Fault prediction and monitoring system of helium mass spectrum leak detection equipment based on data analysis

Abstract

The invention belongs to the technical field of vacuum leak detection, and relates to a helium mass spectrometer leak detection equipment fault prediction and monitoring system based on data analysis, which comprises a data acquisition and preprocessing unit, a database management module, a data analysis module and a state monitoring module. The invention further discloses a using method of the system. According to the invention, the operation efficiency and reliability of the helium mass spectrum leak detection equipment can be obviously improved, the maintenance cost can be effectively reduced, and the service life of the equipment is prolonged.

Description

Technical Field

[0001] The present invention belongs to the technical field of vacuum leak detection, and relates to a fault prediction and monitoring system for helium mass spectrometry leak detection equipment based on data analysis. Background Art

[0002] With the development of industrial technology, helium mass spectrometry leak detection equipment plays a crucial role in many fields such as semiconductor manufacturing, vacuum technology, aerospace, etc. The accuracy and reliability of these devices directly affect product quality and production efficiency. However, traditional maintenance methods mainly rely on regular manual inspections and experience-based fault troubleshooting. This method is not only time-consuming and laborious, but also difficult to achieve real-time monitoring and predictive maintenance of the device health status. Summary of the Invention

[0003] The object of the present invention is to overcome the deficiencies of the prior art, and provide a fault prediction and monitoring system for helium mass spectrometry leak detection equipment based on data analysis, so as to promote the development of traditional detection equipment towards the direction of intelligence and automation, and meet the requirements of modern industry for high precision and high reliability.

[0004] To achieve the above object, the present invention adopts the following technical solutions: A fault prediction and monitoring system for helium mass spectrometry leak detection equipment based on data analysis, comprising: A data acquisition and preprocessing unit, which is used to connect to the programmable logic controller of the helium mass spectrometry leak detection equipment in real time, collect multiple-channel sensor original signals including detection chamber pressure, gas flow rate, temperature, and humidity, and perform preprocessing such as denoising, normalization, and missing value imputation on the signals; A database management module, which is used to store the preprocessed data in a relational database according to the device ID, sensor type, and time sequence, and mark and automatically back up the fault alarms and performance degradation nodes; A data analysis module, which is used to perform fault diagnosis and performance degradation trend prediction on the stored data based on a set threshold judgment and a GRU recurrent neural network model. The GRU model includes an input layer, several hidden layers, and an output layer, and Dropout is applied during training to prevent overfitting; A status monitoring module, which is used to display the original data and the diagnosis and prediction results of the data analysis module on the user interface in real time.

[0005] The data acquisition module connects to the device through TCP / IP, the data analysis module analyzes and processes the acquired data to diagnose whether a fault has occurred, the status monitoring module is responsible for displaying the original data and the diagnosis results together, and the database management module is responsible for storing all the data for convenient statistical call.

[0006] As a preferred embodiment, the data acquisition module further includes: a timestamp synchronization sub-module for adding a high-precision timestamp to the data collected each time.

[0007] As a preferred embodiment, the data analysis module further includes: threshold judgment, which sets adjustable upper and lower threshold values for each sensor parameter; when the real-time acquisition value exceeds the threshold, an alarm is automatically triggered and the abnormal parameter and the occurrence time are recorded in the database.

[0008] As a preferred embodiment, the training process of the GRU recurrent neural network model in the data analysis module includes: Normalize the preprocessed input data; Randomly initialize the weight matrices and bias vectors of the GRU layer and the fully connected layer; Input the training samples in batches, and perform forward propagation and backward propagation based on cross-entropy loss; Use the Adam optimizer to update the parameters and save the optimal model.

[0009] The present invention also discloses a method for using a fault prediction and monitoring system of a helium mass spectrometry leak detection device based on data analysis, including the following steps: S1. The data acquisition module collects multi-channel sensor data in real time and performs preprocessing; S2. Store the preprocessed data in the database management module and mark the fault alarm and performance degradation nodes; S3. The data analysis module performs threshold judgment and GRU prediction to output the fault probabilities at the current and future times; S4. The status monitoring module displays the diagnosis results in real time and issues a warning when the fault risk exceeds the threshold.

[0010] As a preferred embodiment, in step S2, the database management module also performs encrypted storage and automatic archiving of abnormal data.

[0011] As a preferred embodiment, in step S3, the data analysis module performs incremental learning in the online inference stage to dynamically add new fault samples to the fine-tuning training of the GRU model.

[0012] The present invention has the following advantages: The present invention covers the full life cycle management from data acquisition, storage, analysis to status monitoring. By introducing efficient database management and advanced data analysis technologies, the system can not only monitor the running status of the device in real time, but also predict faults in advance and provide scientific and reasonable maintenance suggestions.

[0013] This comprehensive solution can not only significantly improve the operating efficiency and reliability of helium mass spectrometry leak detection equipment, but also effectively reduce maintenance costs, extend the service life of the equipment, and provide a more scientific and reasonable equipment management strategy. Description of the Drawings

[0014] Figure 1 This is the architecture diagram of the health management system for helium leak detection equipment of the present invention; Figure 2 This is the GRU network structure diagram. Detailed Implementation Manner

[0015] The present invention will be described in detail below with reference to the accompanying drawings.

[0016] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0017] The present invention provides a fault prediction and monitoring system for helium mass spectrometry leak detection equipment based on data analysis, including: Data acquisition module: Connection method: Connected to the PLC (programmable logic controller) on the helium mass spectrometry leak detection equipment through Ethernet to ensure stable and efficient data transmission. Supports multiple communication protocols (such as TCP / IP, UDP, etc.) to meet the compatibility requirements of different models of equipment.

[0018] Data type: This module can collect a variety of operating parameters, including but not limited to key indicators such as temperature, pressure, humidity, etc., comprehensively covering the operating conditions of the equipment.

[0019] Real-time and accuracy: Adopt high-speed data acquisition technology, combined with the timestamp mechanism, to ensure the time synchronization and sampling accuracy of the data. At the same time, the system will perform preliminary cleaning and verification on the collected raw data to remove outliers or noise interference, providing a reliable basis for subsequent analysis.

[0020] Database management module: Data storage: The system is built-in with an efficient relational database (MySQL) for storing historical data and real-time data collected from the equipment. The database is organized in a time series for easy subsequent query and analysis.

[0021] Data classification and indexing: Classify and store the collected data, for example, index according to fields such as equipment ID, sensor type, timestamp, etc. to improve query efficiency. At the same time, support marking and archiving important events (such as fault alarms, performance degradation nodes).

[0022] Data Security and Backup: To prevent data loss or corruption, the system has an automatic backup function and supports encrypted storage to ensure data security and integrity. In addition, users can also export data for specific time periods as needed for external analysis or archiving.

[0023] Historical Data Mining: The historical data in the database can be used to train neural network models, optimize threshold settings, and generate device performance trend reports to help users better understand the long-term operating status of the device.

[0024] Data Analysis Module: Threshold Judgment: The collected data is monitored in real time according to the thresholds within the normal operating range of each parameter set in advance. Once data exceeding the threshold is detected, an alarm is immediately triggered and the abnormal situation is recorded.

[0025] Neural Network Model: A neural network model is constructed using deep learning technology. Through training with historical data, complex fault patterns and performance degradation trends are identified.

[0026] Multi-dimensional Analysis: It is not limited to the analysis of a single parameter, but also supports joint analysis across parameters. For example, by combining the change trends of temperature and pressure, potential leakage risks are judged.

[0027] Custom Rule Engine: Allows users to customize analysis rules according to actual needs, such as setting dynamic thresholds or combined conditions to trigger alarms, meeting application requirements in different scenarios.

[0028] Status Monitoring Module: Result Display: The results of data analysis are visually displayed in the form of charts, curves, etc. on the user interface, facilitating operators to quickly grasp the operating status of the device.

[0029] Early Warning Processing: When the system detects potential problems, it will immediately send out early warning messages, which not only include basic anomaly prompts but also give possible fault causes and recommended solutions.

[0030] User Interface Design: Provides a friendly and highly customizable user interface that supports multiple view modes (such as dashboards, trend charts, alarm lists, etc.).

[0031] Fault Cause Display: For detected anomalies, the system will list in detail the possible causes of the problem and provide specific fault location information in combination with historical data in the database and model analysis results. This helps technicians take measures quickly and reduce downtime.

[0032] The entire process starts with the opening and closing of the vacuum chamber door of the helium leak detection equipment as the trigger signal and performs the following steps: The system monitors the opening and closing status of the vacuum chamber door of the helium leak detection equipment, and uses this as the starting signal. Once the door opening signal is detected, the data acquisition process is initiated to obtain the operating parameters from the helium leak detection equipment. The collected data is preprocessed, including the identification and handling of outliers and missing values. The cleaned data will be stored in the database to ensure the quality of the data used for subsequent analysis. The database is not only responsible for storing the processed data, but also undertakes the task of feature extraction from these data. Feature extraction helps to improve the efficiency and accuracy of data analysis and provides strong support for subsequent diagnosis. A neural network model is used to analyze the data after feature extraction. The health status of the helium leak detection equipment is judged through machine learning algorithms, and potential problems are predicted. The data collected each time and the corresponding diagnostic results are displayed in real time on the user interface. The user interface provides an intuitive operation experience, allowing operators to monitor the equipment status and take corresponding measures based on the diagnostic results.

[0033] In terms of the neural network algorithm, the present invention uses GRU (Gated Recurrent Unit) for fault diagnosis. GRU is a variant of the recurrent neural network (RNN). By introducing a gating mechanism to control the flow of information, it can more effectively capture the long-term dependencies in time series data. Its structure is as Figure 2 shown.

[0034] Among them, the update gate determines how much old information should be retained in the hidden state and how much new information should be added. It is calculated through a sigmoid function, and the output value is between 0 and 1, representing the proportion of retaining the old state and the new added information. The reset gate determines how much information from the previous hidden state should be ignored. This enables GRU to understand the relationships between elements in the sequence and at the same time dynamically adjust its internal state according to the time steps of the input data, so as to better adapt to the changes in the sequence.

[0035] Reset gate: , .

[0036] Update gate: , .

[0037] Specific steps for establishing a fault diagnosis model using the GRU neural network: 1) Data preprocessing First, the collected data needs to be preprocessed. It is selected to normalize the data. The formula is as follows: Among them, represents the minimum value of represents the maximum value of 2) Construct a GRU neural network model When constructing the GRU neural network, set the number of nodes in the input layer to the embedding dimension n, and introduce a Dropout layer to avoid overfitting; the output layer uses the softmax activation function, and the number of nodes in the output layer is the number of fault classifications.

[0038] 3) Training process Regarding the training dataset, the training set is denoted as x_train and y_train, where x_train is the training data and y_train is the fault label of the training data.

[0039] Training steps of the GRU neural network: The first step: Initialize the network. Randomly initialize the weight matrices and bias vectors of the GRU layer (update gate, reset gate, candidate hidden state), the weight matrix and bias vector of the fully connected layer, and set the learning rate η.

[0040] The second step: Sample division. Divide the dataset into a training set and a test set, which are used for model training and performance evaluation respectively.

[0041] The third step: Forward propagation and error calculation. Input the training samples into the network, calculate the predicted output through forward propagation, and use an appropriate loss function, cross-entropy, to evaluate the error between the predicted value and the true value.

[0042] The fourth step: Backward propagation and parameter update. Based on the calculated error, use the backpropagation algorithm to calculate the gradients, and update the network parameters (including the weights and biases of the GRU layer and the fully connected layer) through the Adam optimizer. During this process, further improve the model generalization ability through the regularization technique Dropout.

[0043] The fifth step: Iteration and termination judgment. Check whether the termination conditions are met (such as reaching the maximum number of iterations or the validation set error no longer decreases). If not, return to the third step to continue the iteration; otherwise, save the optimal model and end the training.

[0044] 4) Verification of the GRU neural network model Deploy the trained optimal GRU model to the helium leak detection equipment health management system for actual testing to verify whether it can accurately predict faults.

[0045] The present invention is not limited to the foregoing specific embodiments. The present invention extends to any new feature disclosed in this specification or any new combination, as well as any new method or process step disclosed or any new combination.

Claims

1. A fault prediction and monitoring system for a helium mass spectrometer leak detection device based on data analysis, characterized in that, It includes: A data acquisition and preprocessing unit, which is used to connect to the programmable logic controller of the helium mass spectrometry leak detection equipment to collect in real time multiple-channel raw sensor signals including the pressure, gas flow rate, temperature, and humidity of the detection chamber, and perform preprocessing such as denoising, normalization, and missing value imputation on the signals; A database management module, which is used to store the preprocessed data in a relational database according to the device ID, sensor type, and time sequence, and mark and automatically back up the fault alarms and performance degradation nodes; A data analysis module, which is used to perform fault diagnosis and performance degradation trend prediction on the stored data based on the set threshold judgment and the GRU recurrent neural network model. The GRU model includes an input layer, several hidden layers, and an output layer, and applies Dropout during training to prevent overfitting; A status monitoring module, which is used to display the raw data and the diagnosis and prediction results of the data analysis module on the user interface in real time.

2. The fault prediction and monitoring system of the helium mass spectrometry leak detection equipment based on data analysis according to claim 1, characterized in that, The data acquisition module further includes: a timestamp synchronization sub-module, which is used to add high-precision timestamps to the data collected each time.

3. The fault prediction and monitoring system of the helium mass spectrometry leak detection device based on data analysis according to claim 1, characterized in that The data analysis module also includes: threshold judgment, which sets adjustable upper and lower limit thresholds for each sensor parameter; when the real-time collected value exceeds the threshold, it automatically triggers an alarm and records the abnormal parameter and the occurrence time in the database.

4. The fault prediction and monitoring system of the helium mass spectrometry leak detection device based on data analysis according to claim 1, characterized in that, The training process of the GRU recurrent neural network model in the data analysis module includes: Performing normalization processing on the preprocessed input data; Randomly initializing the weight matrices and bias vectors of the GRU layer and the fully connected layer; Inputting the training samples batch by batch, and performing forward propagation and backward propagation based on the cross-entropy loss; Using the Adam optimizer to update the parameters and save the optimal model.

5. The method of using the fault prediction and monitoring system for the helium mass spectrometry leak detection equipment based on data analysis according to any one of claims 1-4, characterized in that, It includes the following steps: S1. The data acquisition module collects multiple-channel sensor data in real time and performs preprocessing; S2. Store the preprocessed data in the database management module and mark the fault alarms and performance degradation nodes; S3. The data analysis module performs threshold judgment and GRU prediction, and outputs the fault probabilities at the current and future times; S4. The status monitoring module displays the diagnosis results in real time and issues a warning when the fault risk exceeds the threshold.

6. The usage method according to claim 5, characterized in that: In step S2, the database management module also performs encrypted storage and automatic archiving on the abnormal data.

7. The usage method according to claim 5, characterized in that: In step S3, the data analysis module performs incremental learning during the online inference stage to dynamically add new fault samples to the fine-tuning training of the GRU model.

Images