Equipment diagnosis method and system based on SVM-RNN model

Abstract

The invention provides an SVM-RNN (support vector machine-recurrent neural network) model-based equipment diagnosis method, which comprises the following steps of: acquiring equipment operation data, carrying out preprocessing including data VMD decomposition and normalization processing, carrying out feature selection and classification by utilizing a support vector machine (SVM) algorithm, and then introducing a recurrent neural network (RNN) model to carry out deeper diagnosis and prediction by combining the output of the SVM. The SVM-RNN model provided by the invention can accurately output the health state or fault information of the equipment, and realizes services such as data transparency, intelligent operation and maintenance, intelligent equipment scheduling and the like; a user can quickly know the state of the equipment, find potential problems in time and take corresponding measures, so that the reliability and the stability of the equipment are improved, the maintenance cost is reduced, and intelligent management, operation and maintenance are realized.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence, building system integration, data analysis and processing, and more particularly to an equipment diagnosis method and system based on a SVM-RNN model. Background Art

[0002] With the continuous development of industrial production and equipment manufacturing, real-time monitoring and fault diagnosis of equipment operation status have become increasingly urgent. Traditional experience-based equipment maintenance methods can no longer meet today's efficient and accurate production needs. Therefore, the introduction of advanced SVM-RNN model algorithms has become an urgent technological breakthrough.

[0003] First, as a supervised learning algorithm, SVM (Support Vector Machine) has a high effect in feature selection and classification. Through the SVM algorithm, key features can be effectively screened from massive amounts of equipment data, and accurate status classification can be performed, thereby achieving rapid judgment and diagnosis of equipment status. This provides important technical support for the prevention and timely handling of equipment failures.

[0004] Secondly, RNN (recurrent neural network), as a deep learning model that can process time series data, has unique advantages in equipment failure prediction and time series analysis. Through the RNN model, the equipment status can be continuously predicted in time, potential failure trends can be identified, and corresponding maintenance measures can be taken in advance, thereby minimizing the adverse effects of equipment failure on production and operations.

[0005] In addition, data preprocessing algorithms play a vital role in the entire analysis process. The collected raw data often contains noise and unnecessary information, so it needs to go through a series of preprocessing steps, such as VMD decomposition and normalization, to improve the quality and accuracy of the data and lay a solid foundation for subsequent feature extraction and model training. Summary of the invention

[0006] The purpose of the present invention is to propose an equipment diagnosis method and system based on the SVM-RNN model, which can not only realize accurate monitoring of equipment status and fault diagnosis, but also achieve a high level of real-time and accuracy, providing important support for the intelligence and automation of industrial production, and has broad application prospects and market potential.

[0007] To achieve the above object, the technical solution of the present invention is: a device diagnosis method based on SVM-RNN model, comprising the following steps:

[0008] S1. Collect equipment operation data;

[0009] S2, preprocessing the collected equipment operation data, including data VMD decomposition and normalization processing;

[0010] S3, build SVM-RNN model and complete model training; use support vector machine SVM model to select and classify features of data processed by S2 to determine key features and classify equipment status; introduce recurrent neural network RNN ​​based on the output of SVM model to diagnose and predict equipment status;

[0011] S4. Based on the diagnosis and prediction results of the SVM-RNN model, the health status or fault information of the equipment is output, and an early warning is issued in time when an abnormality is found or a possible fault is predicted.

[0012] Preferably, the device operation data includes sensor data and device status information, the sensor data is the device operation data collected in real time by sensors installed on the device, including temperature, pressure and vibration data; the device status information is the data of the device operation status obtained through the device interface.

[0013] Preferably, the preprocessing of the collected equipment operation data is specifically:

[0014] S21, setting parameters of the VMD algorithm, including the number of modes of the data, wherein the number of modes is set according to the number of operating status categories of the device; using the VMD algorithm to decompose the collected device operating data to obtain data sequences of different modes;

[0015] S22. Normalize the data of each mode to ensure that the data are within the same scale range.

[0016] Preferably, the support vector machine (SVM) model is used to perform feature selection and classification on the data processed by S2, specifically:

[0017] For linearly separable data, obtain the weight vector ω of the trained SVM model. Each component of the weight vector ω represents the coefficient of the corresponding feature in the decision function. The key feature is determined by the absolute value of the weight vector. Obtain the decision function of the trained SVM model and classify and predict the input data through the decision function:

[0018] f(x)=ω·x+b

[0019] Where ω is the weight vector, x is the input feature vector, b is the bias (or intercept), and f(x) is the predicted category of the input data;

[0020] Among them, for linearly inseparable data, a kernel function is introduced to map the originally linearly inseparable data to a high-dimensional feature space, so that the originally linearly inseparable data becomes linearly separable in the high-dimensional space; the kernel function includes a linear kernel function, a polynomial kernel function and a Gaussian kernel function.

[0021] Preferably, the introduction of a recurrent neural network (RNN) based on the output of the SVM model to diagnose and predict the device status is specifically as follows:

[0022] The output of the SVM model is used as the input of the recurrent neural network RNN, and the hidden state of the current time step is updated according to the current input and the hidden state of the previous time step. The expression of the recurrent neural network RNN ​​is as follows:

[0023] h(t)=f(W xh x(t)+W hh h(t-1)+b h )

[0024] Where x(t) is the input of RNN, W xh is the weight matrix input to the hidden state, W hh is the weight matrix from the previous time step hidden state to the current time step hidden state, b h is the bias term of the hidden state, f is the activation function, and h(t) is the hidden state at the current time step.

[0025] Preferably, the recurrent neural network RNN ​​includes a long short-term memory network LSTM or a gated recurrent unit GRU.

[0026] Preferably, the output device health status or fault information includes:

[0027] Predict equipment failure time: Predict equipment status based on the RNN model and provide the time range in which equipment failure may occur;

[0028] Provide fault cause diagnosis: Identify possible causes of equipment failure through analysis of equipment status data and provide corresponding diagnostic information;

[0029] Recommended maintenance plan: Based on the fault diagnosis results and equipment status prediction, provide targeted maintenance plans or suggestions to solve possible equipment failures.

[0030] An equipment diagnosis system based on SVM-RNN model algorithm includes the following modules:

[0031] Data acquisition module: responsible for collecting equipment operation data, including obtaining data through sensors and equipment interfaces;

[0032] Data preprocessing module: preprocess the collected equipment operation data, including data VMD decomposition and normalization processing;

[0033] SVM module: uses support vector machine algorithm to perform feature selection and classification to determine the equipment status;

[0034] RNN module: introduces a recurrent neural network model for further diagnosis and prediction;

[0035] Result output module: provides users with equipment health status or fault information through a graphical interface or network interface.

[0036] Preferably, the result output module provides the user with equipment health status or fault information through a graphical interface, and the graphical interface provides an intuitive graphical display, including:

[0037] Equipment status diagram: Displays the changing trends of key indicators of the equipment over time, including curve diagrams of temperature, pressure, and vibration parameters, so that users can intuitively understand the operating status of the equipment;

[0038] Fault prediction curve: Predicts the status of equipment based on the RNN model, displays the time range in which equipment failure may occur, and predicts the probability change curve of failure occurrence, helping users take maintenance measures in a timely manner;

[0039] Feature importance graph: After using the SVM algorithm to select features, it shows the importance of each feature in the classification of equipment status;

[0040] Maintenance plan recommendation chart: Based on the fault diagnosis results and equipment status prediction, it provides targeted maintenance plans or suggestions, including a table listing various possible fault causes and corresponding solutions for user reference.

[0041] Preferably, the result output module provides the user with the equipment health status or fault information through the network interface, and transmits the equipment health status or fault information to the user or other systems through the network service or API interface to realize remote monitoring and data sharing.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] Through the SVM-RNN model, the present invention can accurately output the health status or fault information of the equipment, and realize services such as data transparency, intelligent operation and maintenance, and intelligent equipment scheduling. Through this model, users can quickly understand the status of the equipment, discover potential problems in time, and take corresponding measures, thereby improving the reliability and stability of the equipment, reducing maintenance costs, and realizing intelligent management and operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the principle process of the present invention. DETAILED DESCRIPTION

[0045] The following is combined with Figure 1 , the technical solution of the present invention is specifically described.

[0046] The present invention proposes an equipment diagnosis method based on a SVM-RNN model, which is characterized by comprising the following steps:

[0047] S1. Collect equipment operation data; the equipment is any equipment in industrial production, including but not limited to mechanical equipment, electrical equipment, sensors, etc.;

[0048] S2, preprocessing the collected equipment operation data, including data VMD decomposition and normalization processing;

[0049] S3. Build an SVM-RNN model and complete model training. During the training process, use the labeled device status data set for supervised learning of the model. Model verification and tuning is to use the verification set to verify the trained model, evaluate its performance and accuracy, and tune the model according to the verification results, including parameter adjustment and algorithm improvement.

[0050] The support vector machine (SVM) model is used to perform feature selection and classification on the data processed by S2 to determine the key features and classify the equipment status;

[0051] Based on the output of the SVM model, a recurrent neural network (RNN) is introduced to diagnose and predict the equipment status;

[0052] S4. Based on the diagnosis and prediction results of the SVM-RNN model, the health status or fault information of the equipment is output, and an early warning is issued in time when an abnormality is found or a possible fault is predicted.

[0053] The equipment diagnosis method can be used to monitor the equipment operation status in real time and provide real-time warnings. This means that the method can continuously monitor the operation of the equipment and issue warnings in time when an abnormality is found or a possible failure is predicted, so as to reduce the impact of equipment failure on production and operation.

[0054] In this embodiment, the equipment operation data includes sensor data and equipment status information. The sensor data is the equipment operation data collected in real time by sensors installed on the equipment, including temperature, pressure and vibration data; the equipment status information is the equipment status information, control instructions and other data obtained through the equipment interface (communication interface or bus), which realizes the monitoring and control of the equipment operation status. The data collection frequency is set according to the nature of the equipment and the monitoring requirements to ensure the comprehensiveness and accuracy of the data.

[0055] In this embodiment, the preprocessing of the collected equipment operation data is specifically as follows:

[0056] S21, setting parameters of the VMD algorithm, including the number of modes of the data, wherein the number of modes is set according to the number of operating state categories of the device; using the VMD algorithm to decompose the collected original device operating data to obtain data sequences of different modes;

[0057] S22. Normalize the data of each modality to ensure that the data are within the same scale range for subsequent feature selection and classification.

[0058] In this embodiment, the support vector machine SVM model is used to perform feature selection and classification on the data processed by S2, specifically:

[0059] For linearly separable data, obtain the weight vector ω of the trained SVM model. Each component of the weight vector ω represents the coefficient of the corresponding feature in the decision function. The key feature is determined by the absolute value of the weight vector. The key feature refers to the most critical feature for equipment status diagnosis. Obtain the decision function of the trained SVM model and classify and predict the input data through the decision function:

[0060] f(x)=ω·x+b

[0061] Where ω is the weight vector, x is the input feature vector, b is the bias (or intercept), and f(x) is the predicted category for the input data. The goal of the decision function is to maximize the distance from the support vector to the decision boundary, which is called the margin. By adjusting the weight vector ω and the bias b, the SVM model can learn and classify new data points.

[0062] Among them, for linearly inseparable data, a kernel function is introduced to map the originally linearly inseparable data to a high-dimensional feature space, so that the originally linearly inseparable data becomes linearly separable in the high-dimensional space; the kernel function includes a linear kernel function, a polynomial kernel function and a Gaussian kernel function.

[0063] In this embodiment, the recurrent neural network RNN ​​is introduced based on the output of the SVM model to diagnose and predict the device status as follows:

[0064] The output of the SVM model is used as the input of the recurrent neural network RNN, and the hidden state of the current time step is updated according to the current input and the hidden state of the previous time step. The expression of the recurrent neural network RNN ​​is as follows:

[0065] h(t)=f(W xh x(t)+W hh h(t-1)+b h )

[0066] Where x(t) is the input of RNN, W xh is the weight matrix input to the hidden state, W hh is the weight matrix from the previous time step hidden state to the current time step hidden state, b h is the bias term of the hidden state, f is the activation function, and h(t) is the hidden state at the current time step.

[0067] In this embodiment, the recurrent neural network RNN ​​includes a long short-term memory network LSTM or a gated recurrent unit GRU.

[0068] In this embodiment, the output device health status or fault information includes:

[0069] Predict equipment failure time: Predict equipment status based on the RNN model and provide the time range in which equipment failure may occur;

[0070] Provide fault cause diagnosis: Identify possible causes of equipment failure through analysis of equipment status data and provide corresponding diagnostic information;

[0071] Recommended maintenance plan: Based on the fault diagnosis results and equipment status prediction, provide targeted maintenance plans or suggestions to solve possible equipment failures.

[0072] An equipment diagnosis system based on SVM-RNN model algorithm includes the following modules:

[0073] Data acquisition module: responsible for collecting the operating data of the equipment, including obtaining data through sensors and device interfaces; this ensures that the system can flexibly adapt to the data acquisition needs of different types of equipment.

[0074] Data preprocessing module: preprocess the collected equipment operation data, including data VMD decomposition and normalization processing;

[0075] SVM module: uses support vector machine algorithm to perform feature selection and classification to determine the equipment status;

[0076] RNN module: Introduces the recurrent neural network model for further diagnosis and prediction to improve the accuracy of fault diagnosis;

[0077] Result output module: provides users with equipment health status or fault information through a graphical interface or network interface.

[0078] The data acquisition module can obtain equipment operation data through sensors, equipment interfaces, etc., which ensures that the system can flexibly adapt to the data acquisition needs of different types of equipment. The data preprocessing module can include a VMD decomposition algorithm and a normalization processing algorithm; VMD decomposition algorithm: using adaptive, completely non-recursive modal variation and signal processing methods to decompose the data to obtain data sequences of different modes; normalization processing algorithm: normalize the data of each mode to ensure that the data is within the same scale range, so as to facilitate subsequent feature selection and classification. The SVM module and RNN module are implemented through computer programs and can be run on platforms such as computer servers; this means that the system has strong versatility and portability and can be deployed and run on different hardware platforms. The result output module provides users with equipment health status or fault information through a graphical interface or a network interface, which allows users to intuitively understand the operating status and possible fault conditions of the equipment so as to take corresponding measures in time.

[0079] In this embodiment, the result output module provides the user with the health status or fault information of the device through a graphical interface, and enables the user to quickly understand the operation status of the device through a visual interface by providing an intuitive graphical display. The graphical interface provides an intuitive graphical display, including:

[0080] Equipment status diagram: Displays the changing trends of key indicators of the equipment over time, including curve diagrams of temperature, pressure, and vibration parameters, so that users can intuitively understand the operating status of the equipment;

[0081] Fault prediction curve: Predicts the status of equipment based on the RNN model, displays the time range in which equipment failure may occur, and predicts the probability change curve of failure occurrence, helping users take maintenance measures in a timely manner;

[0082] Feature importance graph: After using the SVM algorithm to select features, it shows the importance of each feature in the classification of equipment status;

[0083] Maintenance plan recommendation chart: Based on the fault diagnosis results and equipment status prediction, it provides targeted maintenance plans or suggestions, including a table listing various possible fault causes and corresponding solutions for user reference.

[0084] In this embodiment, the result output module provides the user with equipment health status or fault information through a network interface, and transmits the equipment health status or fault information to the user or other systems through a network service or API interface to achieve remote monitoring and data sharing.

[0085] The present invention can be applied to large commercial building systems such as high-speed railway stations, government office buildings, hotels, hospitals, parks, and office buildings. By deploying the trained model to a computer server or other platform, real-time monitoring and diagnosis of the equipment status can be achieved. The deployed system is continuously monitored and maintained to ensure its stable operation and accurate diagnosis of the equipment status.

[0086] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A device diagnosis method based on SVM-RNN model, characterized in that: The following steps are involved: S1. Collect equipment operation data; S2, preprocessing the collected equipment operation data, including data VMD decomposition and normalization processing; S3, build SVM-RNN model and complete model training; use support vector machine SVM model to select and classify features of data processed by S2 to determine key features and classify equipment status; introduce recurrent neural network RNN ​​based on the output of SVM model to diagnose and predict equipment status; S4. Based on the diagnosis and prediction results of the SVM-RNN model, the health status or fault information of the equipment is output, and an early warning is issued in time when an abnormality is found or a possible fault is predicted.

2. According to the device diagnosis method based on the SVM-RNN model of claim 1, it is characterized in that: The equipment operation data includes sensor data and equipment status information. The sensor data is the equipment operation data collected in real time by sensors installed on the equipment, including temperature, pressure and vibration data; the equipment status information is the data of the equipment operation status obtained through the equipment interface.

3. The device diagnosis method based on the SVM-RNN model according to claim 1 is characterized in that: The preprocessing of the collected equipment operation data is specifically as follows: S21, setting parameters of the VMD algorithm, including the number of modes of the data, wherein the number of modes is set according to the number of operating status categories of the device; using the VMD algorithm to decompose the collected device operating data to obtain data sequences of different modes; S22. Normalize the data of each mode to ensure that the data are within the same scale range.

4. The device diagnosis method based on the SVM-RNN model according to claim 1 is characterized in that: The support vector machine (SVM) model is used to perform feature selection and classification on the data processed by S2, specifically: For linearly separable data, obtain the weight vector ω of the trained SVM model. Each component of the weight vector ω represents the coefficient of the corresponding feature in the decision function. The key feature is determined by the absolute value of the weight vector. Obtain the decision function of the trained SVM model and classify and predict the input data through the decision function: f(x)=ω·x+b Where ω is the weight vector, x is the input feature vector, b is the bias, and f(x) is the predicted category of the input data; Among them, for linearly inseparable data, a kernel function is introduced to map the originally linearly inseparable data to a high-dimensional feature space, so that the originally linearly inseparable data becomes linearly separable in the high-dimensional space; the kernel function includes a linear kernel function, a polynomial kernel function and a Gaussian kernel function.

5. The device diagnosis method based on the SVM-RNN model according to claim 1 is characterized in that: The introduction of the recurrent neural network (RNN) based on the output of the SVM model to diagnose and predict the device status is as follows: The output of the SVM model is used as the input of the recurrent neural network RNN, and the hidden state of the current time step is updated according to the current input and the hidden state of the previous time step. The expression of the recurrent neural network RNN ​​is as follows: h(t)=f(W xh x(t)+W hh h(t-1)+b h ) Where x(t) is the input of RNN, W xh is the weight matrix input to the hidden state, W hh is the weight matrix from the hidden state of the previous time step to the hidden state of the current time step, b h is the bias term of the hidden state, f is the activation function, and h(t) is the hidden state at the current time step.

6. The device diagnosis method based on the SVM-RNN model according to claim 5 is characterized in that: The recurrent neural network RNN ​​includes a long short-term memory network LSTM or a gated recurrent unit GRU.

7. The device diagnosis method based on the SVM-RNN model according to claim 1 is characterized in that: The output device health status or fault information includes: Predict equipment failure time: Predict equipment status based on the RNN model and provide the time range in which equipment failure may occur; Provide fault cause diagnosis: Identify possible causes of equipment failure through analysis of equipment status data and provide corresponding diagnostic information; Recommended maintenance plan: Based on the fault diagnosis results and equipment status prediction, provide targeted maintenance plans or suggestions to solve possible equipment failures.

8. An equipment diagnosis system based on SVM-RNN model algorithm, characterized in that: Includes the following modules: Data acquisition module: responsible for collecting equipment operation data, including obtaining data through sensors and equipment interfaces; Data preprocessing module: preprocess the collected equipment operation data; Including data VMD decomposition and normalization processing; SVM module: uses support vector machine algorithm to perform feature selection and classification to determine the equipment status; RNN module: introduces a recurrent neural network model for further diagnosis and prediction; Result output module: provides users with equipment health status or fault information through a graphical interface or network interface.

9. The equipment diagnosis system based on the SVM-RNN model algorithm according to claim 8, characterized in that: The result output module provides the user with equipment health status or fault information through a graphical interface, which provides an intuitive graphical display, including: Equipment status diagram: Displays the changing trends of key indicators of the equipment over time, including curve diagrams of temperature, pressure, and vibration parameters, so that users can intuitively understand the operating status of the equipment; Fault prediction curve: Predicts the status of equipment based on the RNN model, displays the time range in which equipment failure may occur, and predicts the probability change curve of failure occurrence, helping users take maintenance measures in a timely manner; Feature importance graph: After using the SVM algorithm to select features, it shows the importance of each feature in the classification of equipment status; Maintenance plan recommendation chart: Based on the fault diagnosis results and equipment status prediction, it provides targeted maintenance plans or suggestions, including a table listing various possible fault causes and corresponding solutions for user reference.

10. The equipment diagnosis system based on the SVM-RNN model algorithm according to claim 8, characterized in that: The result output module provides the user with the equipment health status or fault information through the network interface, and transmits the equipment health status or fault information to the user or other systems through the network service or API interface to realize remote monitoring and data sharing.

Images