Forging and pressing monitoring method and system

Abstract

The invention relates to the technical field of forging and pressing monitoring, and discloses a forging and pressing monitoring method and system. According to the method, an Internet of Things sensor is arranged on forging and pressing equipment, various data in the forging and pressing process are collected in real time, and the data are transmitted to a data center through an Internet of Things gateway. In the data center, after data is preprocessed, a deep learning model based on a CTA-GNN algorithm is used for analysis, and whether equipment faults or process risks exist or not is judged. And when the risk probability exceeds a preset threshold value, the system automatically generates early warning information, sends the early warning information to related personnel through a short message, a mail or a system notification and the like, and records a sending state and a receiving confirmation condition at the same time. In addition, the invention further provides a corresponding intelligent monitoring system which comprises a data acquisition module, a data transmission module, a data processing module, an intelligent analysis module and an early warning management module. The system can accurately judge faults, improve monitoring efficiency and accuracy, timely discover and early warn potential risks, and effectively guarantee safety and stability of the forging process.

Description

Technical Field

[0001] The present invention relates to the technical field of forging monitoring, and in particular to a forging monitoring method and system. Background Art

[0002] Forging process is an important link in industrial production, and its stability and safety are directly related to product quality and production efficiency. However, traditional forging process monitoring methods often rely on manual inspections and empirical judgments, and there are problems such as low monitoring efficiency, untimely fault detection, and low early warning accuracy. With the rapid development of the Internet of Things and deep learning technology, new solutions have been provided for forging process monitoring. By collecting various data in the forging process in real time through Internet of Things sensors, and using deep learning algorithms to analyze and predict the data, intelligent monitoring of the forging process can be achieved. This method can not only improve monitoring efficiency, but also detect equipment failures or process risks in a timely manner, providing strong protection for production safety. Therefore, the development of a forging monitoring method and system based on the Internet of Things and deep learning is of great significance to improving the stability and safety of the forging process. Summary of the invention

[0003] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a forging monitoring method and system to solve the problems of low monitoring efficiency, untimely fault detection and low early warning accuracy in the prior art.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a forging monitoring method, the method comprising: Step S100: arranging IoT sensors on the forging equipment to collect forging data in real time during the forging process; Step S200: The collected forging data is transmitted to the data center via the Internet of Things gateway; Step S300: performing data preprocessing on the received forging data in the data center; Step S400: Analyze the preprocessed forging data using a deep learning-based algorithm to determine whether there is equipment failure or process risk; Step S500: When it is determined that there is equipment failure or process risk in forging, an early warning message is sent through the data center.

[0005] Preferably, in a possible implementation manner of the first aspect, the forging data includes forging force data, forging times data, forging current data, forging temperature data and equipment vibration data.

[0006] Preferably, in a possible implementation manner of the first aspect, step S200 specifically includes: The IoT sensor sends the collected forging data to the IoT gateway wirelessly; The IoT gateway encapsulates and encrypts the received forging data; The IoT gateway transmits the encapsulated and encrypted forging data to the data center.

[0007] Preferably, in a possible implementation manner of the first aspect, the data preprocessing includes data cleaning, outlier detection and data normalization.

[0008] Preferably, in a possible implementation of the first aspect, the deep learning-based algorithm adopts a CTA-GNN algorithm, and uses historical forging data, equipment failure data and forging result data to train the CTA-GNN model.

[0009] Preferably, in a possible implementation of the first aspect, the CTA-GNN model structure includes: The input layer is used to receive forging data and construct the corresponding graph structure according to the physical structure of the forging equipment. The nodes represent the components of the forging equipment, and the edges represent the interactions between the components. The feature extraction layer uses graph neural network to extract the data features of the forging process from the input forging data and graph structure; The causal-shortcut feature separation layer splits the data features obtained by the feature extraction layer into causal features and shortcut features through a soft mask mechanism. The causal features are directly related to the actual physical process in the forging process, while the shortcut features contain noise and misleading information. The parameterized adjustment layer performs parameterized backdoor adjustments on causal features and shortcut features to reduce the negative impact of shortcut features on prediction results; The output layer uses the classifier to predict the probability of each equipment failure or process risk.

[0010] Preferably, in a possible implementation manner of the first aspect, step S500 specifically includes: When the probability of equipment failure or process risk output by the deep learning-based algorithm is greater than the preset threshold, the data center generates corresponding warning information; Early warning information is sent to relevant personnel via SMS, email or system notification; The data center records the sending status and receipt confirmation of the warning information.

[0011] In a second aspect, the present invention provides a forging monitoring system, the system comprising: The data acquisition module is used to arrange IoT sensors on the forging equipment to collect forging data in real time during the forging process, including forging pressure data, forging times data, forging current data, forging temperature data, and equipment vibration data information; The data transmission module is used to send the forging data collected by the IoT sensor to the IoT gateway wirelessly. The IoT gateway encapsulates and encrypts the received data and transmits it to the data center. A data processing module is used to perform data preprocessing on the received forging data in the data center, including data cleaning, outlier detection and data normalization processing; The intelligent analysis module uses a deep learning model based on the CTA-GNN algorithm to analyze the preprocessed forging data to determine whether there is equipment failure or process risk. The CTA-GNN model structure includes an input layer, a feature extraction layer, a causal-shortcut feature separation layer, a parameterized adjustment layer, and an output layer. Early warning management module: When the probability of equipment failure or process risk output by the intelligent analysis module is greater than the preset threshold, the corresponding early warning information is generated according to the early warning rules and sent to relevant personnel via SMS, email or system notification, and the sending status and receipt confirmation of the early warning information are recorded at the same time.

[0012] The beneficial effects of the present invention are: by collecting a variety of data in the forging process in real time through the Internet of Things sensor, and through data preprocessing and deep learning algorithm analysis, it is possible to accurately judge equipment failures or process risks. This method not only improves the accuracy and efficiency of monitoring, but also realizes the timely discovery and early warning of faults, effectively avoiding potential safety hazards. In addition, by adopting the CTA-GNN algorithm, the present invention can more deeply mine the characteristic information in the forging data and improve the accuracy of fault prediction. At the same time, the early warning management module can automatically send early warning information, and record the sending status and receiving confirmation, providing timely and reliable early warning information for relevant personnel, which helps to ensure the safety and stability of the forging process. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 A flow chart of a forging monitoring method is provided for this application.

[0015] Figure 2 A structural diagram of a forging monitoring system is provided for this application.

[0016] Explanation of the accompanying drawings: 1-data acquisition module, 2-data transmission module, 3-data processing module, 4-intelligent analysis module, 5-early warning management module. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1: Figure 1 As shown, the present invention provides a forging monitoring method, the method comprising: Step S100: IoT sensors are arranged on the forging equipment to collect forging data during the forging process in real time.

[0019] Specifically, the IoT sensors include forging pressure sensors, current sensors, thermocouple temperature sensors and acceleration sensors, and the forging data include forging pressure data, forging times data, forging current data, forging temperature data and equipment vibration data.

[0020] In this embodiment, the forging pressure sensor is model XYZ-1000, which has high precision, high stability and good impact resistance. It is deployed at the position where the pressure head of the forging press contacts the workpiece, and is used to monitor the pressure changes during the forging process in real time. A current sensor model ABC-500 is installed in the control system of the forging press to monitor the current consumption of the motor during the forging process, thereby indirectly reflecting the load condition of the forging press. A thermocouple temperature sensor model DEF-200 is used to obtain the forging temperature, which fits tightly on the forging die to ensure that the temperature changes during the forging process can be accurately measured. The equipment vibration data is provided by an acceleration sensor model GHI-300, which is installed on the forging platform of the forging press to capture and analyze the vibration characteristics of the equipment during the forging process.

[0021] Forging data, including forging pressure data, forging times data, forging current data, forging temperature data and equipment vibration data, are important bases for evaluating the stability of the forging process and the health status of the equipment. Forging pressure data can directly reflect the pressure distribution during the forging process, which helps to determine whether the workpiece has obtained uniform plastic deformation; the number of forgings records the frequency of forging operations, which is of great significance for evaluating production efficiency and equipment wear, and is obtained through historical records in the data center; changes in forging current can indirectly reflect the energy consumption and load changes of the forging machine; forging temperature is a key factor affecting the plasticity and deformation resistance of the material, which is crucial for controlling the forging quality and avoiding defects caused by overheating or overcooling; and equipment vibration data can reveal the operating status of the equipment and promptly warn of potential mechanical failures.

[0022] Step S200: The collected forging data is transmitted to the data center via the Internet of Things gateway.

[0023] Specifically, the IoT sensor sends the collected forging data to the IoT gateway wirelessly; the IoT gateway encapsulates and encrypts the received forging data; and the IoT gateway transmits the encapsulated and encrypted forging data to the data center.

[0024] In this embodiment, after completing the collection of forging data, the IoT sensor transmits the data to the IoT gateway through the built-in wireless communication module.

[0025] The IoT gateway encapsulates and encrypts the received forging data. Encapsulation refers to packaging the original data sent by the sensor in a specific format to facilitate subsequent data processing and analysis; encryption is to ensure that the data is not illegally intercepted or tampered with during transmission, thereby ensuring the security of the data. In this embodiment, the IoT gateway uses the AES encryption algorithm to encrypt the forging data.

[0026] After completing the encapsulation and encryption, the IoT gateway transmits the forging data to the data center through the HTTPS protocol. The data center can receive and store the forging data from the IoT gateway in real time.

[0027] In addition, the communication between the IoT gateway and the data center uses a redundant design and a failover mechanism to ensure that there is no single point of failure during data transmission. When an IoT gateway fails or the network connection is unstable, other gateways can take over the data transmission task to ensure data continuity and integrity.

[0028] Step S300: preprocessing the received forging data in the data center.

[0029] Specifically, data preprocessing includes data cleaning, outlier detection, and data normalization.

[0030] Data cleaning aims to remove or correct erroneous, duplicate or missing values ​​in forging data. Since sensors may be affected by environmental interference or equipment failure during the acquisition process, abnormal or missing data may appear. Therefore, during the data cleaning stage, the received forging data will be checked to identify and correct erroneous data values, and duplicate or missing data records will be filled or deleted to ensure data integrity and consistency.

[0031] Outlier detection performs statistical analysis on forging data, determines the normal distribution range of the data, and identifies data values ​​that exceed the normal range as outliers. For identified outliers, the system marks or deletes them to avoid adverse effects on subsequent analysis.

[0032] Data normalization aims to convert forging data into a unified scale for subsequent analysis and comparison. Since the various parameters in the forging data (such as forging pressure, forging temperature, etc.) have different dimensions and value ranges, direct analysis may lead to inaccurate or difficult to interpret results. Therefore, in the data normalization stage, they are converted into dimensionless values ​​and ensure that the value ranges of all parameters are within a comparable range.

[0033] In this embodiment, the drop_duplicates() function in the Pandas library is first used to remove duplicate records in the data set. By comparing the records in the data set row by row, identical records are identified and deleted, thereby ensuring the uniqueness of each record in the data set.

[0034] Secondly, for missing values ​​in the data, we use the mean filling method to process them, using the fillna() function in the Pandas library and passing in data.mean() as the filling parameter. This function calculates the mean of the non-missing values ​​of each numeric column containing missing values ​​and replaces all missing values ​​in the column with this mean.

[0035] Next, we use the Z-score method to detect outliers in the dataset. Specifically, we use the stats.zscore() function in the SciPy library to calculate the Z-score of each numerical feature in the dataset. The Z-score represents the difference between each feature value and its mean divided by the standard deviation. Then we set a threshold of 3, and treat records with an absolute value of the Z-score greater than the threshold as outliers.

[0036] For the detected outliers, we chose to delete them. Specifically, we created a Boolean DataFrame abnormal_data based on the Z score returned by the stats.zscore() function, where True indicates that the record is an outlier. Then, we deleted all records containing outliers using the Boolean indexing function in the Pandas library.

[0037] Finally, the Min-Max normalization method is used to achieve data normalization. Specifically, a lambda function is defined, which accepts a numeric Series as input, calculates the minimum and maximum values ​​of the Series, and then subtracts the minimum value from each value and divides it by the difference between the maximum and the minimum value to obtain the normalized value. Then, the apply() function in the Pandas library is used to apply the anonymous function to each numeric feature in the dataset to achieve data normalization.

[0038] Step S400: Analyze the preprocessed forging data using a deep learning-based algorithm to determine whether there is equipment failure or process risk.

[0039] Specifically, the deep learning-based algorithm adopts the CTA-GNN algorithm, and uses historical forging data, equipment failure data and forging result data to train the CTA-GNN model. The historical forging data, equipment failure data and forging result data are obtained from the data center.

[0040] The historical forging data includes forging force data, historical forging times data, historical forging current data, historical forging temperature data and historical equipment vibration data.

[0041] The CTA-GNN model structure includes input layer, feature extraction layer, causal-shortcut feature separation layer, parameter adjustment layer and output layer.

[0042] The algorithm receives preprocessed forging data through the input layer, which includes forging force data, forging times data, forging current data, forging temperature data, and equipment vibration data. Then, according to the physical structure of the forging equipment and the interaction between the components, a corresponding graph structure is constructed, in which the nodes represent the various components of the forging equipment, and the edges represent the interactions and connections between these components.

[0043] In the feature extraction layer, graph neural networks are used to extract data features from the input forging data and the constructed graph structure. These features not only include the changes in various physical parameters during the forging process, but also reflect the interactions and influences between equipment components.

[0044] The causal-shortcut feature separation layer splits the data features obtained by the feature extraction layer into causal features and shortcut features through a soft mask mechanism. Among them, the causal features are directly related to the actual physical process in the forging process and can accurately reflect the changes in equipment status and process parameters; while the shortcut features contain noise, misleading information or complex associations with other factors. These features may be inaccurate or disruptive for fault prediction. Through this layer of separation, it is possible to more accurately focus on the features that actually contribute to fault prediction.

[0045] The parameterized adjustment layer performs parameterized backdoor adjustments on causal features and shortcut features to reduce the negative impact of shortcut features on prediction results. By optimizing the parameter settings in the algorithm, the model can more accurately identify and use causal features for prediction, while reducing the interference of shortcut features on prediction results.

[0046] Finally, the output layer uses a classifier to predict the probability of each equipment failure or process risk. This classifier is trained and optimized based on the features extracted and processed by the previous layers, and can output a prediction result that includes various failure types and risk probabilities.

[0047] In this embodiment, the structure of the CTA-GNN model is first defined, which includes an input layer, a feature extraction layer, a causal-shortcut feature separation layer, a parameterization adjustment layer, and an output layer.

[0048] The input layer receives the node feature matrix x and the edge index matrix edge_index as input. The dimension of the node feature matrix x is (num_nodes, input_dim), indicating that there are num_nodes nodes and each node has input_dim features; the dimension of the edge index matrix edge_index is (2, num_edges), indicating the connection relationship between nodes.

[0049] In the feature extraction layer, the graph convolution layer (GCNConv) is used to extract features from the input graph data, and the node feature matrix x and the edge index matrix edge_index are input into the graph convolution layer to obtain the extracted feature h. Then the ReLU activation function is used to perform nonlinear transformation on the feature h to enhance the expression ability of the model.

[0050] Next, we enter the causal-shortcut feature separation layer and input the extracted feature h into two fully connected layers to obtain the causal feature causal_features and shortcut feature shortcut_features. The output dimensions of these two fully connected layers are both hidden_dim / / 2, which achieves dimensionality reduction and separation of features.

[0051] Then enter the parameter adjustment layer, use a fully connected layer to perform parameter adjustment on the extracted feature h, and obtain the adjusted feature adjusted_features. The output dimension of this layer is the same as the input dimension, that is, hidden_dim.

[0052] Finally, the output layer is entered to perform average pooling on the adjusted features adjusted_features to obtain the global feature representation. Then, the global feature representation is input into a fully connected layer to obtain the classification result logits. Finally, the classification result logits is normalized using the softmax function to obtain the predicted probability probs.

[0053] Step S500: When it is determined that there is equipment failure or process risk in forging, an early warning message is sent through the data center.

[0054] Specifically, when the probability of equipment failure or process risk output by the deep learning-based algorithm is greater than the preset threshold, the data center generates corresponding warning information; the warning information is sent to relevant personnel via SMS, email or system notification; the data center records the sending status and receipt confirmation of the warning information.

[0055] In this embodiment, after the CTA-GNN model analyzes the preprocessed forging data, if it is determined that the probability of a certain equipment failure or process risk in the forging process exceeds 50%, the data center will respond immediately and generate detailed warning information.

[0056] First, the warning information will be sent to the mobile phones of relevant personnel through the SMS platform, including production supervisors, equipment maintenance personnel and other staff; secondly, the warning information will also be sent to the mailboxes of relevant personnel through the email system; in addition, the data center will also send system notifications through the company's internal management system, which will appear in the form of notification bar messages on the computer screens or mobile devices of relevant personnel, further ensuring the visibility and accessibility of the warning information.

[0057] The data center will also record the sending status and receipt confirmation of the warning information. After sending the warning information, the data center will monitor the sending status of the information in real time, including whether it was successfully sent, whether it failed to be sent due to network problems or equipment failures, etc. At the same time, the warning information has a built-in confirmation mechanism, and the recipient can confirm the operation after reading the information.

[0058] If the warning information fails to be successfully delivered or the recipient fails to confirm it within the specified time, the data center triggers a further reminder mechanism, which includes sending the warning information again and contacting the relevant personnel directly by phone to ensure that the warning information can be responded to and processed in a timely manner.

[0059] Embodiment 2: Figure 2 As shown, the present invention provides a forging monitoring system, the system comprising: Data acquisition module 1, used to arrange IoT sensors on the forging equipment to collect forging data in real time during the forging process, including forging pressure data, forging times data, forging current data, forging temperature data and equipment vibration data information; Data transmission module 2, used to send the forging data collected by the IoT sensor to the IoT gateway by wireless means, and the IoT gateway encapsulates and encrypts the received data and transmits it to the data center; Data processing module 3, used for preprocessing the received forging data in the data center, including data cleaning, outlier detection and data normalization; Intelligent analysis module 4 uses a deep learning model based on the CTA-GNN algorithm to analyze the preprocessed forging data to determine whether there is equipment failure or process risk. The CTA-GNN model structure includes an input layer, a feature extraction layer, a causal-shortcut feature separation layer, a parameterized adjustment layer, and an output layer; Early warning management module 5, when the probability of equipment failure or process risk output by the intelligent analysis module is greater than the preset threshold, the corresponding early warning information is generated according to the early warning rules, and sent to relevant personnel via SMS, email or system notification, and the sending status and receipt confirmation of the early warning information are recorded at the same time.

[0060] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A forging monitoring method, characterized in that: The method comprises: Step S100: arranging IoT sensors on the forging equipment to collect forging data in real time during the forging process; Step S200: The collected forging data is transmitted to the data center via the Internet of Things gateway; Step S300: performing data preprocessing on the received forging data in the data center; Step S400: Analyze the preprocessed forging data using a deep learning-based algorithm to determine whether there is equipment failure or process risk; Step S500: When it is determined that there is equipment failure or process risk in forging, an early warning message is sent through the data center.

2. The forging monitoring method according to claim 1, characterized in that: The forging data includes forging force data, forging times data, forging current data, forging temperature data and equipment vibration data.

3. The forging monitoring method according to claim 1, characterized in that: The step S200 specifically includes: The IoT sensor sends the collected forging data to the IoT gateway wirelessly; The IoT gateway encapsulates and encrypts the received forging data; The IoT gateway transmits the encapsulated and encrypted forging data to the data center.

4. The forging monitoring method according to claim 1, characterized in that: The data preprocessing includes data cleaning, outlier detection and data normalization.

5. The forging monitoring method according to claim 1, characterized in that: The deep learning-based algorithm adopts the CTA-GNN algorithm and uses historical forging data, equipment failure data and forging result data to train the CTA-GNN model.

6. The forging monitoring method according to claim 5, characterized in that: The CTA-GNN model structure includes: The input layer is used to receive forging data and construct the corresponding graph structure according to the physical structure of the forging equipment. The nodes represent the components of the forging equipment, and the edges represent the interactions between the components. The feature extraction layer uses graph neural network to extract the data features of the forging process from the input forging data and graph structure; The causal-shortcut feature separation layer splits the data features obtained by the feature extraction layer into causal features and shortcut features through a soft mask mechanism. The causal features are directly related to the actual physical process in the forging process, while the shortcut features contain noise and misleading information. The parameterized adjustment layer performs parameterized backdoor adjustments on causal features and shortcut features to reduce the negative impact of shortcut features on prediction results; The output layer uses the classifier to predict the probability of each equipment failure or process risk.

7. The forging monitoring method according to claim 6, characterized in that: Step S500 specifically includes: When the probability of equipment failure or process risk output by the deep learning-based algorithm is greater than the preset threshold, the data center generates corresponding warning information; Early warning information is sent to relevant personnel via SMS, email or system notification; The data center records the sending status and receipt confirmation of the warning information.

8. A forging monitoring system, characterized in that: The system comprises: The data acquisition module is used to arrange IoT sensors on the forging equipment to collect forging data in real time during the forging process, including forging pressure data, forging times data, forging current data, forging temperature data, and equipment vibration data information; The data transmission module is used to send the forging data collected by the IoT sensor to the IoT gateway wirelessly. The IoT gateway encapsulates and encrypts the received data and transmits it to the data center. A data processing module is used to perform data preprocessing on the received forging data in the data center, including data cleaning, outlier detection and data normalization processing; The intelligent analysis module uses a deep learning model based on the CTA-GNN algorithm to analyze the preprocessed forging data to determine whether there is equipment failure or process risk. The CTA-GNN model structure includes an input layer, a feature extraction layer, a causal-shortcut feature separation layer, a parameterized adjustment layer, and an output layer. Early warning management module: When the probability of equipment failure or process risk output by the intelligent analysis module is greater than the preset threshold, the corresponding early warning information is generated according to the early warning rules and sent to relevant personnel via SMS, email or system notification, while recording the sending status and receipt confirmation of the early warning information.

Images