnormal probability analysis model is configured as a Bayesian network model, comprising the following steps:

[0091] Define nodes and variables: Determine the nodes in the Bayesian network, primarily including environmental level nodes and sensor operating status nodes. Environmental level nodes use different environmental levels, such as low, medium, and high, as their status values, as determined by k-means clustering. Sensor operating status nodes use "normal" and "abnormal" as their status values. These nodes represent key factors in the model. Each node is assigned a corresponding variable to facilitate subsequent probability calculation and inference.

[0092] Determine the network structure: A Bayesian network structure is constructed based on the potential causal relationship between environmental level and sensor operating status. Generally speaking, changes in environmental level will affect sensor operating status. Therefore, the environmental level node is the parent node, the sensor operating status node is the child node, and directed edges are established from the environmental level node to the sensor operating status node. This structure reflects our initial understanding of the logical relationship between these two factors and provides a framework for subsequent probability calculations.

[0093] Estimating the conditional probability table: For each node, its conditional probability table (CPT) needs to be estimated. For environmental level nodes, the prior probability distribution, that is, the probability of each environmental level occurring, is estimated based on the frequency of each environmental level in the samples of the environmental parameter detection dataset over the historical period. For sensor operating status nodes, given different environmental levels, the conditional probability is determined based on the proportion of "normal" or "abnormal" sensor operating states at that environmental level in the historical data. For example, in the "high" environmental level, the probability of the sensor operating state being "abnormal" is 0.3, and the probability of being "normal" is 0.7. These conditional probability tables quantify the dependencies between nodes.

[0094] Model Validation and Optimization: Using a subset of historical data as a validation set and the environmental level as input, a Bayesian network model is used to calculate the probability of an abnormal sensor operating state and compare it with the actual operating state. Error metrics, such as precision and recall, are calculated between the predicted and actual results. The model is optimized based on these error metrics. Probability values in the conditional probability table can be adjusted, and the network structure can be fine-tuned, such as by adding or removing edges, to improve the model's prediction accuracy.

[0095] Input and Inference: The current environmental level of preliminary environmental parameters is obtained and input into the optimized Bayesian network model. Using a Bayesian inference algorithm, such as variable elimination or the joint tree algorithm, combined with the established conditional probability table, the probability of the sensor operating state being abnormal under this environmental level is calculated. This effectively assesses the probability of abnormal sensor operating conditions based on historical data and the constructed model.

[0096] Furthermore, the adjusting of the sensor detection parameters according to the working state to obtain the correction environment parameters includes the following steps:

[0097] Obtain historical data samples of the probability of abnormal operating conditions and the corresponding detection deviations: This step aims to collect historical data as a foundation for subsequent analysis. By recording past sensor operation, we can compile data on the probability of abnormal operating conditions. We also record the deviation between sensor detection results and actual environmental parameters each time an abnormal operating condition occurs. This deviation reflects the degree of sensor detection inaccuracy during abnormal conditions. Collecting a large amount of such sample data allows us to fully understand the correlation between the probability of abnormal operating conditions and detection deviations.

[0098] Establish a transformation model for the relationship between the probability of an abnormal operating state and detection deviation: Based on the data samples obtained in the previous step, use mathematical or statistical methods to construct a model that describes the inherent relationship between the probability of an abnormal operating state and detection deviation. For example, regression analysis or machine learning algorithms (such as generalized additive models, linear regression, and decision tree regression) can be used to determine this relationship. The model building process involves discovering patterns from large amounts of data and finding a function or algorithm that can predict detection deviation based on the probability of an abnormality. This model will become the key basis for adjusting sensor detection parameters.

[0099] Input the probability of the current sensor operating state being abnormal into the relational transformation model and output the detection bias: Once the probability of the current sensor operating state being abnormal is known, this probability value is substituted into the established relational transformation model. The model processes the input abnormality probability based on previously learned patterns and outputs a corresponding detection bias value. This detection bias value represents the potential detection error of the current sensor, predicted based on historical data and the model.

[0100] Based on the output detection deviation, the sensor's detection parameters are adjusted to obtain corrected environmental parameters. After obtaining the detection deviation, this deviation is used to adjust the sensor's current detection parameters based on the sensor's operating principle and structural characteristics. For example, if the detection deviation indicates that the sensor's detection value is too high, the sensor's sensitivity can be appropriately reduced or a subtraction operation can be performed on the detection result. Through such adjustments, the sensor's detection results are closer to the actual environmental parameters, resulting in corrected environmental parameters and improving the accuracy and reliability of sensor detection.

[0101] Furthermore, the relationship transformation model is configured as a generalized additive model, comprising the following steps:

[0102] Data preparation: Collect historical data samples on the probability of abnormal working status and the corresponding detection deviations, and then clean and preprocess this data. The cleaning process includes removing noise, outliers, and missing values from the data. Missing values can be filled with the mean, median, or interpolated based on the data distribution. Preprocessing also includes standardizing or normalizing the data to map the abnormality probability and detection deviation data to an appropriate scale. This ensures that the model can better process the data and improves model stability and convergence speed.

[0103] Model structure setting: The generalized additive model (GAM) assumes that the relationship between the response variable (detection bias) and the explanatory variable (the probability of the working status being abnormal) can be expressed as the sum of multiple smooth functions. In this scenario, the model structure is set to a detection bias equal to a constant term plus a smooth function of the probability of the working status being abnormal. Smooth functions are usually approximated by spline functions, which have good fitting capabilities and can flexibly capture nonlinear relationships between variables. Select an appropriate spline type, such as a natural spline or a cubic spline, and determine the number and location of the spline nodes. These parameters will affect the complexity and fitting effect of the model.

[0104] Model Fitting: Fit a generalized additive model using the prepared data. Using iterative algorithms, such as locally weighted regression scatter point smoothing (LOWESS) or penalized likelihood estimation, the parameters of the smoothing function are continuously adjusted to minimize the error between the model's predicted values and the actual detection deviations. During the fitting process, it is important to select an appropriate penalty parameter to balance the model's goodness of fit and complexity. If the penalty parameter is too large, the model will be too smooth and may not capture subtle changes in the data; if the penalty parameter is too small, the model will be too complex and prone to overfitting. The optimal penalty parameter is selected through methods such as cross-validation.

[0105] Model evaluation: Use an independent validation dataset to evaluate the fitted generalized additive model. Calculate the model evaluation metrics such as mean square error (MSE), mean absolute error (MAE), or coefficient of determination (R 2 ), etc. The mean squared error measures the average squared deviation between the model's predictions and actual values, the mean absolute error measures the average absolute deviation, and the coefficient of determination indicates how well the model explains the data. These metrics are used to evaluate the model's predictive performance and generalization ability, determining whether the model can accurately predict the detection deviation corresponding to the probability of an abnormal working state.

[0106] Prediction and Application: The probability of the current sensor operating state being abnormal is input into an evaluated and optimized generalized additive model. The model calculates and outputs the corresponding detection deviation prediction based on the fitted smooth function. Based on this predicted detection deviation, the sensor's detection parameters are adjusted to obtain the corrected environmental parameters. During actual application, the model's prediction results are continuously monitored and updated and optimized based on new data to ensure that the model can adapt to environmental changes and sensor performance drift.

[0107] Furthermore, the generalized additive model is calculated as follows:

[0108] The generalized additive model is expressed as:

[0109]

[0110] In the formula, g(u y ) is an identity connection function with additive properties; u y is the response variable, i.e., the detection deviation; s(x i , k i ) is a smooth function, the smooth function is configured as a cubic spline function, x i is the explanatory variable, i.e., the abnormal probability, k i is the number of nodes of the cubic spline function, which controls the number of intervals within the numerical range of the explanatory variable. Each interval has a curve fitted by the cubic spline function. n is the number of explanatory variables (1 in this embodiment).

[0111] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent adaptive environmental perception sensor, characterized by: It includes an environment detection module, an environment analysis module and a parameter adjustment module, which are communicatively connected to each other, wherein: The environmental detection module is used to detect and read the preliminary environmental parameters of the current sensor; The environmental analysis module is used to analyze the working status of the sensor according to preliminary environmental parameters; The parameter adjustment module is used to adjust the sensor detection parameters according to the working state to obtain the correction environment parameters.

2. The intelligent adaptive environment perception sensor according to claim 1, characterized in that: The environmental detection module includes a main detection unit and several redundant detection units, and its specific operations are as follows: When the preliminary environmental parameters detected do not exceed the set environmental tolerance threshold, the detection results of the main detection unit are adopted; When the preliminary environmental parameters detected exceed the set environmental tolerance threshold, all redundant detection units are used for measurement, and the optimal detection results are analyzed through the multi-layer perceptron model.

3. The intelligent adaptive environment perception sensor according to claim 2, characterized in that: The method of analyzing the optimal detection result by the multi-layer perceptron model includes the following steps: S1. Determine initial weights: Determine initial weights for each redundant detection unit by comprehensively considering the hardware characteristics of the redundant detection unit, the accuracy of past detection data, and stability under different environmental conditions; S2. Feature extraction and data preprocessing: Fourier transform and wavelet transform methods are used to extract key features from the multi-parameter environmental detection data obtained by redundant detection units; the data is normalized to unify the data of different parameter ranges to the same scale; S3. Build a neural network model: Build a multi-layer perceptron model so that the number of nodes in the input layer corresponds to the number of extracted features. Adjust the number of layers and nodes in the hidden layer based on data complexity and computing resources, and use nonlinear activation functions to transform the input. S4. Model training and optimization: The pre-processed data features of each redundant detection unit are used as input, and the known accurate environmental parameters are used as labels, which are input into the neural network model for training. The error between the model prediction results and the labels is calculated through the back-propagation algorithm, and the weights of each layer of the network are adjusted according to the error. The stochastic gradient descent optimization algorithm is used, and it is iterated continuously until the error converges to the preset value. S5. Determine the optimal detection result: input the detection data features of the current redundant detection unit into the trained and optimized neural network model to obtain the result output by the model.

4. The intelligent adaptive environment perception sensor according to claim 1, characterized in that: Analyzing the working status of the sensor according to the preliminary environmental parameters includes the following steps: Obtaining samples of environmental parameter detection datasets in historical periods and the working status of corresponding sensors, where the working status is divided into normal and abnormal; The k-means clustering method was used to divide the data set into several environmental levels; Taking the environmental level as the explanatory variable and the corresponding working status as the response variable, an abnormal probability analysis model is constructed; Obtain the environmental level of the preliminary environmental parameters, input the abnormal probability analysis model, and analyze the probability that the working state of the sensor is abnormal.

5. The intelligent adaptive environment perception sensor according to claim 4, characterized in that: The k-means clustering method comprises the following steps: K1. Determine the number of clusters: Based on the characteristics of the data, use empirical rules or the elbow method to assist in determining the number of clusters k; K2, randomly initialize k cluster centers: randomly select k data points from the environmental parameter detection data set as the initial cluster centers as the clustering starting points; K3, assign data points to the nearest cluster center: calculate the distance between each data point in the data set and k cluster centers, assign the data points to the corresponding cluster according to the principle of the closest distance, and form a preliminary clustering result; K4, update cluster center: calculate the mean of all data points in each cluster and use it as the new cluster center; K5. Repeat steps K3 and K4 until convergence: Repeat the data point allocation and cluster center update steps until the cluster center change is less than the threshold or the maximum number of iterations is reached, completing the clustering of the data set.

6. The intelligent adaptive environment perception sensor according to claim 4, characterized in that: The abnormal probability analysis model is configured as a Bayesian network model, comprising the following steps: Define nodes and variables: Define the nodes of the Bayesian network model, namely the environmental level and sensor working status, and assign corresponding variables to each node to determine the model factors; Determine the network structure: Based on the causal relationship between the environmental level and the sensor working status, construct the network structure, using directed edges to indicate the logical connection between nodes; Estimated conditional probability table: For each node, the probability distribution is estimated based on historical data. The environmental level node determines the prior probability, and the sensor working status node determines the conditional probability under a given environmental level. Model validation and optimization: Validate the model using some historical data, compare the prediction with the actual working status, and adjust the probability value or network structure based on the error index; Input and reasoning: The current environmental level is input into the optimized model, and the Bayesian inference algorithm is combined with the conditional probability table to calculate the probability that the sensor's working state is abnormal.

7. The intelligent adaptive environment perception sensor according to claim 6, characterized in that: The step of adjusting the sensor detection parameters according to the working state to obtain the correction environment parameters includes the following steps: Obtain the probability of abnormal working status in the historical period and the corresponding detection deviation data samples; Establish a relationship transformation model between the probability of abnormal working status and detection deviation; The probability that the current sensor working state is abnormal is input into the relational transformation model, and the detection deviation is output; According to the output detection deviation, the detection parameters of the sensor are adjusted to obtain the correction environment parameters.

8. The intelligent adaptive environment perception sensor according to claim 7, characterized in that: The relational transformation model is configured as a generalized additive model and includes the following steps: Data preparation: Collect historical data on anomaly probability and detection deviation, clean it to remove noise, anomalies, and missing values, and standardize or normalize it; Model structure setting: Use the generalized additive model to construct the relationship between abnormal probability and detection deviation, use the spline function as the smoothing function, and determine the spline type, number of nodes, and location; Model fitting: Use an iterative algorithm to fit the model, adjust the smoothing function parameters to minimize the prediction error, select appropriate penalty parameters to balance fitting and complexity, and determine the optimal value through cross-validation; Model evaluation: Use an independent validation set to evaluate the model and calculate the coefficient of determination to determine the model's prediction and generalization capabilities.

9. The intelligent adaptive environment perception sensor according to claim 8, characterized in that: The generalized additive model is calculated as follows: In the formula, g(u y ) is an identity connection function with additive properties; u y is the response variable, i.e., the detection deviation; s(x i , k i ) is a smooth function, the smooth function is configured as a cubic spline function, x i is the explanatory variable, i.e., the abnormal probability, k i is the number of nodes of the cubic spline function, which controls the number of intervals within the numerical range of the explanatory variable. Each interval has a curve fitted by the cubic spline function, and n is the number of explanatory variables.

Images