Method for detecting service life of trigger board
Through the combination of random forest regression model and data acquisition module, the problem of inability to predict trigger plate life and failure in the existing technology is solved, and high-accurate life estimation and fault warning are achieved, which improves the operating efficiency and safety of the equipment .
Patent Information
- Application Number
- CN202410979226.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-12
AI Technical Summary
The existing trigger board detection systems lack the integrated advanced data processing and analysis capabilities, and cannot achieve real-time fault prediction or residual life estimation, resulting in reduced system reliability.
The random forest regression model is used to combine data acquisition, preprocessing, feature extraction, model construction, real-time monitoring and early warning modules. The trigger board status data is collected through the data acquisition module, preprocessing and feature extraction, build feature matrix and target vector, and train random Forest regression model predicts the remaining life, and optimizes model parameters in the event of a failure to trigger early warning.
It improves the accuracy of estimating the remaining life of the trigger board, can promptly detect equipment status abnormalities and make early warning responses, and improves the equipment's operating efficiency, safety and life cycle management capabilities.
Smart Images

Figure CN118940170A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of electronic technology, and in particular relates to a trigger plate life detection method. Background Art
[0002] Trigger boards are usually used in various automated control systems, especially in situations where electrical signals are needed to trigger specific operations or events. After long-term use, the trigger board may have circuit aging, unstable connections, or decreased sensitivity, resulting in reduced system reliability. Therefore, detecting and monitoring the life of the trigger board is of great significance to the stability of the automated control system.
[0003] Current trigger plate inspection systems typically lack integrated advanced data processing and analysis capabilities to achieve real-time failure prediction or remaining life estimation. Summary of the invention
[0004] The purpose of the present invention is to provide a trigger plate life detection method, which solves the technical problem of predicting the life status of the trigger plate and giving early warning of the occurrence of failures.
[0005] To achieve the above object, the present invention adopts the following technical solution:
[0006] A method for detecting the life of a trigger plate comprises the following steps:
[0007] Step 1: Establish a data acquisition module and a prediction server. Establish a data preprocessing module, a feature engineering module, a model building module, a real-time monitoring module, an early warning module, an optimization module, and a remote monitoring module in the prediction server. The data acquisition module collects various status data of the trigger board, generates a status data packet, and uploads the status data packet to the prediction server via the Internet.
[0008] Step 2: After the prediction server receives the status data packet, the data preprocessing module preprocesses the status data packet to obtain preprocessed data;
[0009] Step 3: The feature engineering module extracts features from the preprocessed data and constructs a feature set;
[0010] Step 4: The model building module constructs a feature matrix and a target vector based on the feature set; builds and trains a random forest regression model and adjusts its hyperparameters; the target vector is used to represent the remaining useful life of each sample in the feature matrix;
[0011] Step 5: The real-time monitoring module obtains the real-time status data packet collected by the data acquisition module and inputs it into the trained random forest regression model to predict the remaining life of the current trigger board;
[0012] Step 6: The early warning module compares the predicted remaining life with the preset early warning threshold. If it is lower than the threshold, an early warning is triggered and early warning information is generated;
[0013] Step 7: The optimization module records the event information of each time a fault occurs on the trigger board, generates a historical fault data set, uses the fault type and duration as retrieval tags, extracts fault-related status data from the historical data set, and constructs a fault training set; the fault training set is fed back to the model construction module, the random forest regression model is optimized and trained, and the model parameters are adjusted;
[0014] Step 8: The remote monitoring module obtains the data in the remaining life and real-time status data packet, generates an operation interface, and displays it to the client user.
[0015] Preferably, the data in the status data packet includes the input voltage V in 、Input current I in , output voltage V out , output current I out , trigger pulse frequency f, trigger duration t triggert , cumulative running time t run 、N number of triggers trigger , trigger board temperature T board , working environment temperature T env , Working environment humidity H env .
[0016] Preferably, when executing step 2, preprocessing the status data packet includes filling missing values using an interpolation method and filtering the status data packet using a Kalman filter.
[0017] Preferably, when executing step 4, a feature matrix for the state data packet is constructed, and then a target vector is constructed according to the feature matrix to represent the remaining service life corresponding to each sample in the feature matrix; one sample contains a state data packet, and the specific mathematical model of the feature matrix is as follows:
[0018]
[0019] Where X is the feature matrix, each row in the feature matrix represents a sample, that is, the data in a state data packet, and each column represents a different feature;
[0020] The specific mathematical model of the target vector is as follows:
[0021]
[0022] Where y is the target vector, which contains the remaining useful life corresponding to each sample, RUL represents the remaining useful life, and the subscript n represents the sample number;
[0023] The random forest regression model is as follows:
[0024]
[0025] Where M is the number of trees, h m is the mth regression tree;
[0026] The hyperparameters are M, the maximum depth of the tree, and the minimum number of sample splits.
[0027] Preferably, the data acquisition module includes a sensor group, a controller and a communication module, the sensor group includes multiple voltage sensors, multiple current sensors, a temperature sensor, a humidity sensor and a pulse signal detection interface, and the multiple voltage sensors, multiple current sensors, temperature sensors, humidity sensors, pulse signal detection interface and communication module are all connected to the controller.
[0028] Preferably, the controller is an ARM development board, the specific model is
[0029] STM32F4DISCOVERY, the voltage sensor model is TI INA219, the current sensor model is ACS712, the temperature sensor model is DS18B20, the humidity sensor model is DHT22, the pulse signal detection interface is an open input signal interface, and the communication module is an Ethernet module, model WIZnet W5500.
[0030] A trigger plate life detection method described in the present invention solves the technical problem of predicting the life status of a trigger plate and giving early warning of a fault. The present invention is based on a random forest regression model and continuously improves and optimizes the random forest model, thereby improving the accuracy of estimating the remaining life of the trigger plate, being able to promptly detect abnormal equipment status and make early warning responses, thereby overall improving the operating efficiency, safety and life cycle management capabilities of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 is a system architecture diagram of the present invention;
[0032] Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0033] Depend on Figure 1-Figure 2 A trigger plate life detection method shown includes the following steps:
[0034] Step 1: Establish a data acquisition module and a prediction server. Establish a data preprocessing module, a feature engineering module, a model building module, a real-time monitoring module, an early warning module, an optimization module and a remote monitoring module in the prediction server. The data acquisition module collects various status data of the trigger board, generates a status data packet, and uploads the status data packet to the prediction server via the Internet; the data in the status data packet includes the input voltage V in 、Input current I in , output voltage V out , output current I out , trigger pulse frequency f, trigger duration t triggert , cumulative running time t run 、N number of triggers trigger , trigger board temperature T board , working environment temperature T env , Working environment humidity H env .
[0035] Step 2: After the prediction server receives the status data packet, the data preprocessing module preprocesses the status data packet to obtain preprocessed data; the preprocessing of the status data packet includes using interpolation to fill in missing values and using a Kalman filter to filter the status data packet.
[0036] Step 3: The feature engineering module extracts features from the preprocessed data and constructs a feature set;
[0037] In this embodiment, the preprocessed data can be used for feature extraction by selecting a suitable feature extraction method in the prior art, such as a statistical feature extraction method, based on various data collected by the trigger board (such as voltage, current, frequency, duration, temperature, etc.).
[0038] Step 4: The model building module constructs a feature matrix and a target vector based on the feature set; builds and trains a random forest regression model and adjusts its hyperparameters; the target vector is used to represent the remaining useful life of each sample in the feature matrix;
[0039] Specifically, a feature matrix for the state data packet is constructed, and then a target vector is constructed based on the feature matrix to represent the remaining service life corresponding to each sample in the feature matrix; one sample contains a state data packet, and the specific mathematical model of the feature matrix is as follows:
[0040]
[0041] Where X is the feature matrix, each row in the feature matrix represents a sample, that is, the data in a state data packet, and each column represents a different feature;
[0042] The specific mathematical model of the target vector is as follows:
[0043]
[0044] Where y is the target vector, which contains the remaining useful life corresponding to each sample, RUL represents the remaining useful life, and the subscript n represents the sample number;
[0045] The random forest regression model is as follows:
[0046]
[0047] Where M is the number of trees, h m is the mth regression tree;
[0048] The hyperparameters are M, the maximum depth of the tree, and the minimum number of sample splits.
[0049] In this embodiment, the decision tree in the random forest regression model is constructed by recursively splitting the data set. Specifically, each time the data is split, a feature X is selected from the feature matrix X. j , and choose a threshold θ to maximize the purity of the data:
[0050] For feature X j And threshold θ, select the combination that minimizes the classification error:
[0051]
[0052] Among them, I is an indicator function, indicating the truth or falsity of the condition, x i represents the feature vector of the i-th sample (or data point) in the data set, x i,j represents the jth eigenvalue of the sample in the feature space, and N represents the total number of samples in the data set, that is, the number of data points.
[0053] I(x i,j ≤θ) and I(x i,j >θ) represents the indicator function of the data points in the left and right child nodes after splitting feature j according to a specific threshold θ.
[0054] Gini(L) and Gini(R) are the Gini impurities of the left and right child nodes, defined as:
[0055]
[0056] Among them, p k is the proportion of samples belonging to class k in node t, and K is the total number of classes.
[0057] In this embodiment, the maximum depth of the tree is set to limit the complexity of the tree and prevent overfitting.
[0058] Step 5: The real-time monitoring module obtains the real-time status data packet collected by the data acquisition module and inputs it into the trained random forest regression model to predict the remaining life of the current trigger board;
[0059] Step 6: The early warning module compares the predicted remaining life with the preset early warning threshold. If it is lower than the threshold, an early warning is triggered and an early warning message is generated. Specifically, the early warning module presets a remaining life early warning threshold, obtains the remaining life obtained in step 5, and compares it with the remaining life early warning threshold: if the predicted remaining life is lower than the early warning threshold, an early warning is triggered and an early warning message is generated; otherwise, no action is taken.
[0060] Step 7: The optimization module records the event information of each time a fault occurs on the trigger board, generates a historical fault data set, uses the fault type and duration as retrieval tags, extracts fault-related status data from the historical data set, and constructs a fault training set; the fault training set is fed back to the model construction module, the random forest regression model is optimized and trained, and the model parameters are adjusted;
[0061] In this embodiment, the optimization module uses the fault type and duration as retrieval tags, extracts the voltage, current, frequency, duration and temperature data related to the fault from the historical data set, and constructs a fault training set.
[0062] Step 8: The remote monitoring module obtains the data in the remaining life and real-time status data packet, generates an operation interface, and displays it to the client user.
[0063] In this embodiment, the data acquisition module includes a sensor group, a controller and a communication module. The sensor group includes multiple voltage sensors, multiple current sensors, a temperature sensor, a humidity sensor and a pulse signal detection interface. The multiple voltage sensors, multiple current sensors, a temperature sensor, a humidity sensor, a pulse signal detection interface and the communication module are all connected to the controller.
[0064] The controller is an ARM development board, specifically model STM32F4DISCOVERY, the voltage sensor model is TIINA219, the current sensor model is ACS712, the temperature sensor model is DS18B20, the humidity sensor model is DHT22, the pulse signal detection interface is an open input signal interface, and the communication module is an Ethernet module, model WIZnet W5500.
[0065] The voltage sensor is connected to the controller via the I2C interface, the current sensor is connected to an AD interface of the controller, and the humidity sensor is connected to the controller via the 1-Wire bus; there are at least 2 voltage sensors, which measure the input voltage V in and output voltage V outThe trigger pulse frequency f is obtained by counting the pulse signal detected by the pulse signal detection interface, and there are at least two current sensors, which measure the input current I in and output current I out There are at least two temperature sensors, which measure the trigger board temperature T board and operating environment temperature T env , humidity sensor is used to measure the working environment humidity H env , trigger duration t triggert , cumulative running time t run and trigger times N trigger All are obtained by controller counting.
[0066] A trigger plate life detection method described in the present invention solves the technical problem of predicting the life status of a trigger plate and giving early warning of a fault. The present invention is based on a random forest regression model and continuously improves and optimizes the random forest model, thereby improving the accuracy of estimating the remaining life of the trigger plate, being able to promptly detect abnormal equipment status and make early warning responses, thereby overall improving the operating efficiency, safety and life cycle management capabilities of the equipment.
Claims
1. A trigger plate life detection method, characterized in that: The steps include: Step 1: Establish a data acquisition module and a prediction server. Establish a data preprocessing module, a feature engineering module, a model building module, a real-time monitoring module, an early warning module, an optimization module, and a remote monitoring module in the prediction server. The data acquisition module collects various status data of the trigger board, generates a status data packet, and uploads the status data packet to the prediction server via the Internet. Step 2: After the prediction server receives the status data packet, the data preprocessing module preprocesses the status data packet to obtain preprocessed data; Step 3: The feature engineering module extracts features from the preprocessed data and constructs a feature set; Step 4: The model building module constructs a feature matrix and a target vector based on the feature set; builds and trains a random forest regression model and adjusts its hyperparameters; the target vector is used to represent the remaining useful life of each sample in the feature matrix; Step 5: The real-time monitoring module obtains the real-time status data packet collected by the data acquisition module and inputs it into the trained random forest regression model to predict the remaining life of the current trigger board; Step 6: The early warning module compares the predicted remaining life with the preset early warning threshold. If it is lower than the threshold, an early warning is triggered and early warning information is generated; Step 7: The optimization module records the event information of each time a fault occurs on the trigger board, generates a historical fault data set, uses the fault type and duration as retrieval tags, extracts fault-related status data from the historical data set, and constructs a fault training set; the fault training set is fed back to the model construction module, the random forest regression model is optimized and trained, and the model parameters are adjusted; Step 8: The remote monitoring module obtains the data in the remaining life and real-time status data packet, generates an operation interface, and displays it to the client user.
2. A trigger plate life detection method as claimed in claim 1, characterized in that: The data in the status data packet includes the input voltage V in 、Input current I in , output voltage V out , output current I out , trigger pulse frequency f, trigger duration t triggert , cumulative running time t run 、N number of triggers trigger , trigger board temperature T board , working environment temperature T env , Working environment humidity H env .
3. A trigger plate life detection method as claimed in claim 2, characterized in that: When executing step 2, the preprocessing of the state data packet includes filling the missing values using the interpolation method and filtering the state data packet using the Kalman filter.
4. A trigger plate life detection method as claimed in claim 2, characterized in that: When executing step 4, a feature matrix for the state data packet is constructed, and then a target vector is constructed based on the feature matrix to represent the remaining service life corresponding to each sample in the feature matrix; one sample contains a state data packet, and the specific mathematical model of the feature matrix is as follows: Where X is the feature matrix, each row in the feature matrix represents a sample, that is, the data in a state data packet, and each column represents a different feature; The specific mathematical model of the target vector is as follows: Where y is the target vector, which contains the remaining useful life corresponding to each sample, RUL represents the remaining useful life, and the subscript n represents the sample number; The random forest regression model is as follows: Where M is the number of trees, h m is the mth regression tree; The hyperparameters are M, the maximum depth of the tree, and the minimum number of sample splits.
5. A trigger plate life detection method as claimed in claim 2, characterized in that: The data acquisition module includes a sensor group, a controller and a communication module. The sensor group includes multiple voltage sensors, multiple current sensors, a temperature sensor, a humidity sensor and a pulse signal detection interface. The multiple voltage sensors, multiple current sensors, a temperature sensor, a humidity sensor, a pulse signal detection interface and the communication module are all connected to the controller.
6. A trigger plate life detection method as claimed in claim 5, characterized in that: The controller is an ARM development board, specifically model STM32F4DISCOVERY, the voltage sensor model is TI INA219, the current sensor model is ACS712, the temperature sensor model is DS18B20, the humidity sensor model is DHT22, the pulse signal detection interface is an open input signal interface, and the communication module is an Ethernet module, model WIZnet W5500.
