Abnormity detection method for magnesium-based alloy production raw material conveying system

Abstract

The invention relates to the technical field of magnesium-based alloy manufacturing data processing, and provides an anomaly detection method for a magnesium-based alloy production raw material conveying system, which comprises the following steps: collecting basic data of the magnesium-based alloy production raw material conveying system, including equipment operation parameters, raw material state parameters, conveying environment parameters and the like, and carrying out preprocessing and feature extraction; and constructing a data feature library. Then, abnormal points and trend change information thereof in the data are identified by using a statistical analysis model, and a potential abnormal mode is formed. And based on the potential abnormal mode and the historical operation data, training a neural network model to realize abnormal prediction and classification of the real-time data. And finally, by comparing a model prediction result with actual data, the model is optimized by using a genetic algorithm, and the detection accuracy is improved. According to the method, the anomaly detection accuracy of the magnesium-based alloy production raw material conveying system is effectively improved, and a powerful guarantee is provided for stable operation of the magnesium-based alloy manufacturing and producing process.

Description

Technical Field

[0001] The invention relates to the technical field of magnesium-based alloy production data processing, and in particular to an abnormality detection method for a magnesium-based alloy production raw material conveying system. Background Art

[0002] The raw material conveying system for magnesium-based alloy production is responsible for efficiently and safely conveying raw materials such as dolomite, ferrosilicon (reducing agent), fluorite (catalyst), etc. from the storage location to the magnesium-based alloy production line in the magnesium-based alloy production process. The raw material conveying system for magnesium-based alloy production includes various conveying equipment, such as belt conveyors, bucket elevators, etc. In the production process of magnesium-based alloys, the stability and efficiency of the raw material conveying system for magnesium-based alloy production are directly related to the smooth progress of subsequent smelting, casting and other processes. For example, in the magnesium smelting production line, belt conveyors are widely used to convey loose materials or finished items. According to the requirements of the conveying process, it can be a single conveyor or multiple conveyors to form a horizontal or inclined conveying system to meet the needs of different layout types of production lines. By accurately controlling the conveying volume and conveying speed, the raw material conveying system for magnesium-based alloy production ensures the continuity of production.

[0003] The existing abnormality detection method of the raw material conveying system for magnesium-based alloy production uses integrated sensor technology to monitor the operating parameters of the raw material conveying equipment for magnesium-based alloy production in real time. The operating parameters of the conveying equipment include speed, temperature, vibration and other parameters. When the operating parameters of the raw material conveying equipment for magnesium-based alloy production exceed the preset normal range, the existing raw material conveying system for magnesium-based alloy production immediately triggers an alarm to prompt the operator to check and repair. In this way, abnormal conditions in the conveying system can be discovered and handled in a timely manner.

[0004] In the abnormality detection method of the raw material conveying system for the production of magnesium-based alloys, when erroneous data appears in the raw material conveying system for the production of magnesium-based alloys, the existing raw material conveying system for the production of magnesium-based alloys will not be able to extract useful information from massive data, affecting the accuracy of identifying abnormal patterns of the raw material conveying system for the production of magnesium-based alloys. The existing data analysis method cannot cope with the changeable data characteristics. In order to solve this technical problem, the present invention provides an abnormality detection method for the raw material conveying system for the production of magnesium-based alloys. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an abnormality detection method for a magnesium-based alloy production raw material conveying system, which solves the problem that when erroneous data appears in the magnesium-based alloy production raw material conveying system, the existing magnesium-based alloy production raw material conveying system will be unable to extract useful information from massive data, affecting the accuracy of identifying abnormal patterns of the magnesium-based alloy production raw material conveying system, and the existing data analysis method is unable to cope with the problem of changeable data characteristics.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The present invention provides a method for detecting anomalies in a raw material conveying system for magnesium-based alloy production, comprising: Step S101, obtaining basic data of a raw material delivery system for magnesium-based alloy production, wherein the basic data of the raw material delivery system for magnesium-based alloy production includes operating parameters of raw material delivery equipment for magnesium-based alloy production, state parameters of raw materials for magnesium-based alloy production, environmental parameters of raw material delivery for magnesium-based alloy production, state data of the raw material delivery system for magnesium-based alloy production, error data of the raw material delivery system for magnesium-based alloy production, and historical operating data of the raw material delivery system for magnesium-based alloy production; Step S102: preprocessing the basic data of the magnesium-based alloy production raw material transportation system to obtain the preprocessed basic data of the magnesium-based alloy production raw material transportation system, extracting characteristic indicators according to the preprocessed basic data of the magnesium-based alloy production raw material transportation system, and constructing a data characteristic library; Step S103: using a preset statistical analysis model to identify data features in the data feature library, obtaining abnormal points in the data and abnormal point data trend change information, and using the abnormal points in the data and abnormal point data trend change information as potential abnormal patterns; Step S104: training a neural network model according to potential abnormal patterns and historical operation data of the raw material delivery system for magnesium-based alloy production to obtain an abnormality detection model, which is used to perform abnormality prediction and abnormal data classification on real-time data of the raw material delivery system for magnesium-based alloy production; Step S105: collecting real-time data of the raw material transportation system for magnesium-based alloy production, and substituting the real-time data of the raw material transportation system for magnesium-based alloy production as input into the anomaly detection model, the anomaly detection model outputs the anomaly prediction result of the real-time data of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, retrieving the time information of the real-time data anomaly prediction result of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, and obtaining the first time information of the data to be compared, collecting the actual data of the raw material transportation for magnesium-based alloy production according to the time information of the data to be compared, and obtaining the second time information of the data to be compared, comparing the first time information of the data to be compared with the second time information of the data to be compared, and obtaining the error data of the raw material transportation for magnesium-based alloy production, optimizing the anomaly detection model using a genetic algorithm based on the error data of the raw material transportation for magnesium-based alloy production, and obtaining the optimized anomaly detection model, and deploying the optimized anomaly detection model to the raw material transportation system for magnesium-based alloy production.

[0007] Furthermore, in the method for detecting anomalies in a raw material conveying system for magnesium-based alloy production described in the present invention, in step S102, the data feature library includes operating characteristics of raw material conveying equipment for magnesium-based alloy production, state characteristics of raw materials for magnesium-based alloy production, environmental characteristics of raw material conveying environment for magnesium-based alloy production, state characteristics of raw material conveying system for magnesium-based alloy production, and error data characteristics of raw material conveying system for magnesium-based alloy production.

[0008] Furthermore, the abnormality detection method for the magnesium-based alloy production raw material conveying system of the present invention, the step S103 comprises: According to the data feature library, obtain the data feature set to be analyzed and load the preset statistical analysis model; The data feature set to be analyzed is used as input and substituted into a preset statistical analysis model. The preset statistical analysis model calculates the data feature set and identifies a set of abnormal points in the data feature set to be analyzed; Based on the set of abnormal points, a preset statistical analysis model is used to analyze the data trend change information of the abnormal points, and the data trend change information of the abnormal points includes the frequency of occurrence and the change rate of the abnormal points; The abnormal point set and its data trend change information are integrated to obtain the potential abnormal pattern.

[0009] Furthermore, the abnormality detection method for the magnesium-based alloy production raw material conveying system of the present invention, the step S104 comprises: According to the potential abnormal pattern, historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal pattern is obtained from a database of the magnesium-based alloy production raw material conveying system; Merging the potential abnormal mode and the historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal mode to obtain a preprocessed training data set, a test set, and a validation set; Initialize the weights and bias parameters of the neural network model, use the training data set as input, and substitute it into the neural network model for training; Through the iterative gradient descent method, the parameters of the neural network model are adjusted to make the neural network model fit the abnormal pattern in the training data. After the training is completed, the neural network model is evaluated using the test set to obtain the performance index of the neural network model. According to the performance index of the neural network model, the neural network model with the best performance is selected as the anomaly detection model; The selected anomaly detection model is deployed in the raw material delivery system for magnesium-based alloy production to perform anomaly prediction and abnormal data classification on the real-time data of the raw material delivery system for magnesium-based alloy production.

[0010] Furthermore, the abnormality detection method for the magnesium-based alloy production raw material conveying system of the present invention, the step S105 comprises: The real-time data of the raw material conveying system for magnesium-based alloy production includes real-time operating parameters of the raw material conveying equipment for magnesium-based alloy production, real-time status parameters of raw materials for magnesium-based alloy production, real-time environmental parameters of raw material conveying for magnesium-based alloy production, real-time status data of the raw material conveying system for magnesium-based alloy production, and real-time error data of the raw material conveying system for magnesium-based alloy production. The real-time data of the raw material conveying system for magnesium-based alloy production is identified and abnormal data is marked using an anomaly detection model. The marked abnormal data is summarized to form an abnormal data report, which includes information on the type of abnormality, the time when the abnormality occurred, and the scope of the impact of the abnormality.

[0011] Furthermore, the abnormality detection method for the raw material conveying system for magnesium-based alloy production according to the present invention, the step S105 comprises: According to the time information of the first data to be compared, the actual data of the magnesium-based alloy production raw material delivery system at the time point corresponding to the time information of the first data to be compared is extracted from the actual data collection record of the magnesium-based alloy production raw material delivery system, and the time information of the actual data of the magnesium-based alloy production raw material delivery system is recorded as the second time information of the data to be compared; Aligning the time information of the first data to be compared with the time information of the second data to be compared, so that the time information of the first data to be compared corresponds to the time information of the second data to be compared; If there is a time mismatch, the mismatching time point is recorded; For each aligned time point, comparing the time information of the first data to be compared with the time information of the second data to be compared; Calculating the difference between the time information of the first data to be compared and the time information of the second data to be compared to obtain preliminary error data; The preliminary error data are processed, and the preliminary error data are processed including removing abnormal values, calculating the average value or the standard deviation, and obtaining the raw material transportation error data for magnesium-based alloy production.

[0012] Furthermore, the abnormality detection method for the raw material conveying system for magnesium-based alloy production according to the present invention, the step S105 comprises: The obtained magnesium-based alloy production raw material transportation error data is recorded as an error data set; Receive and set optimization goals of genetic algorithms; Initialize the population: According to the parameter space of the anomaly detection model, initialize the population of the genetic algorithm. Each individual in the population represents a set of model parameters. Use the error data set to calculate the fitness of each individual in the population; Selection operation: According to the fitness, the selection operation is performed to select individuals with good performance as parents; Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals; Mutation operation, performing mutation operation on offspring individuals; Update the population, use the new individuals after crossover and mutation operations to update the population and form a new population; Iterative optimization, repeatedly executing the steps of calculating fitness, selection, crossover, mutation and updating population until the preset number of iterations is reached or the optimization target reaches the preset target, and the optimized population is obtained; The individual with the highest fitness is extracted from the optimized population. The parameters of the individual with the highest fitness are the optimized anomaly detection model parameters. The optimized anomaly detection model parameters are used to construct the optimized anomaly detection model.

[0013] Beneficial effects of the present invention: The present invention constructs a data feature library to extract various feature information related to the raw material delivery system for magnesium-based alloy production, including equipment operation features, raw material state features, delivery environment features, system state features, and error data features. These feature information provides comprehensive data support for subsequent anomaly detection, improving the accuracy of anomaly detection. By combining statistical analysis models with neural network models, potential abnormal patterns are first identified through statistical analysis models, and then abnormal prediction and classification of real-time data are performed using neural network models, further improving the accuracy of anomaly detection.

[0014] The present invention utilizes a genetic algorithm to optimize the neural network model, and by continuously iteratively adjusting the model parameters, enables the model to better adapt to variable data characteristics, thereby improving the robustness and generalization ability of the model.

[0015] During the model training process, a combination of training sets, validation sets and test sets is used to comprehensively evaluate the model performance.

[0016] The present invention can collect data from the raw material delivery system for magnesium-based alloy production in real time, and use the trained anomaly detection model to predict and classify anomalies in real-time data, and issue anomaly warnings or provide anomaly handling suggestions in a timely manner. By deploying the optimized anomaly detection model to the raw material delivery system for magnesium-based alloy production, continuous monitoring of the system status and anomaly detection are achieved, providing strong support for production decision-making.

[0017] The present invention timely discovers and handles abnormal conditions in the raw material conveying system for magnesium-based alloy production, thereby avoiding production interruptions and safety accidents caused by equipment failure or raw material problems, and improving production efficiency and safety.

[0018] The implementation of the present invention is helpful to promote the digital transformation and upgrading of the magnesium-based alloy production industry and enhance the overall competitiveness and innovation ability of the industry. In summary, the anomaly detection method of the magnesium-based alloy production raw material delivery system provided by the present invention has significant beneficial effects in improving anomaly detection accuracy, enhancing model robustness, realizing real-time anomaly detection and feedback, improving production efficiency and safety, and promoting intelligent production. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0020] Figure 1 A schematic flow chart of an abnormality detection method for a raw material conveying system for magnesium-based alloy production provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. 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. The technical solutions provided by the embodiments of the present invention are described below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described below.

[0022] The present invention provides an abnormality detection method for a raw material conveying system for magnesium-based alloy production, comprising: Step S101, obtaining basic data of a raw material delivery system for magnesium-based alloy production, wherein the basic data of the raw material delivery system for magnesium-based alloy production includes operating parameters of raw material delivery equipment for magnesium-based alloy production, state parameters of raw materials for magnesium-based alloy production, environmental parameters of raw material delivery for magnesium-based alloy production, state data of the raw material delivery system for magnesium-based alloy production, error data of the raw material delivery system for magnesium-based alloy production, and historical operating data of the raw material delivery system for magnesium-based alloy production; Equipment operating parameter data source: Determine the sensor location and data interface of all key conveying equipment (such as belt conveyors, bucket elevators, etc.) in the raw material conveying system of magnesium-based alloy production, and the operating parameters of the equipment, such as speed, temperature, vibration, current, voltage, etc.

[0023] Raw material status parameter data source: determine the monitoring points during the raw material storage and transportation process, and obtain the raw material status parameters such as weight, humidity, particle size distribution, etc. by installing corresponding sensors or detection devices.

[0024] Transportation environment parameter data source: Set up environmental monitoring points on the transportation line to obtain environmental parameters that affect the transportation process, such as temperature, humidity, dust concentration, noise level, etc.

[0025] System status data data source: obtain the overall status data of the magnesium-based alloy production raw material conveying system from the system control center or data management platform, including equipment operation status, fault alarm information, system maintenance records, etc.

[0026] Error data data source: Establish an error data recording mechanism to automatically capture and record error data that occurs during system operation, such as sensor failure, data transmission interruption, data anomaly, etc.

[0027] Historical operation data data source: Past operation data, including normal operation data and abnormal operation data, are extracted from the system's historical database for subsequent data analysis and model training.

[0028] Depending on the data source, appropriate data collection methods are used, such as directly reading sensor data, extracting data from system logs, querying historical data from the database, etc.

[0029] Develop a unified data transmission protocol so that data collected from different data sources can be transmitted to the data processing center in a unified format to facilitate subsequent data integration and analysis.

[0030] Determine a reasonable data collection frequency based on the system's real-time requirements and the frequency of data changes.

[0031] The data collected from different data sources are integrated to form a complete basic data set for the raw material transportation system of magnesium-based alloy production. During the integration process, the data needs to be pre-processed by deduplication, denoising, format conversion and other operations.

[0032] Store the integrated data in a secure and reliable data storage system, such as a relational database, NoSQL database, or distributed file system. At the same time, establish a data backup and recovery mechanism to ensure data security and availability.

[0033] Regularly check the integrity of the collected data to ensure that there is no missing or lost data. Verify the accuracy of the collected data by comparing historical data, cross-validation, etc. Correct or process any data errors or outliers found in a timely manner. Monitor the timeliness of the data so that the collected data can reflect the latest status of the system in a timely manner, and clean or mark outdated or invalid data in a timely manner.

[0034] Step S102: preprocessing the basic data of the magnesium-based alloy production raw material transportation system to obtain the preprocessed basic data of the magnesium-based alloy production raw material transportation system, extracting characteristic indicators according to the preprocessed basic data of the magnesium-based alloy production raw material transportation system, and constructing a data characteristic library; Data cleaning in data preprocessing; Missing value processing: Check whether there are missing values ​​in the data. Missing values ​​can be processed by filling (such as mean filling, median filling, front and back value filling, etc.), interpolation or deletion.

[0035] Outlier processing: Identify and process outliers in the data, which may be caused by sensor failure, data transmission errors, or unexpected situations in actual operation. Statistical methods (such as 3σ criteria, box plots, etc.) or methods based on domain knowledge can be used to judge and process outliers.

[0036] Duplicate value processing: Check whether there are duplicate records in the data. Duplicate records can be directly deleted or merged.

[0037] Data standardization / normalization in data preprocessing; Standardization: Convert data of different dimensions to the same dimension, usually by subtracting the mean and dividing by the standard deviation, so that the data follows a standard normal distribution.

[0038] Normalization: Scale the data to a specified range (such as [0,1]) to facilitate subsequent data processing and analysis.

[0039] Data dimensionality reduction in data preprocessing; For high-dimensional data, methods such as principal component analysis (PCA) and linear discriminant analysis (LDA) can be used for dimensionality reduction to reduce data redundancy and improve data processing efficiency.

[0040] Feature extraction in data preprocessing: Select features with high correlation with target variables (such as abnormal conditions) by calculating feature correlation coefficients, mutual information, etc. Based on the expertise and experience of raw material delivery systems for magnesium-based alloy production, select features that have a significant impact on system operation.

[0041] New features are constructed based on the original features to better describe the state of the system operation. For example, features such as the rate of change of speed and the range of temperature fluctuation can be constructed.

[0042] Transform the original features to extract more useful information. For example, operations such as differencing, smoothing, or Fourier transform can be performed on time series data.

[0043] The extracted feature indicators are classified and stored, such as equipment operation characteristics, raw material status characteristics, transportation environment characteristics, system status characteristics, and error data characteristics. Provide detailed descriptions and labels for each feature, including feature name, physical meaning, calculation method, data type and other information, to facilitate subsequent data analysis and model training. As the system runs and data accumulates, regularly update and maintain the data feature library, add new feature indicators, and delete features that are no longer important.

[0044] Step S103: using a preset statistical analysis model to identify data features in the data feature library, obtaining abnormal points in the data and abnormal point data trend change information, and using the abnormal points in the data and abnormal point data trend change information as potential abnormal patterns; Load the preset statistical analysis model; Model selection: Select an appropriate statistical analysis model based on the characteristics of the data and analysis requirements. Common statistical analysis models include distance-based models (such as DBSCAN, K-means, etc.), density-based models (such as LOF, IsolationForest, etc.), and time series analysis models (such as ARIMA, Holt-Winters, etc.).

[0045] Parameter setting: Set the parameters of the selected statistical analysis model, including distance threshold, number of clusters, outlier definition, etc., to ensure that the model can effectively identify outliers in the data.

[0046] The data feature sets to be analyzed are obtained from the data feature library. These feature sets should cover various aspects of the raw material transportation system for magnesium-based alloy production, such as equipment operating parameters, raw material state parameters, transportation environment parameters, etc.

[0047] The data feature set to be analyzed is used as input and substituted into the preset statistical analysis model. The model calculates and analyzes the data and identifies a set of abnormal points in the data.

[0048] Analysis of abnormal point data trend change information: Based on the identified abnormal point set, use the statistical analysis model to further analyze the abnormal point data trend change information. This trend change information includes the frequency of occurrence, change rate, duration, etc. of the abnormal point. Based on the abnormal point data trend change information, identify potential trend patterns. These trend patterns may represent some abnormal state or potential risks of system operation.

[0049] Integration of abnormal point set and trend change information: Integrate the identified abnormal point set and its data trend change information to form a potential abnormal pattern. The potential abnormal pattern is a preliminary description and summary of the abnormal state of the system. Provide a label and description for each potential abnormal pattern, including abnormal type, occurrence time, impact range and other information, to facilitate subsequent analysis and processing.

[0050] Iterative optimization of outlier identification and trend analysis: Evaluate the preset statistical analysis model based on the accuracy and practicality of potential anomaly patterns. If the model performs poorly, adjust the model or select a more appropriate model. As new data continues to accumulate, update the data feature library regularly and re-execute the outlier identification and trend analysis steps. Identifying potential trend patterns based on the data trend change information of outliers is an important process involving data analysis and pattern recognition. The following are the detailed steps and methods: Analysis of data trend change information: First, organize the set of identified abnormal points, including the value of the abnormal point, timestamp and other basic information. Calculate the rate of change of the abnormal point value over time, which helps to identify whether the development trend of the abnormal point is a rapid increase, a rapid decrease or a slow change. Record the length of time that the abnormal point persists. The abnormal point that exists for a long time may represent a more serious system problem. Check whether the abnormal point shows the characteristics of periodic occurrence, which may be related to the periodic factors of system operation (such as production cycle, maintenance cycle, etc.).

[0051] Use charts (such as line charts, scatter plots, box plots, etc.) to visualize the data trend change information of abnormal points, which helps to identify potential trend patterns more intuitively.

[0052] Potential trend pattern recognition: If the value of an abnormal point gradually increases over time, it may represent the deterioration of a certain performance indicator (such as equipment wear, increased energy consumption, etc.). If the value of an abnormal point gradually decreases over time, it may represent the improvement of a certain performance indicator or the recovery of the system status. If the value of an abnormal point shows a fluctuating change, it may represent the instability of the system status or external interference.

[0053] Trend pattern recognition method: Use statistical indicators (such as mean, standard deviation, skewness, kurtosis, etc.) to describe and analyze the data trend change information of abnormal points, and identify trend patterns that significantly deviate from the normal range. Train machine learning models (such as time series prediction models, classification models, etc.) to predict and classify the data trend change information of abnormal points and identify potential trend patterns.

[0054] When identifying potential trend patterns, it is necessary to analyze them in combination with the business logic of the raw material delivery system for magnesium-based alloy production. For example, the change trend of certain abnormal points may be related to a specific link or operation in the raw material delivery process.

[0055] Use new data sets to verify the identified potential trend patterns. If it is found that the identified trend patterns do not match the actual system operation, the trend patterns need to be adjusted or re-identified.

[0056] Practical application: When potential trend patterns are identified, they can be applied to abnormal warning systems, and warning signals can be issued in a timely manner when the system operation status approaches or reaches the abnormal threshold. Potential trend patterns can provide decision support for system maintenance, optimization and upgrades, helping engineers and managers better understand the system operation status and formulate effective response measures.

[0057] Step S104: training a neural network model according to potential abnormal patterns and historical operation data of the raw material delivery system for magnesium-based alloy production to obtain an abnormality detection model, which is used to perform abnormality prediction and abnormal data classification on real-time data of the raw material delivery system for magnesium-based alloy production; The potential abnormal patterns identified in step S103 are integrated to form label data for model training. These label data will serve as output targets of the neural network model.

[0058] Cleaning and preprocessing of historical operation data: Clean and preprocess the historical operation data of the raw material delivery system for magnesium-based alloy production, including missing value processing, outlier processing, data standardization / normalization, etc., to ensure the quality and consistency of the data.

[0059] Data partitioning: The cleaned and preprocessed historical running data is divided into training set, validation set and test set for training, validation and testing of neural network models.

[0060] According to the characteristics and requirements of the anomaly detection task, select a suitable neural network model. Commonly used neural network models include multi-layer perceptron (MLP), convolutional neural network (CNN), recurrent neural network (RNN) and its variants (such as LSTM, GRU), etc.

[0061] Design the structure and parameters of the neural network model according to data characteristics and task requirements, including the dimension of the input layer, the number of hidden layers and neurons, the dimension of the output layer, the choice of activation function, the definition of the loss function, etc.

[0062] Training process: Use the training set to train the neural network model, and adjust the weights and biases of the model through the backpropagation algorithm and optimization algorithms (such as gradient descent, Adam, etc.) to minimize the loss function.

[0063] Verification and tuning: Use the validation set to verify the model during training and evaluate the model's performance (such as accuracy, recall, F1 score, etc.). Tune the model based on the validation results, including adjusting the model structure, parameters, learning rate, etc.

[0064] Anomaly detection model construction: Combine the trained neural network model with the potential anomaly pattern to build an anomaly detection model. The anomaly detection model can receive real-time data from the raw material delivery system of magnesium-based alloy production as input and output anomaly prediction results and abnormal data classification.

[0065] Set anomaly detection thresholds based on historical operating data and business needs. When the anomaly score of real-time data exceeds the threshold, anomaly warning or anomaly classification is triggered.

[0066] Use the test set to test the anomaly detection model and evaluate the model's performance indicators (such as accuracy, recall, F1 score, ROC curve, AUC value, etc.). Optimize the anomaly detection model based on the test results, including adjusting the threshold, improving the model structure, adding features, etc., to improve the performance and stability of the model.

[0067] The trained anomaly detection model is deployed to the real-time monitoring system of the raw material delivery system for magnesium-based alloy production, and integrated with modules such as data acquisition, data transmission, and data storage. The data of the raw material delivery system for magnesium-based alloy production is received in real time, and the anomaly detection model is used to predict anomalies and classify abnormal data, and abnormal warnings are issued in a timely manner or suggestions for abnormal handling are provided.

[0068] Step S105: collecting real-time data of the raw material transportation system for magnesium-based alloy production, and substituting the real-time data of the raw material transportation system for magnesium-based alloy production as input into the anomaly detection model, the anomaly detection model outputs the anomaly prediction result of the real-time data of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, retrieving the time information of the real-time data anomaly prediction result of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, and obtaining the first time information of the data to be compared, collecting the actual data of the raw material transportation for magnesium-based alloy production according to the time information of the data to be compared, and obtaining the second time information of the data to be compared, comparing the first time information of the data to be compared with the second time information of the data to be compared, and obtaining the error data of the raw material transportation for magnesium-based alloy production, optimizing the anomaly detection model using a genetic algorithm based on the error data of the raw material transportation for magnesium-based alloy production, and obtaining the optimized anomaly detection model, and deploying the optimized anomaly detection model to the raw material transportation system for magnesium-based alloy production.

[0069] Determine the data collection method of the raw material delivery system for magnesium-based alloy production, such as sensor collection, system log extraction, database query, etc. Pre-process the collected real-time data, including data cleaning (noise removal, missing value filling), data conversion (format conversion, unit unification) and data standardization / normalization to ensure data quality and consistency.

[0070] Model input: Substitute the preprocessed real-time data as input into the anomaly detection model.

[0071] Model output: The anomaly detection model outputs the anomaly prediction results of real-time data (such as anomaly probability, anomaly score) and abnormal data classification (such as anomaly type, anomaly cause).

[0072] Result storage: The abnormal prediction results and abnormal data are classified and stored in the database, and the corresponding time information is recorded at the same time to form the time information of the first data to be compared.

[0073] According to the first data time information to be compared, actual data at the corresponding time point is collected from the actual operation record of the magnesium-based alloy production raw material delivery system to form the second data time information to be compared. The first data time information to be compared is compared with the second data time information to be compared, and the error between the predicted value and the actual value is calculated to obtain the magnesium-based alloy production raw material delivery error data.

[0074] Conduct statistical analysis on the error data to identify the main sources and distribution patterns of the errors.

[0075] Optimization goal setting: According to the error analysis results, set the model optimization goals, such as reducing the error mean, reducing the error variance, and improving the anomaly detection accuracy.

[0076] Genetic algorithm selection: Select a suitable genetic algorithm framework, such as the basic genetic algorithm, improved genetic algorithm (such as genetic algorithm with elite strategy, adaptive genetic algorithm, etc.).

[0077] Coding method: Determine the encoding method of model parameters, such as binary encoding, real number encoding, etc.

[0078] Fitness function design: Design a fitness function based on the optimization objective to evaluate the quality of model parameters.

[0079] Genetic operations: Perform genetic operations such as selection, crossover, and mutation to generate new model parameter combinations.

[0080] Iterative optimization: Through multiple iterative optimizations, the model parameter combination that maximizes the fitness function value is found, that is, the optimized anomaly detection model.

[0081] The optimized anomaly detection model is deployed to the raw material delivery system for magnesium-based alloy production to replace the original model. The deployed optimized model is verified, and the performance of the model is evaluated through real-time data collection, anomaly prediction and comparative analysis to see if it has been improved. The optimized model is continuously monitored, its performance is regularly evaluated, and necessary updates and adjustments are made based on actual operation conditions.

[0082] Specifically, in the method for detecting anomalies in a raw material conveying system for magnesium-based alloy production described in the present invention, in step S102, the data feature library includes operating characteristics of the raw material conveying equipment for magnesium-based alloy production, state characteristics of raw materials for magnesium-based alloy production, environmental characteristics of raw material conveying for magnesium-based alloy production, state characteristics of the raw material conveying system for magnesium-based alloy production, and error data characteristics of the raw material conveying system for magnesium-based alloy production.

[0083] Operation characteristics of raw material conveying equipment for magnesium-based alloy production: Equipment operating parameters: such as speed, temperature, pressure, vibration, etc. These parameters reflect the operating status and working performance of the equipment.

[0084] Equipment operating time: records the cumulative operating time of the equipment, which is used to evaluate the degree of wear and tear of the equipment and the remaining service life.

[0085] Equipment maintenance records: including equipment maintenance history, parts replacement records, etc., providing a basis for equipment failure prediction and maintenance planning.

[0086] Characteristics of raw material status in magnesium-based alloy production: Raw material composition: record the chemical composition, physical properties, etc. of the raw materials to ensure that the raw materials meet production requirements.

[0087] Raw material temperature and humidity: monitor the storage environment of raw materials to prevent them from deteriorating due to environmental factors.

[0088] Raw material flow and speed: Monitor the flow and speed of raw materials during transportation to ensure the stability and efficiency of the transportation process.

[0089] Characteristics of raw material transportation environment for magnesium-based alloy production: Ambient temperature and humidity: reflect the climatic conditions of the environment in which the conveying system is located, and have an important impact on the operation of the equipment and the storage of raw materials.

[0090] Environmental noise and vibration: Monitor environmental noise and vibration levels to prevent adverse effects on equipment operation and personnel health.

[0091] Environmental cleanliness: Maintain the cleanliness of the conveying environment to prevent debris from entering the system and causing malfunctions.

[0092] Characteristics of raw material delivery system in magnesium-based alloy production: System pressure and flow: monitor the overall pressure and flow of the system to ensure stable operation of the system.

[0093] System control parameters: such as control signals, set values, etc., reflect the control status and response capabilities of the system.

[0094] System operation status indication: such as running, stopping, fault and other status indications, providing operators with intuitive system status information.

[0095] Characteristics of error data in raw material delivery system for magnesium-based alloy production: Error code and error message: record the error code and error message of the system to provide a basis for troubleshooting and repair.

[0096] Error occurrence time and frequency: Count the time and frequency of error occurrence, and analyze the patterns and trends of errors.

[0097] Error impact scope: Assess the impact of the error on system operation and production, and determine the severity and priority of the fault.

[0098] Specifically, the method for detecting anomalies in a raw material conveying system for magnesium-based alloy production according to the present invention, step S103, comprises: According to the data feature library, obtain the data feature set to be analyzed and load the preset statistical analysis model; The data feature set to be analyzed is used as input and substituted into a preset statistical analysis model. The preset statistical analysis model calculates the data feature set and identifies a set of abnormal points in the data feature set to be analyzed; Based on the set of abnormal points, a preset statistical analysis model is used to analyze the data trend change information of the abnormal points, and the data trend change information of the abnormal points includes the frequency of occurrence and the change rate of the abnormal points; The abnormal point set and its data trend change information are integrated to obtain the potential abnormal pattern.

[0099] According to the needs of anomaly detection and the characteristics of data features, select a suitable statistical analysis model. The statistical analysis model can be a statistical-based algorithm, such as mean-variance analysis, box plot method, Z-score method, etc., or a machine learning algorithm, such as support vector machine, random forest, etc. (Note: If a machine learning algorithm is used, the model must be trained in advance).

[0100] The data feature set to be analyzed is used as input and substituted into the preset statistical analysis model. The statistical analysis model calculates the data feature set and identifies anomalies in the data based on the model's algorithm and logic. Anomalies refer to data points that are significantly different from the normal data pattern, indicating an abnormal state or failure of the system.

[0101] Outlier set output: The identified outliers are grouped into an outlier set to facilitate subsequent analysis and processing.

[0102] Use the preset statistical analysis model to perform trend analysis on outliers. Analyze the frequency of outliers, that is, the number of times or proportion of outliers in the data set. Analyze the rate of change of outliers, that is, the trend of outlier data values ​​changing over time or other variables.

[0103] Extract the data trend change information of abnormal points from trend analysis. The data trend change information should include the frequency of occurrence, change rate and possible change mode of abnormal points.

[0104] Integrate the abnormal point set and its data trend change information. The integrated information should be able to fully reflect the characteristics and trends of the abnormal points.

[0105] Based on the integrated information, potential abnormal patterns are identified. Potential abnormal patterns may be the precursor or cause of system failure and require further analysis and processing. Potential abnormal patterns are presented to operators or decision makers in a visual way.

[0106] Specifically, the method for detecting anomalies in a raw material conveying system for magnesium-based alloy production according to the present invention, step S104, comprises: According to the potential abnormal pattern, historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal pattern is obtained from a database of the magnesium-based alloy production raw material conveying system; Merging the potential abnormal mode and the historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal mode to obtain a preprocessed training data set, a test set, and a validation set; Initialize the weights and bias parameters of the neural network model, use the training data set as input, and substitute it into the neural network model for training; Through the iterative gradient descent method, the parameters of the neural network model are adjusted to make the neural network model fit the abnormal pattern in the training data. After the training is completed, the neural network model is evaluated using the test set to obtain the performance index of the neural network model. According to the performance index of the neural network model, the neural network model with the best performance is selected as the anomaly detection model; The selected anomaly detection model is deployed in the raw material delivery system for magnesium-based alloy production to perform anomaly prediction and abnormal data classification on the real-time data of the raw material delivery system for magnesium-based alloy production.

[0107] According to the potential abnormal patterns identified in step S103, historical operation data corresponding to these patterns are extracted from the database of the raw material delivery system for magnesium-based alloy production.

[0108] Clean the extracted historical operation data to remove missing values, outliers or noise data.

[0109] Normalize or standardize the data to ensure that the data is on the same scale to facilitate subsequent model training.

[0110] The potential abnormal patterns are merged with the corresponding historical operation data to form a comprehensive data set containing abnormal patterns and normal patterns.

[0111] Dataset division: The comprehensive dataset is randomly divided into training dataset, test set and validation set, so that the data distribution among the datasets is uniform and contains enough abnormal mode and normal mode samples.

[0112] Initialize the neural network model: Model selection: According to the requirements of the anomaly detection task, choose a suitable neural network model architecture, such as multi-layer perceptron (MLP), convolutional neural network (CNN) or recurrent neural network (RNN).

[0113] Parameter initialization: Initialize the weights and bias parameters of the neural network model, usually using random initialization or pre-training initialization methods.

[0114] The training data set is used as input and substituted into the neural network model for training. The parameters of the neural network model are optimized using iterative gradient descent methods (such as SGD, Adam, etc.). Appropriate loss functions, such as cross entropy loss and mean square error loss, are set to measure the degree of fit of the model to the training data. The gradient is calculated through the back propagation algorithm, and the parameters of the model are updated so that the model gradually fits the abnormal patterns in the training data.

[0115] Monitor indicators such as loss function value and accuracy during training to evaluate the training effect of the model.

[0116] Use the validation set to validate the model to prevent overfitting or underfitting of the model.

[0117] Model evaluation: Use the test set to evaluate the trained neural network model and obtain the model's performance indicators, such as accuracy, recall, F1 score, etc.

[0118] Model selection: Based on the model's performance indicators, select the best performing neural network model as the anomaly detection model. If multiple models perform similarly, further factors such as model complexity and training time can be considered for selection.

[0119] Model deployment: The selected anomaly detection model is deployed into the raw material delivery system for magnesium-based alloy production and connected to the real-time data acquisition module of the system.

[0120] Anomaly prediction and classification: Use the deployed anomaly detection model to predict anomalies and classify abnormal data in real-time data of the raw material delivery system for magnesium-based alloy production. When the model detects abnormal data, it will promptly issue an alarm or notify the operator to handle it.

[0121] Specifically, the abnormality detection method of the raw material conveying system for magnesium-based alloy production according to the present invention, step S105 comprises: The real-time data of the raw material conveying system for magnesium-based alloy production includes real-time operating parameters of the raw material conveying equipment for magnesium-based alloy production, real-time status parameters of raw materials for magnesium-based alloy production, real-time environmental parameters of raw material conveying for magnesium-based alloy production, real-time status data of the raw material conveying system for magnesium-based alloy production, and real-time error data of the raw material conveying system for magnesium-based alloy production. The real-time data of the raw material conveying system for magnesium-based alloy production is identified and abnormal data is marked using an anomaly detection model. The marked abnormal data is summarized to form an abnormal data report, which includes information on the type of abnormality, the time when the abnormality occurred, and the scope of the impact of the abnormality.

[0122] Source of real-time data: The real-time data of the magnesium-based alloy production raw material conveying system includes but is not limited to: real-time magnesium-based alloy production raw material conveying equipment operating parameters (such as speed, temperature, pressure, etc.), real-time magnesium-based alloy production raw material status parameters (such as composition, particle size, humidity, etc.), real-time magnesium-based alloy production raw material conveying environmental parameters (such as temperature, humidity, air pressure, etc.), real-time magnesium-based alloy production raw material conveying system status data (such as equipment operating status, system load, etc.) and real-time magnesium-based alloy production raw material conveying system error data (such as fault codes, alarm information, etc.).

[0123] The collected real-time data is preprocessed, including data cleaning, denoising, normalization or standardization, etc. The processed real-time data is organized according to a predetermined format and structure to facilitate subsequent anomaly detection model processing.

[0124] The anomaly detection model trained and selected in step S104 is used to perform real-time recognition on real-time data.

[0125] The real-time data is used as input and substituted into the anomaly detection model. The model makes judgments on the real-time data based on the learned anomaly patterns.

[0126] When the model identifies abnormal data in real-time data, the abnormal data is immediately marked, including information such as the type of abnormality and the degree of abnormality.

[0127] Summarize the marked abnormal data, including the source of the abnormal data, the abnormal type, the time when the abnormality occurred, etc. Classify and organize the summarized abnormal data to facilitate subsequent analysis and report generation.

[0128] Generate an abnormal data report based on the summarized abnormal data. The abnormal data report should include the abnormal type, abnormal occurrence time, abnormal impact range information (such as affected production lines, equipment or raw materials, etc.), recommended treatment measures, etc.

[0129] The generated abnormal data report will be promptly released to relevant personnel through the system interface, email, text message, etc.

[0130] Specifically, the abnormality detection method of the raw material conveying system for magnesium-based alloy production according to the present invention, step S105 comprises: According to the time information of the first data to be compared, the actual data of the magnesium-based alloy production raw material delivery system at the time point corresponding to the time information of the first data to be compared is extracted from the actual data collection record of the magnesium-based alloy production raw material delivery system, and the time information of the actual data of the magnesium-based alloy production raw material delivery system is recorded as the second time information of the data to be compared; Aligning the time information of the first data to be compared with the time information of the second data to be compared, so that the time information of the first data to be compared corresponds to the time information of the second data to be compared; If there is a time mismatch, the mismatching time point is recorded; For each aligned time point, comparing the time information of the first data to be compared with the time information of the second data to be compared; Calculating the difference between the time information of the first data to be compared and the time information of the second data to be compared to obtain preliminary error data; The preliminary error data are processed, and the preliminary error data are processed including removing abnormal values, calculating the average value or the standard deviation, and obtaining the raw material transportation error data for magnesium-based alloy production.

[0131] Determine the time information of the first data to be compared: clarify the time range or specific time point of the data to be compared as the time information of the first data to be compared.

[0132] Extracting actual data: According to the time information of the first data to be compared, the actual data of the magnesium-based alloy production raw material delivery system at the corresponding time point is extracted from the actual data collection record of the magnesium-based alloy production raw material delivery system. The extracted actual data and its corresponding time information are recorded as the second data to be compared time information.

[0133] Time alignment operation: Align the time information of the first data to be compared with the time information of the second data to be compared to ensure that the two are consistent in time. If there is a difference in time format or time zone, a unified conversion is required to ensure the accuracy of time comparison.

[0134] Record mismatched time points: If time mismatches are found during the time alignment process (for example, a time point is missing in the actual data or is not included in the data to be compared), record these mismatched time points for subsequent analysis.

[0135] Point-by-point comparison: for each aligned time point, the data value of the time information of the first data to be compared is compared with the data value of the time information of the second data to be compared.

[0136] Calculate the difference: Calculate the difference between the time information of the first data to be compared and the time information of the second data to be compared by mathematical operations (such as subtraction, percentage difference, etc.) to obtain preliminary error data.

[0137] Remove outliers: Perform outlier detection on preliminary error data, such as using statistical methods (such as the 3σ rule) or machine learning algorithms to identify and remove outliers. Outliers may be caused by equipment failure, data collection errors, etc. Removing outliers can improve the accuracy and reliability of error data.

[0138] Calculate statistical indicators: Calculate statistical indicators such as mean value and standard deviation of the processed error data to quantify the size and fluctuation of the error. These statistical indicators can be used to evaluate the performance stability of the raw material delivery system for magnesium-based alloy production and provide a basis for subsequent abnormality detection and analysis.

[0139] After the above processing steps, the final magnesium-based alloy production raw material conveying error data is obtained, which is used for subsequent abnormality judgment, fault diagnosis or system optimization.

[0140] Specifically, the abnormality detection method of the raw material conveying system for magnesium-based alloy production according to the present invention, step S105 comprises: The obtained magnesium-based alloy production raw material transportation error data is recorded as an error data set; Receive and set optimization goals of genetic algorithms; Initialize the population: According to the parameter space of the anomaly detection model, initialize the population of the genetic algorithm. Each individual in the population represents a set of model parameters. Use the error data set to calculate the fitness of each individual in the population; Selection operation: According to the fitness, the selection operation is performed to select individuals with good performance as parents; Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals; Mutation operation, performing mutation operation on offspring individuals; Update the population, use the new individuals after crossover and mutation operations to update the population and form a new population; Iterative optimization, repeatedly executing the steps of calculating fitness, selection, crossover, mutation and updating population until the preset number of iterations is reached or the optimization target reaches the preset target, and the optimized population is obtained; The individual with the highest fitness is extracted from the optimized population. The parameters of the individual with the highest fitness are the optimized anomaly detection model parameters. The optimized anomaly detection model parameters are used to construct the optimized anomaly detection model.

[0141] Record error data set: The previously calculated raw material transportation error data for magnesium-based alloy production is organized into an error data set as the basic data for genetic algorithm optimization.

[0142] Receive optimization goals: According to the actual needs of anomaly detection, receive and clarify the optimization goals of the genetic algorithm, such as minimizing error, maximizing accuracy, etc.

[0143] Set optimization target parameters: According to the optimization goal, set specific objective functions or evaluation indicators to measure the quality of model parameters.

[0144] Determine parameter space: Analyze the parameter space of the anomaly detection model and determine the possible value range of each parameter.

[0145] Initialize the population: According to the parameter space, randomly generate a set of initial model parameters as the population of the genetic algorithm. Each individual in the population represents a set of model parameters.

[0146] Calculate fitness using the error data set: Substitute each individual in the population (i.e., a set of model parameters) into the anomaly detection model, and use the error data set to calculate the fitness of the model, such as by calculating indicators such as model prediction error and accuracy.

[0147] Execute selection operation: According to the fitness, perform selection operation and select individuals with good performance as parents. Common selection methods include roulette selection, tournament selection, etc.

[0148] Perform crossover operation: Perform crossover operation on the selected parent individuals to generate new offspring individuals by exchanging some gene fragments of the parent individuals. The crossover operation helps to combine the good characteristics of the parent individuals.

[0149] Perform mutation operations: Perform mutation operations on offspring individuals to increase the diversity of the population by randomly changing some genes of offspring individuals. Mutation operations help to jump out of local optimal solutions and explore a wider solution space.

[0150] Update the population: Update the population with new individuals after crossover and mutation operations to form a new population. The new population will serve as the starting point for the next round of iteration.

[0151] Repeat the iteration steps: Repeat the steps of calculating fitness, selection, crossover, mutation, and updating the population until the preset number of iterations is reached or the optimization goal reaches the preset goal.

[0152] Monitor the optimization process: During the iteration process, monitor the fitness changes of the population and record the fitness and parameters of the optimal individuals for subsequent analysis.

[0153] Extract the best individual: Extract the individual with the highest fitness from the optimized population. The parameters of this individual are the optimized anomaly detection model parameters.

[0154] Using the optimized anomaly detection model parameters, an optimized anomaly detection model is constructed. This model will have better performance and accuracy and is suitable for anomaly detection in raw material delivery systems for magnesium-based alloy production.

[0155] The model and construction steps involved in the technical solution of the present invention are as follows: Statistical analysis model: The statistical analysis model is used to identify anomalies in the data feature library and analyze the data trend change information of the anomalies.

[0156] Steps for building a statistical analysis model: Select an appropriate statistical analysis model based on the characteristics of the data and analysis requirements. Commonly used models include distance-based models (such as DBSCAN), density-based models (such as IsolationForest), time series analysis models (such as ARIMA), etc.

[0157] Set parameters for the selected statistical analysis model, such as distance threshold, number of clusters, definition of outliers, etc., to ensure that the model can effectively identify outliers in the data.

[0158] Obtain the data feature sets to be analyzed from the data feature library, and substitute these feature sets as input into the statistical analysis model.

[0159] The statistical analysis model calculates and analyzes a set of data features to identify a set of outliers in the data.

[0160] Based on the identified abnormal point set, the statistical analysis model is used to further analyze the data trend change information of the abnormal points, including the frequency of occurrence and change rate of the abnormal points. The abnormal point set and its data trend change information are integrated to form a potential abnormal pattern.

[0161] Neural network model: The neural network model is used to predict anomalies and classify abnormal data in real-time data of the raw material delivery system for magnesium-based alloy production.

[0162] The steps of building a neural network model are as follows: According to the potential abnormal pattern, the corresponding historical operation data is obtained from the database. The potential abnormal pattern and the historical operation data are merged to form a training data set, a test set, and a validation set.

[0163] According to the characteristics and requirements of the anomaly detection task, select a suitable neural network model, such as multi-layer perceptron (MLP), convolutional neural network (CNN), recurrent neural network (RNN) and its variants (such as LSTM, GRU), etc.

[0164] Design the structure and parameters of the neural network model, including the dimension of the input layer, the number of hidden layers and neurons, the dimension of the output layer, the choice of activation function, the definition of the loss function, etc. Initialize the weights and bias parameters of the neural network model.

[0165] The neural network model is trained using the training data set, and the weights and biases of the model are adjusted through the back-propagation algorithm and optimization algorithms (such as gradient descent, Adam, etc.) to minimize the loss function.

[0166] Use the validation set to validate the model during training, evaluate the model's performance (such as accuracy, recall, F1 score, etc.), and tune the model based on the validation results.

[0167] Use the test set to evaluate the trained neural network model and obtain the performance indicators of the model.

[0168] Based on the evaluation results of the test set, the best performing neural network model is selected as the anomaly detection model.

[0169] The present invention provides an anomaly detection method for a raw material conveying system for magnesium-based alloy production, which effectively solves the above-mentioned problems through systematic data collection, preprocessing, feature extraction, model training and optimization steps.

[0170] First, obtain comprehensive basic data of the raw material conveying system for magnesium-based alloy production, including but not limited to conveying equipment operating parameters, raw material status parameters, conveying environment parameters, system status data, error data, and historical operating data (step S101). By comprehensively collecting data, ensure that subsequent analysis can cover all factors that may affect system abnormalities, laying a solid foundation for subsequent feature extraction and anomaly detection.

[0171] The collected basic data is preprocessed, including data cleaning and denoising, to obtain preprocessed basic data (step S102). Feature indicators are extracted from the preprocessed data to build a data feature library. The feature library contains equipment operation features, raw material status features, transportation environment features, system status features, and error data features (step S102).

[0172] Through feature extraction, the raw data is converted into more meaningful feature representations to facilitate subsequent model recognition and analysis.

[0173] The data features in the data feature library are identified using a preset statistical analysis model to find out the abnormal points in the data and the data trend change information of the abnormal points (step S103).

[0174] The trend change information of abnormal point data includes the frequency of occurrence and change rate of abnormal points, which helps to understand the occurrence patterns and potential causes of abnormalities.

[0175] The abnormal points and their data trend change information are integrated into potential abnormal patterns to provide input for subsequent model training.

[0176] According to the potential abnormal pattern and the historical operation data of the system, the neural network model is trained to obtain an abnormal detection model (step S104). The abnormal detection model is used to predict abnormalities and classify abnormal data of the system real-time data, and can timely discover and handle abnormal situations in the system.

[0177] Collect real-time data from the system and substitute it as input into the anomaly detection model to obtain real-time data anomaly prediction results and abnormal data classification (step S105). Compare the prediction results with the actual data to obtain error data. Based on the error data, use a genetic algorithm to optimize the anomaly detection model to improve the accuracy and robustness of the model (step S105). Deploy the optimized anomaly detection model to the raw material delivery system for magnesium-based alloy production to achieve continuous and effective anomaly detection.

[0178] In summary, the technical solution of the present invention effectively solves the problem that when erroneous data appears in the raw material delivery system for magnesium-based alloy production, the existing system is unable to extract useful information from massive data and the existing data analysis method is unable to cope with the variable data characteristics through systematic data collection, preprocessing, feature extraction, model training and optimization steps.

Claims

1. A method for detecting abnormality in a raw material conveying system for magnesium-based alloy production, characterized in that: include: Step S101, obtaining basic data of a raw material delivery system for magnesium-based alloy production, wherein the basic data of the raw material delivery system for magnesium-based alloy production includes operating parameters of raw material delivery equipment for magnesium-based alloy production, state parameters of raw materials for magnesium-based alloy production, environmental parameters of raw material delivery for magnesium-based alloy production, state data of the raw material delivery system for magnesium-based alloy production, error data of the raw material delivery system for magnesium-based alloy production, and historical operating data of the raw material delivery system for magnesium-based alloy production; Step S102: preprocessing the basic data of the magnesium-based alloy production raw material transportation system to obtain the preprocessed basic data of the magnesium-based alloy production raw material transportation system, extracting characteristic indicators according to the preprocessed basic data of the magnesium-based alloy production raw material transportation system, and constructing a data characteristic library; Step S103: using a preset statistical analysis model to identify data features in the data feature library, obtaining abnormal points in the data and abnormal point data trend change information, and using the abnormal points in the data and abnormal point data trend change information as potential abnormal patterns; Step S104: training a neural network model according to potential abnormal patterns and historical operation data of the raw material delivery system for magnesium-based alloy production to obtain an abnormality detection model, which is used to perform abnormality prediction and abnormal data classification on real-time data of the raw material delivery system for magnesium-based alloy production; Step S105: collecting real-time data of the raw material transportation system for magnesium-based alloy production, and substituting the real-time data of the raw material transportation system for magnesium-based alloy production as input into the anomaly detection model, the anomaly detection model outputs the anomaly prediction result of the real-time data of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, retrieving the time information of the real-time data anomaly prediction result of the raw material transportation system for magnesium-based alloy production and the classification of the real-time anomaly data of the raw material transportation for magnesium-based alloy production, and obtaining the first time information of the data to be compared, collecting the actual data of the raw material transportation for magnesium-based alloy production according to the time information of the data to be compared, and obtaining the second time information of the data to be compared, comparing the first time information of the data to be compared with the second time information of the data to be compared, and obtaining the error data of the raw material transportation for magnesium-based alloy production, optimizing the anomaly detection model using a genetic algorithm based on the error data of the raw material transportation for magnesium-based alloy production, and obtaining the optimized anomaly detection model, and deploying the optimized anomaly detection model to the raw material transportation system for magnesium-based alloy production.

2. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 1, characterized in that: In step S102, the data feature library includes operation characteristics of magnesium-based alloy production raw material conveying equipment, state characteristics of magnesium-based alloy production raw materials, conveying environment characteristics of magnesium-based alloy production raw materials, state characteristics of magnesium-based alloy production raw material conveying system, and error data characteristics of magnesium-based alloy production raw material conveying system.

3. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 1, characterized in that: The step S103 includes: According to the data feature library, obtain the data feature set to be analyzed and load the preset statistical analysis model; The data feature set to be analyzed is used as input and substituted into a preset statistical analysis model. The preset statistical analysis model calculates the data feature set and identifies a set of abnormal points in the data feature set to be analyzed; Based on the set of abnormal points, a preset statistical analysis model is used to analyze the data trend change information of the abnormal points, and the data trend change information of the abnormal points includes the frequency of occurrence and the change rate of the abnormal points; The abnormal point set and its data trend change information are integrated to obtain the potential abnormal pattern.

4. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 1, characterized in that: The step S104 includes: According to the potential abnormal pattern, historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal pattern is obtained from a database of the magnesium-based alloy production raw material conveying system; Merging the potential abnormal mode and the historical operation data of the magnesium-based alloy production raw material conveying system corresponding to the potential abnormal mode to obtain a preprocessed training data set, a test set, and a validation set; Initialize the weights and bias parameters of the neural network model, use the training data set as input, and substitute it into the neural network model for training; Through the iterative gradient descent method, the parameters of the neural network model are adjusted to make the neural network model fit the abnormal pattern in the training data. After the training is completed, the neural network model is evaluated using the test set to obtain the performance index of the neural network model. According to the performance index of the neural network model, the neural network model with the best performance is selected as the anomaly detection model; The selected anomaly detection model is deployed in the raw material delivery system for magnesium-based alloy production to perform anomaly prediction and abnormal data classification on the real-time data of the raw material delivery system for magnesium-based alloy production.

5. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 1, characterized in that: The step S105 includes: The real-time data of the raw material conveying system for magnesium-based alloy production includes real-time operating parameters of the raw material conveying equipment for magnesium-based alloy production, real-time status parameters of raw materials for magnesium-based alloy production, real-time environmental parameters of raw material conveying for magnesium-based alloy production, real-time status data of the raw material conveying system for magnesium-based alloy production, and real-time error data of the raw material conveying system for magnesium-based alloy production. The real-time data of the raw material conveying system for magnesium-based alloy production is identified and abnormal data is marked using an anomaly detection model. The marked abnormal data is summarized to form an abnormal data report, which includes information on the type of abnormality, the time when the abnormality occurred, and the scope of the impact of the abnormality.

6. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 5, characterized in that: The step S105 includes: According to the time information of the first data to be compared, the actual data of the magnesium-based alloy production raw material delivery system at the time point corresponding to the time information of the first data to be compared is extracted from the actual data collection record of the magnesium-based alloy production raw material delivery system, and the time information of the actual data of the magnesium-based alloy production raw material delivery system is recorded as the second time information of the data to be compared; Aligning the time information of the first data to be compared with the time information of the second data to be compared, so that the time information of the first data to be compared corresponds to the time information of the second data to be compared; If there is a time mismatch, the mismatching time point is recorded; For each aligned time point, comparing the time information of the first data to be compared with the time information of the second data to be compared; Calculating the difference between the time information of the first data to be compared and the time information of the second data to be compared to obtain preliminary error data; The preliminary error data are processed, and the preliminary error data are processed including removing abnormal values, calculating the average value or the standard deviation, and obtaining the raw material transportation error data for magnesium-based alloy production.

7. The abnormality detection method for the raw material conveying system for magnesium-based alloy production according to claim 6, characterized in that: The step S105 includes: The obtained magnesium-based alloy production raw material transportation error data is recorded as an error data set; Receive and set optimization goals of genetic algorithms; Initialize the population: According to the parameter space of the anomaly detection model, initialize the population of the genetic algorithm. Each individual in the population represents a set of model parameters. Use the error data set to calculate the fitness of each individual in the population; Selection operation: According to the fitness, the selection operation is performed to select individuals with good performance as parents; Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals; Mutation operation, performing mutation operation on offspring individuals; Update the population, use the new individuals after crossover and mutation operations to update the population and form a new population; Iterative optimization, repeatedly executing the steps of calculating fitness, selection, crossover, mutation and updating population until the preset number of iterations is reached or the optimization target reaches the preset target, and the optimized population is obtained; The individual with the highest fitness is extracted from the optimized population. The parameters of the individual with the highest fitness are the optimized anomaly detection model parameters. The optimized anomaly detection model parameters are used to construct the optimized anomaly detection model.

Images