Fault prediction and self-adaptive control method and system for forging and pressing equipment based on Internet of Things

Abstract

The invention discloses a forging and pressing equipment fault prediction and self-adaptive control method and system based on the Internet of Things, and relates to the related field of forging and pressing equipment maintenance technology.According to the method, an Internet of Things platform is built, basic data of forging and pressing equipment is uploaded, stored and processed, and the method is composed of an equipment layer, an edge computing layer, a network layer and a cloud platform layer; a plurality of sensors of the equipment layer collect various data in operation of the forging and pressing equipment in real time, the data are converted into multivariable time sequence data, the preprocessed time sequence data are input into a pre-training time sequence anomaly detection model through the edge calculation layer, and a fault prediction result of the equipment is output; and the network layer uploads the fault prediction result of the forging and pressing equipment to the cloud platform layer, operates an intelligent algorithm to analyze the prediction result and the operation state of the equipment, calculates an optimal control parameter, generates an optimal control instruction, and realizes self-adaptive control of the forging and pressing equipment. The problems of response lag, insufficient data analysis capability and low efficiency of an existing method are solved, and the prediction precision and the control efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of forging equipment maintenance technology, and in particular to a method and system for fault prediction and adaptive control of forging equipment based on the Internet of Things. Background Technology

[0002] As core equipment in the manufacturing industry, the operating status of forging equipment is directly related to production efficiency and product quality. With the advancement of industrial technology, fault prediction and health management of forging equipment have become increasingly important.

[0003] Traditional methods of periodic maintenance and reactive repair for forging equipment are ill-suited to the pace of modern production. Periodic maintenance may result in either over-maintenance or under-maintenance, while reactive repair is often delayed. The intelligent transformation of the manufacturing industry is driving the intelligent operation and maintenance of forging equipment. Existing technologies can assess the health status of equipment and predict the failure of core components through real-time data monitoring, thereby enabling predictive maintenance, reducing unplanned downtime, and optimizing maintenance resources. However, equipment failures exhibit abnormal patterns across various data sources, and existing methods struggle to effectively integrate multi-source data and establish data correlations for accurate fault prediction. Furthermore, the control parameters of forging equipment (such as pressure, speed, and position) are usually set during the commissioning phase and remain fixed once set. However, during long-term operation, the equipment performance will change due to wear, aging, and component heating. This static control mode lacks a data analysis process and cannot adapt to these dynamic changes, leading to fluctuations in product quality and a decrease in the pass rate. Operators often rely on experience to manually adjust parameters, which lacks scientific rigor and consistency. Summary of the Invention

[0004] To address the technical problems of the prior art, this application provides a method and system for fault prediction and adaptive control of forging equipment based on the Internet of Things (IoT). By building an IoT platform and applying a time-series anomaly detection model based on deep learning, the operating data of the forging equipment is monitored in real time, enabling predictive maintenance of the forging equipment, timely detection of potential equipment faults, and adaptive control of the forging equipment through dynamic optimization algorithms.

[0005] This application provides a fault prediction and adaptive control method for forging equipment based on the Internet of Things, including: (1) Build an Internet of Things platform to upload, store and process basic data of forging equipment. The platform consists of an equipment layer, an edge computing layer, a network layer and a cloud platform layer. (2) Activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data in real time during the operation of the forging equipment; (3) Convert the forging equipment operation data collected by the sensor into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the equipment fault prediction results; (4) The network layer of the Internet of Things platform provides a network channel to upload the fault prediction results of the forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters; (5) Generate optimized control instructions based on the calculated optimal control parameters, transmit them to the edge computing layer, realize adaptive control of the forging equipment, and feed back the execution results to optimize the parameter configuration of the control algorithm.

[0006] Furthermore, the IoT platform adopts a classic four-layer architecture of "cloud-pipe-edge-device" to ensure comprehensive data collection, efficient transmission, and intelligent processing. The device layer deploys various sensors to sense the operating status of the forging equipment; the edge computing layer performs preliminary processing and intelligent decision-making at the data source, achieving low-latency response; the network layer establishes a secure, reliable, and bidirectional data transmission channel, connecting to industrial fiber optic leased lines and combining with commercial broadband to form a primary and backup link; the cloud platform layer provides massive data storage, computational analysis, and visualization, enabling adaptive control optimization.

[0007] Furthermore, the time series anomaly detection model employs a variational autoencoder and a long short-term memory network to achieve short-window and long-term trend time series analysis. Given a time series... ,in Indicates the first The time-series value of timestamps, with a dimension size of [missing information]. ,Include Information from different channels will be used to define the timing anomaly detection task as follows: time, , using a length of The historical time series, that is ,in, and They represent the first and Using time-series values ​​at each moment to predict a binary output ,in express An anomaly occurred, thus enabling online anomaly detection.

[0008] This application also provides an Internet of Things-based fault prediction and adaptive control system for forging equipment, including: IoT Platform Building Module: Used to build an IoT platform, upload, store and process basic data of forging equipment. The platform consists of a device layer, an edge computing layer, a network layer and a cloud platform layer. Forging equipment operation data acquisition module: used to activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data during the operation of the forging equipment in real time; Forging equipment fault prediction module: It is used to convert the operating data of forging equipment collected by sensors into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the fault prediction results of the equipment. Forging equipment control parameter optimization module: used to upload the fault prediction results of forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters; Control command generation and execution module: This module generates optimized control commands based on the calculated optimal control parameters, transmits them to the edge computing layer, enables adaptive control of the forging equipment, and feeds back the execution results to optimize the parameter configuration of the control algorithm.

[0009] This application also proposes an IoT-based forging equipment fault prediction and adaptive control device, the device comprising: a memory, a processor, and programs such as an IoT-based forging equipment fault prediction and adaptive control algorithm stored in the memory and executable on the processor, wherein the IoT-based forging equipment fault prediction and adaptive control algorithm and other programs are steps for implementing the IoT-based forging equipment fault prediction and adaptive control method described above.

[0010] This application also provides a computer program product, which includes programs such as an IoT-based fault prediction and adaptive control algorithm for forging equipment. When the IoT-based fault prediction and adaptive control algorithm is executed by a processor, it implements the IoT-based fault prediction and adaptive control method for forging equipment as described above.

[0011] This application discloses the following technical effects: This application provides a method and system for fault prediction and adaptive control of forging equipment based on the Internet of Things (IoT). An IoT platform is built, leveraging a cloud-edge collaborative structure to store massive amounts of forging equipment data while providing powerful computing and data analysis capabilities. Fault prediction of forging equipment is achieved through time-series anomaly detection. Specifically, time-series anomaly detection is implemented using a deep learning-based anomaly detection model. This model employs a hybrid anomaly detection method, combining the representation learning capabilities of deep generative models with the time modeling capabilities of recurrent neural networks, improving the accuracy of fault prediction and significantly reducing manual inspection time. After receiving the fault prediction results output by the model, the cloud platform layer of the IoT platform automatically triggers intelligent algorithms to evaluate the predicted product quality values ​​under different control parameters. It quickly searches for the optimal control parameter configuration strategy that maximizes product quality and operating efficiency, ensuring rapid response of adaptive control, avoiding prolonged downtime for maintenance, and reducing resource consumption. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0013] Figure 1 This is a flowchart illustrating the IoT-based fault prediction and adaptive control method for forging equipment provided in an embodiment of this application.

[0014] Figure 2 This is a schematic diagram of the structure of the IoT-based fault prediction and adaptive control system for forging equipment provided in the embodiments of this application. Detailed Implementation

[0015] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the following description, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0018] Example 1: This application provides a method for fault prediction and adaptive control of forging equipment based on the Internet of Things, such as... Figure 1 As shown, the method includes: Step S10: Build an IoT platform to upload, store, and process basic data of the forging equipment. The platform consists of an equipment layer, an edge computing layer, a network layer, and a cloud platform layer.

[0019] In this embodiment, the IoT platform adopts the classic "cloud-pipe-edge-device" four-layer architecture to ensure comprehensive data collection, efficient transmission and intelligent processing.

[0020] The equipment layer is used to collect the physical status and operating parameters of the equipment, and consists of various sensors and PLC controllers. Based on the needs of predictive maintenance, the physical quantities to be monitored and the accuracy requirements are determined, and appropriate sensor models are selected. Sensor types include vibration sensors, current sensors, temperature sensors, pressure transmitters, and displacement sensors. The physical quantities are converted into digital signals that are easy for the model to process. According to the equipment structure diagram and the fault mode of the equipment to be detected, the sensors are installed in the optimal position. Standard industrial protocols are configured for the PLC controller to achieve communication connection with the network layer.

[0021] The edge computing layer performs preliminary processing and intelligent decision-making at the data source, achieving low-latency response. It consists of an edge intelligent gateway and edge computing software. The gateway has multiple interfaces to connect to various sensors and PLC controllers. The edge computing software consists of a data acquisition module, a data preprocessing module, and a data analysis module. The data acquisition module configures and starts the data acquisition software to read digital signals output by the sensors from the data acquisition card. The data preprocessing module performs preliminary processing on the data to facilitate calculation and analysis. The data analysis module embeds pre-trained timing anomaly detection to perform real-time processing and analysis on the received sensor data.

[0022] The network layer establishes a secure, reliable, and bidirectional data transmission channel, connects to industrial fiber optic leased lines, and combines commercial broadband to form a primary and backup link. When the wired network is interrupted, it automatically switches to the wireless network to ensure that no data is lost. Hardware devices include switches, routers, and industrial firewalls. Network bandwidth is planned by assessing data volume. The industrial firewall only allows the edge gateway to access the cloud platform's specific IP address through specific ports.

[0023] The cloud platform layer provides massive data storage, computing analysis, and visualization, enabling adaptive control optimization. It serves as the center for data aggregation, storage, in-depth analysis, and application, containing multiple time-series and relational databases to store data from various forging and pressing equipment. It also provides a big data and AI platform to run intelligent optimization algorithms and create environments for training and evaluating time-series anomaly detection models. Furthermore, it offers application services to enable fault warnings and data API queries.

[0024] Step S20: Activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data during the operation of the forging equipment in real time.

[0025] In this embodiment, the main drive bearing is used as the fault prediction target. Bearing faults manifest as fatigue spalling and wear, generating high-frequency vibration and impact signals and a sharp increase in temperature. The data sources are determined to be vibration data, temperature data, speed data, and load data. Vibration data, as a core feature for fault prediction, needs to be acquired at high frequency. It is obtained by analyzing the impact signal through an accelerometer installed on the bearing housing. Temperature data is obtained through a temperature sensor installed on the bearing housing. The speed and load data are the spindle speed and working pressure read from the PLC controller and used for normalization analysis to eliminate interference from changes in working conditions. Set the sensor's range, accuracy, and frequency response; create a data list based on the data source, labeling the data source, data point name, and sampling frequency; correctly connect the sensor's output signal line to the data acquisition unit; power all devices; configure parameters for each connected channel in the edge gateway to match the sensor; receive real-time data acquired by the data acquisition unit; and synchronously upload the data to the cloud platform layer for storage.

[0026] Step S30: Convert the operating data of the forging equipment collected by the sensor into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the fault prediction result of the equipment.

[0027] In this embodiment, the forging equipment operation data received by the edge computing layer arrives asynchronously and unaligned. Since the sampling frequencies of different data sources vary, their arrival times differ, necessitating time alignment processing. Based on the system clock of the edge gateway, a fixed time interval is defined to generate a series of consecutive timestamps. For each time interval, all data points arriving during this period are marked as belonging to the current time, and data resampling is performed. For data with a sampling frequency lower than the threshold, if there is no new data at the current moment, the previous valid value is used to fill the gap; for data with a sampling frequency higher than the threshold, an average value is taken within the current time interval to represent the value at the current moment.

[0028] At each aligned time point, a data vector is created containing timestamps and various sensor data; a sliding window is used to construct the model input, setting the window size and sliding step size, and continuously... The data vectors at each time point are stacked to form two-dimensional data, with the dimension represented as... , This represents the number of data types, which are used as inputs to the time series anomaly detection model.

[0029] Data preprocessing includes data cleaning, feature engineering, and data standardization: Data cleaning processes outliers and missing values ​​in multivariate time series data. Based on the threshold values ​​of each physical quantity during the operation of the forging equipment, outlier data points are identified and filled using linear interpolation of the preceding and following valid values. For missing values ​​caused by brief communication interruptions, forward filling is used for repair. In cases of severe missing values, the data for that time period is marked as unusable. Feature engineering, as a key step, extracts features from multivariate time series data that better reflect the state of forging equipment, reducing data dimensionality. Features include time-domain features and frequency-domain features. Taking the fault prediction of the main drive bearing as an example, the time-domain features are the mean and root mean square values ​​of the vibration data, reflecting the average energy level of the pulse signal. Kurtosis, waveform factor, peak factor, and impulse factor are extracted from the pulse signal. These indicators are sensitive to impact-type faults and facilitate model identification. The vibration data is converted from the time domain to the frequency domain through fast Fourier transform, and specific frequency band energy values ​​are extracted as frequency features. The extracted features are used as additional input features for the time-series anomaly detection model.

[0030] The pre-trained time-series anomaly detection model is obtained through the cloud platform layer in the Internet of Things platform. Normal samples from the historical operation data of forging equipment stored in the cloud platform layer are used as training data. Specific sensor data are selected according to the fault prediction target. After time alignment, preprocessing and key feature extraction, multivariate time-series data and key feature data are obtained to form a dataset. The cloud platform provides a model training environment. In the training environment, a time series anomaly detection model is built through code programming, hyperparameters of each part of the model are set and model parameters are randomly initialized, and the optimizer type, initial learning rate and learning rate strategy for model training are configured. The model is trained according to the set training period and sample batch, and the model parameters are updated. The cloud platform automatically records the changes in the loss function and evaluates the model performance after each training period. When the loss function converges, the model parameters with the best performance are saved as the pre-trained time series anomaly detection model.

[0031] In step S40, the network layer of the IoT platform provides a network channel to upload the fault prediction results of the forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters.

[0032] In this embodiment, the cloud platform layer, based on the prediction results and a more comprehensive equipment status, utilizes the powerful computing power of the cloud to calculate the optimal control parameters to compensate for the performance degradation of the forging equipment or optimize product quality.

[0033] Taking the bearing failure of the forging equipment as an example, the cloud platform receives the bearing failure warning information sent by the edge computing layer. The adaptive control service in the cloud is triggered by the warning information, aggregates and calculates the required panoramic data from the database, including the process parameters of the current production task, vibration data within a specific time period, motor current data and temperature data, reads the current position loop PID parameters and speed feedforward gain from the PLC controller from the edge gateway, and calls the digital twin model of the forging equipment. The control parameters to be optimized are the proportional gain, integral gain, and speed feedforward gain of the PLC controller. A search space, i.e. a reasonable adjustment range, is set for each parameter. A comprehensive objective function is defined with the goal of minimizing the controller tracking error, overshoot, settling time, and product thickness variance. The constraints are set as follows: the maximum pressure of the forging equipment does not exceed the safety threshold, and the single cycle time does not exceed the maximum value allowed by the process.

[0034] Run the improved particle swarm optimization algorithm in the digital twin model to perform simulations and find the optimal solution: Create a swarm of particles, each representing a random combination of control parameters, and initialize the particle velocity, individual optimal solution, and global optimal solution of the improved particle swarm optimization algorithm. For each particle, the current parameters are injected into the digital twin model, which has been corrected based on the current fault prediction results to simulate abnormal operating data of the forging equipment. A complete stamping cycle is simulated in a digital twin model, including the entire process of slide acceleration and descent, contact with the workpiece, pressurization, pressure holding and return, to realize the working simulation of the forging equipment; Based on the simulation results, the position tracking curve, pressure curve, product thickness and adjustment time data are extracted, and substituted into the objective function to calculate the fitness value of the current particle. The smaller the value, the closer the set of parameters is to the optimal solution. Each particle compares its fitness value with its own historical best fitness value to update its individual optimal solution. The entire particle swarm compares the fitness values ​​of all particles to find the current global optimal solution. The inertia weights are dynamically adjusted according to the current iteration progress, the velocity and position of each particle are updated, and new parameter combinations are generated. If the global optimal solution does not improve after several consecutive iterations, some particles are randomly selected for mutation, their parameters are randomly perturbed, and a local optimum is found. When the maximum number of iterations is reached, the global optimal solution at this point is used as the optimal combination of control parameters to adjust the operating status of the forging equipment and reduce the possibility of failure.

[0035] Step S50: Generate optimized control instructions based on the calculated optimal control parameters, transmit them to the edge computing layer to achieve adaptive control of the forging equipment, and feed back the execution results to optimize the parameter configuration of the control algorithm.

[0036] Example 2: This embodiment of the invention provides a detailed structure of a time-series anomaly detection model, which consists of a variational autoencoder and a long short-term memory network: A variational autoencoder (VAE) is a generative probabilistic model that simulates the normal patterns of multivariate time-series data from the operation of forging equipment to extract local features. A VAE consists of an encoder and a decoder. A time-series local window of consecutive readings is used as input, and the encoder estimates... The low-dimensional embedding of the window is then modeled through a long short-term memory network to establish temporal context relationships, and the original window is reconstructed through a decoder.

[0037] During training, the variational autoencoder uses normal forging equipment operation data as input features. After multiple nonlinear transformations, linear layer mapping and multi-scale convolution are applied to map the input features to the latent space and calculate the mean vector. Sum of logarithmic variance vector , to obtain input features The corresponding normal distribution in the latent space:

[0038] in, Represents a probability distribution. This represents the probabilistic representation of the input features in a low-dimensional latent space. The standard deviation represents the normal distribution. The identity matrix is ​​used; to achieve backpropagation, differentiable samples are taken from the normal distribution, and the probabilistic representation is further processed as follows: , This indicates element-wise multiplication. It follows a normal distribution with a mean of 0 and a variance of 1; during the training of the variational autoencoder, a rolling window and window sequence are generated from the training data. Represented as , and They represent the first and The time series value at any given moment; For training data ,generate Train an autoencoder model using a rolling sequence. The Long Short-Term Memory (LSTM) network is trained using a rolling window. Running on the low-dimensional embedding of the variational autoencoder output, for the LSTM input sequence ,use express The corresponding embedded LSTM from Take the middle before The model parameters are optimized by minimizing the prediction error of the embedded vectors.

[0039] A pre-trained temporal anomaly detection model was used for real-time fault detection in forging equipment. At any given time, the model analyzes the multivariate time series of the input forging equipment's operation, which includes previous data. The pre-trained model first evaluates the embedded sequences in the sequence using the encoder in the variational autoencoder, and then evaluates the first few historical readings of the embedded sequences. Each embedded vector is used for LSTM to predict the future. The predicted inline vectors are used to reconstruct the data, and finally, the predicted inline vectors and the decoder part of the variational autoencoder are used to reconstruct the data. Using a window sequence and the reconstructed sequence, anomaly scores are calculated. :

[0040] in, Indicates the window sequence index. and They represent the first The reconstructed sequence and the sequence readings of the original input sequence at each time step are used to define a threshold for the anomaly score. This threshold is defined using a validation set that includes both normal and anomaly data. The threshold is then used to determine the anomaly score. It continuously marks abnormal warnings of forging and pressing equipment to enable fault prediction.

[0041] Example 3: The IoT-based fault prediction and adaptive control system for forging equipment provided in this embodiment of the invention can execute the IoT-based fault prediction and adaptive control method for forging equipment provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method, such as... Figure 2 As shown, it includes the following modules: IoT Platform Building Module: Used to build an IoT platform, upload, store and process basic data of forging equipment. The platform consists of a device layer, an edge computing layer, a network layer and a cloud platform layer. Forging equipment operation data acquisition module: used to activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data during the operation of the forging equipment in real time; Forging equipment fault prediction module: It is used to convert the operating data of forging equipment collected by sensors into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the fault prediction results of the equipment. Forging equipment control parameter optimization module: used to upload the fault prediction results of forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters; Control command generation and execution module: This module generates optimized control commands based on the calculated optimal control parameters, transmits them to the edge computing layer, enables adaptive control of the forging equipment, and feeds back the execution results to optimize the parameter configuration of the control algorithm.

[0042] Example 4: This application provides an IoT-based fault prediction and adaptive control device for forging equipment. The IoT-based fault prediction and adaptive control device for forging equipment includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the IoT-based fault prediction and adaptive control methods for forging equipment in Examples 1 and 2 above.

[0043] Example 5: This application provides a computer program product including a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication system, or installed from a storage system. When the computer program is executed by a processing system, it performs the functions defined in the methods of Examples 1 and 2 of this application.

[0044] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A fault prediction and adaptive control method for forging equipment based on the Internet of Things, characterized in that, The method includes: (1) Build an Internet of Things platform to upload, store and process basic data of forging equipment. The platform consists of an equipment layer, an edge computing layer, a network layer and a cloud platform layer. (2) Activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data in real time during the operation of the forging equipment; (3) Convert the forging equipment operation data collected by the sensor into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the equipment fault prediction results; (4) The network layer of the Internet of Things platform provides a network channel to upload the fault prediction results of the forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters; (5) Generate optimized control instructions based on the calculated optimal control parameters, transmit them to the edge computing layer, realize adaptive control of the forging equipment, and feed back the execution results to optimize the parameter configuration of the control algorithm.

2. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 1, characterized in that, In step (1), the equipment layer is used to collect the physical state and operating parameters of the equipment, and consists of a variety of sensors and a PLC controller; the types of sensors include vibration sensors, current sensors, temperature sensors, pressure transmitters and displacement sensors, which convert physical quantities into digital signals that are easy for the model to process. The edge computing layer performs preliminary data processing and intelligent decision-making, and consists of an edge intelligent gateway and edge computing software. The gateway has multiple interfaces to connect to various sensors and PLC controllers. Edge computing software consists of a data acquisition module, a data preprocessing module, and a data analysis module; The network layer establishes a bidirectional data transmission channel, connects to an industrial fiber optic leased line, and combines with commercial broadband to form a primary and backup link. When the wired network is interrupted, it automatically switches to the wireless network to ensure that data is not lost. The hardware devices include switches, routers, and industrial firewalls. The cloud platform layer provides massive data storage, computational analysis, and visualization, enabling adaptive control optimization. It serves as the center for data aggregation, storage, in-depth analysis, and application, containing multiple time-series and relational databases to store data from various forging and pressing equipment. It also provides a big data and AI platform to run intelligent optimization algorithms and create environments for training and evaluating time-series anomaly detection models.

3. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 1, characterized in that, In step (2), the range, accuracy and frequency response of the sensor are set based on the fault prediction target. A data list is prepared according to the data source, and the data source, data point name and sampling frequency are marked. The sensor output signal line is correctly connected to the data acquisition device, power is supplied to all devices, and parameters are configured for each access channel in the edge gateway to match the sensor. The real-time data obtained by the data acquisition device is received and uploaded to the cloud platform layer for storage.

4. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 1, characterized in that, In step (3), a pre-trained time-series anomaly detection model is obtained through the cloud platform layer in the Internet of Things platform. Normal samples in the historical data of forging equipment stored in the cloud platform layer are used as training data. Specific sensor data are selected according to the fault prediction target. After time alignment, preprocessing and key feature extraction, multivariate time-series data and key feature data are obtained to form a dataset for model training. The cloud platform provides a model training environment. In the training environment, a time series anomaly detection model is built through code programming, hyperparameters of each part of the model are set and model parameters are randomly initialized, and the optimizer type, initial learning rate and learning rate strategy for model training are configured. The model is trained according to the set training period and sample batch, and the model parameters are updated. The cloud platform automatically records the changes in the loss function and evaluates the model performance after each training period. When the loss function converges, the model parameters with the best performance are saved as the pre-trained time series anomaly detection model.

5. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 4, characterized in that, The time-series anomaly detection model consists of a variational autoencoder and a long short-term memory network. A variational autoencoder is a generative probabilistic model that simulates the normal patterns of multivariate time series data of forging equipment operation in order to extract local features; The variational autoencoder consists of an encoder and a decoder. It takes a temporal local window of p consecutive readings as input, estimates the q-dimensional low-dimensional embedding through the encoder, models the temporal context relationship through a long short-term memory network, and reconstructs the original window through the decoder.

6. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 5, characterized in that, The variational autoencoder, during training, uses normal forging equipment operating data as input features. After multiple nonlinear transformations, linear layer mapping and multi-scale convolution are applied to map the input features to the latent space, and the mean vector is calculated. Sum of logarithmic variance vector , to obtain input features The corresponding normal distribution in the latent space: in, Represents a probability distribution. This represents the probabilistic representation of the input features in a low-dimensional latent space. This represents the mean of a normal distribution. The standard deviation represents the normal distribution. The identity matrix is ​​used; to achieve backpropagation, differentiable samples are taken from the normal distribution, and the probabilistic representation is further processed as follows: , This indicates element-wise multiplication. It follows a normal distribution with a mean of 0 and a variance of 1; The original input feature structure is recovered through multi-layer nonlinear mapping by the decoder, generating reconstructed features. By calculating the reconstruction loss, the reconstructed features output by the decoder are forced to approach the original input features, thus optimizing the model parameters of the variational autoencoder.

7. The IoT-based fault prediction and adaptive control method for forging equipment as described in claim 1, characterized in that, In step (4), the intelligent algorithm is an improved particle swarm optimization algorithm. The optimal solution is found through simulation, including the following steps: Create a swarm of particles, each representing a random combination of control parameters, and initialize the particle velocity, individual optimal solution, and global optimal solution of the improved particle swarm optimization algorithm. For each particle, the current parameters are injected into the digital twin model, which has been corrected based on the current fault prediction results to simulate abnormal operating data of the forging equipment. A complete stamping cycle is simulated in a digital twin model, including the entire process of slide acceleration and descent, contact with the workpiece, pressurization, pressure holding and return, to realize the working simulation of the forging equipment; Based on the simulation results, the position tracking curve, pressure curve, product thickness and adjustment time data are extracted, and substituted into the objective function to calculate the fitness value of the current particle. The smaller the value, the closer the current parameter combination is to the optimal solution. Each particle compares its fitness value with its own historical best fitness value to update its individual optimal solution. The entire particle swarm compares the fitness values ​​of all particles to find the current global optimal solution. The inertia weights are dynamically adjusted according to the current iteration progress, the velocity and position of each particle are updated, and new parameter combinations are generated. If the global optimal solution does not improve after several consecutive iterations, some particles are randomly selected for mutation, their parameters are randomly perturbed, and a local optimum is found. When the maximum number of iterations is reached, the global optimal solution at this point is used as the optimal combination of control parameters to adjust the operating status of the forging equipment and reduce the possibility of failure.

8. A fault prediction and adaptive control system for forging equipment based on the Internet of Things, characterized in that, The system is used to implement the IoT-based fault prediction and adaptive control method for forging equipment as described in any one of claims 1-7, and the system comprises: IoT Platform Building Module: Used to build an IoT platform, upload, store and process basic data of forging equipment. The platform consists of a device layer, an edge computing layer, a network layer and a cloud platform layer. Forging equipment operation data acquisition module: used to activate various sensors in the IoT platform device layer, determine the data source based on the fault prediction target, and collect various data during the operation of the forging equipment in real time; Forging equipment fault prediction module: It is used to convert the operating data of forging equipment collected by sensors into multivariate time series data, preprocess it through the edge computing layer of the Internet of Things platform, and input the preprocessed time series data into the pre-trained time series anomaly detection model to output the fault prediction results of the equipment. Forging equipment control parameter optimization module: used to upload the fault prediction results of forging equipment to the cloud platform layer, run intelligent algorithms to analyze the prediction results and equipment operating status, and calculate the optimal control parameters; Control command generation and execution module: This module generates optimized control commands based on the calculated optimal control parameters, transmits them to the edge computing layer, enables adaptive control of the forging equipment, and feeds back the execution results to optimize the parameter configuration of the control algorithm.

9. A fault prediction and adaptive control device for forging equipment based on the Internet of Things, characterized in that, The IoT-based forging equipment fault prediction and adaptive control device includes: The system includes a memory, a processor, and an IoT-based forging equipment fault prediction and adaptive control program stored in the memory and executable on the processor. When the IoT-based forging equipment fault prediction and adaptive control program is executed by the processor, it implements the IoT-based forging equipment fault prediction and adaptive control method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes an Internet of Things (IoT) based forging equipment fault prediction and adaptive control program. When the IoT-based forging equipment fault prediction and adaptive control program is executed by the processor, it implements the IoT-based forging equipment fault prediction and adaptive control method as described in any one of claims 1 to 7.

Images