Cylinder body fault early warning system and method based on mechanism and data dual drive

By adopting a feature screening method based on mechanism and data in the industrial equipment monitoring system, combined with statistics and machine learning, the problems of poor interpretability of fault warning and limitations of feature extraction in the existing technology are solved, and more accurate and efficient Fault prediction and maintenance improve equipment monitoring and maintenance efficiency.

CN118915654APending Publication Date: 2024-11-08CENTRAL UNIVERSITY OF FINANCE AND ECONOMICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410961149.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing industrial equipment monitoring system has poor interpretability and feature extraction limitations in fault warning, and cannot effectively capture the equipment operating status and cause of failure, resulting in low maintenance efficiency and high false alarm rate.

Method used

Using a feature screening method based on mechanism and data dual drive, combined with statistical methods and machine learning, sensor data is analyzed through a central server, equipment failures are identified and intuitive feature trend change diagrams are provided, improving the interpretability and accuracy of fault warnings. .

Benefits of technology

It improves the understanding of the operating status of the equipment and the failure prediction ability, reduces the false alarm rate and missed alarm rate, improves the monitoring and maintenance efficiency of industrial equipment, and reduces labor costs and production downtime losses.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915654A_ABST
    Figure CN118915654A_ABST
Patent Text Reader

Abstract

According to the mechanism and data dual-drive-based cylinder body fault early warning system and method provided by the invention, the system comprises an architecture module composed of a production workshop, a central server and a mobile terminal user, and data acquisition, storage, analysis and early warning functions are realized from a sensor assembly of the production workshop to the mobile terminal user; the central server is responsible for receiving and storing data from the sensor assembly in the production workshop, and the stored data is used for subsequent fault detection and analysis and feeding back the working condition of the workshop. According to the mechanism and data dual-drive-based cylinder body fault early warning system and method provided by the invention, a novel feature screening method combining a statistical method and a machine learning method is adopted, so that large-scale data can be effectively processed, understanding of machine operation can be improved, and hidden complex relationships in the data can be mined; and the interpretability of the operation state and fault early warning of the equipment is provided, so that the monitoring and maintenance efficiency of the industrial equipment is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial equipment fault warning, and in particular relates to a cylinder fault warning system and method based on dual drive of mechanism and data. Background Art

[0002] Over the past few decades, with the development of industrial automation and informatization, the monitoring and maintenance systems of industrial equipment have made significant progress. Early monitoring systems mainly relied on manual inspections and basic instrument readings. With the development of technology, automatic monitoring systems based on computer and communication technologies have emerged. These systems usually include multiple parts such as data acquisition, data transmission, data processing and data display, which can provide certain monitoring functions for the operating status of the equipment. Even these advanced monitoring systems still have some problems and limitations. For example, many systems have limited data analysis capabilities and can only provide basic status information, but cannot perform in-depth data analysis and fault prediction; the early warning functions of many systems are not intelligent enough, which may result in high false alarm rates or missed alarm rates; due to the lack of effective system integration solutions, the sharing and integration of data and information between different systems is a challenge.

[0003] In recent years, the development of new technologies such as the Internet of Things (IoT), big data, and artificial intelligence has made it possible to solve these problems. These systems usually contain a group of IoT devices consisting of sensors, controllers, and other devices, which are connected to a central server through a network. The server is responsible for data processing and analysis, and users can access the server through a computer or mobile device to view device status information and warnings. However, these systems usually only provide basic monitoring functions, such as equipment status display and basic fault warnings, but the data obtained by workshop sensors is too complex, and simple data prediction cannot be well applied to such high-frequency data. In addition, the interpretability is poor, and the cause of the fault cannot be quickly checked. Manual feedback is required for cause screening, making it difficult for maintenance personnel to make maintenance or repair decisions quickly and accurately.

[0004] In summary, the existing industrial equipment monitoring and maintenance systems have shortcomings that seriously affect the performance and reliability of the system. Specifically, there are the following points:

[0005] (1) Poor interpretability of the early warning function: Most existing fault warning technologies can only detect faults and then issue an alarm. After receiving the alarm, the staff needs to manually check the parameters and data of the workshop machines one by one. This process not only wastes manpower, but also leads to excessive energy consumption in the production workshop and reduced production efficiency.

[0006] (2) Limitations of feature extraction: Most existing systems use a single feature screening method, which is unable to fully and accurately capture and analyze the operating status of equipment. For example, although the feature screening method based on statistics is efficient in processing large amounts of data, it may not be able to capture the nonlinear relationships hidden in the data. Although the feature screening method based on machine learning can discover complex relationships in the data, when faced with high-dimensional and large-scale data, the computational complexity may be too high, affecting the analysis speed. Therefore, this limits the in-depth understanding of the equipment status. The lack of an effective feature extraction method means that even if a large amount of data is collected, it cannot be converted into useful information, and thus cannot provide accurate input for the early warning system. In addition, only from the perspective of data features, the actual operation of the machine is not taken into account. Different machines have certain differences, and the emphasis of features is also different.

[0007] In view of the above problems, the present invention proposes a cylinder fault warning system and method based on dual drive of mechanism and data, so as to solve the technical problems of limitations in data processing and insufficient explainability of faults in the prior art. Summary of the invention

[0008] The purpose of the present invention is to provide a cylinder fault warning system and method based on dual drive of mechanism and data, which solves the technical problems of limitations in data processing and insufficient explainability of faults in the prior art. The system adopts a new feature screening method that combines statistical methods and machine learning methods, which can effectively process large-scale data, improve the understanding of machine operation, and mine the complex relationships hidden in the data. In addition, it provides the explainability of equipment operating status and fault warnings, thereby improving the monitoring and maintenance efficiency of industrial equipment.

[0009] A cylinder fault early warning system based on dual drive of mechanism and data, including an architecture module consisting of a production workshop, a central server and a mobile user, realizing data collection, storage, analysis and early warning functions from the sensor components in the production workshop to the mobile user;

[0010] The central server is responsible for receiving and storing data from sensor components in the production workshop. The stored data is used for subsequent fault detection and analysis, as well as feedback on workshop working conditions.

[0011] The production workshop is installed with any one of a temperature sensor, a humidity sensor, and a pressure sensor, or a combination of at least two or more of them, for sensing the environmental data of the workshop and transmitting the real-time collected data to a central server via a network.

[0012] A fault detection and analysis module is deployed at the end of the central server, and the fault detection and analysis module identifies whether there is a device fault by analyzing the stored data;

[0013] When the fault detection and analysis module detects a device fault, it triggers a fault warning sending module for mobile users, and the fault warning sending module sends fault warning information to the mobile user;

[0014] At the same time, the characteristic trend change graph sending module on the central server screens out the characteristic trend change graph and sends it to the mobile user. The characteristic trend change graph intuitively displays the abnormal trend of the data, helping the operator to better understand the equipment status and make corresponding adjustments.

[0015] The mobile user uses a mobile application to communicate with the central server through the mobile device. The mobile application receives the fault warning information sent by the central server and displays the fault warning information to the operator of the mobile user in real time. The operator views the fault details through the mobile application and takes corresponding measures in time.

[0016] Through such an architecture, we have achieved real-time monitoring of production workshop environmental data, timely detection and early warning of faults, and intuitive data visualization through mobile terminals, allowing operators to respond to and handle equipment failures quickly and effectively.

[0017] The server includes a data layer, a processing layer and an application layer;

[0018] The data layer includes a data acquisition and cleaning component and a database module. The data acquisition and cleaning component is responsible for collecting raw data from devices or sensors, and performing necessary cleaning and preprocessing to ensure data quality and availability. The database module is where the cleaned and processed data is stored, providing support for the model analysis of the processing layer and retaining historical data for future analysis and learning.

[0019] The processing layer includes a feature recognition module, a fault detection module, and a feature visualization module. The feature recognition module recognizes the features of the cleaned data based on the dual drive of mechanism and data;

[0020] Based on the mechanism drive, the cleaned data is identified, and according to the expert opinions and practical experience, the features that are strongly related to the machine operation are defined to obtain the feature set X m ,The engine cylinder block casting is an extremely complex process. From the perspective of industrial manufacturing principles, the cylinder block defects involve multiple factors, and need to comprehensively consider five factors, namely sand core preparation, mold design, gating system design, gating temperature control and metal composition uniformity, which are respectively denoted as A1…A5;

[0021] Based on data-driven, the features of the cleaned data are identified. From the perspective of data, the feature set X1 is obtained by using the lasso technology. Y is whether the cylinder body is faulty, X is all the features obtained from the sensors in the production workshop, β is the regression coefficient of the feature and whether it is faulty, λ is the penalty parameter, ||·||1 represents the L1 norm, that is, the sum of the absolute values ​​of the coefficients. The size of λ determines the sparsity of the regression coefficient. When λ is larger, the regression coefficient is sparser, that is, more features are excluded. The L1 regularization term makes part of the parameters in the parameter vector of the model become zero, thereby realizing the function of feature selection.

[0022] min imize||Y-Xβ|| 2 +λ||β||1 (1)

[0023] Formula (1) is the calculation formula for feature selection, which is a constraint condition.

[0024] Finally, the features obtained in the above two steps are combined to obtain the feature set X that is most conducive to feedback on the machine operation status and accurate fault prediction. f ;

[0025]

[0026] X1=x i ∈X|Lasso(X,Y,λ) i≠0 (3)

[0027] X f =X m ∪X1 (4)

[0028] Among them, X m It is a feature set obtained based on the mechanism, Ai refers to five mechanism levels respectively, and F is a function summary that represents the process of feature screening;

[0029] X1 is the feature set obtained by lasso, where lasso is the abbreviation of the function, Y is the fault index, X is all feature sets, and λ is the penalty parameter of the lasso method;

[0030] X f is the final feature set, taking X1 and X m The union of .

[0031] The fault detection module targets the fusion feature dataset X f , using deep learning models for feature learning and predicting failures in production in the next time period;

[0032] The feature visualization module is used to fusion feature dataset X fConduct data statistics and visualization, and generate relevant reports to help staff gain an intuitive understanding of the machine's operating status.

[0033] The application layer includes a feature push module and a fault warning module;

[0034] The feature push module visualizes and summarizes the working status of the machine and sends it to the mobile terminal user, helping the staff to intuitively view the changing trends and abnormal points of the workshop data;

[0035] When the detection module at the processing layer finds a fault, the fault warning module triggers a warning and sends a warning signal to the mobile terminal user, prompting the staff or maintenance team to take measures in advance.

[0036] The entire server architecture design reflects the complete process of data from collection and cleaning to analysis and judgment to early warning feedback. Each level and component plays a specific role and works together to achieve the goals of the system.

[0037] The production workshop includes a data layer and an application layer;

[0038] The data layer includes a workshop sensor module, a data collection module and a data transmission module; the workshop sensor is responsible for real-time monitoring of the operating status of the equipment in the workshop and sending the collected data to the data collection component; the sensor may include a thermometer, a pressure sensor, a vibration sensor, etc. The data collection module is responsible for receiving the raw data from the sensor and performing preprocessing; the data transmission module sends the collected data to the central server for further analysis and processing to ensure timely and accurate data transmission;

[0039] The application layer includes a feature display module and a fault warning module. The feature display module is responsible for displaying the data status and change trends of machine features to users. These data include key parameters such as temperature, vibration, and pressure. The fault warning module receives analysis results from the data layer. If a production failure occurs, the module sends a fault warning signal to the staff to remind them to respond to potential problems in a timely manner.

[0040] The entire workshop-side architecture design reflects the complete process of data from original collection to display and early warning. Various components work together to ensure that the status of workshop equipment can be monitored and analyzed in real time, thereby achieving early detection and early warning of faults.

[0041] The mobile user includes a data layer and an application layer;

[0042] The data layer includes a data receiving module and a data sending module; the data receiving module is responsible for receiving data from the central server, including real-time feature data and fault warning information of the equipment. Data reception ensures that users can obtain the latest equipment status information on the mobile terminal; the data sending module is used to send the user's operations and feedback in the mobile terminal application to the central server, including the user's response to the warning, maintenance request or other key information upload;

[0043] The application layer includes a feature display module and a fault warning module; the feature display module will display the received feature data information and visualization diagrams to the staff for viewing, thereby assisting the staff to quickly understand the production situation in the workshop; the fault warning module is used to provide users with real-time fault or abnormality warnings. When the server analyzes potential problems and sends warning information, the module will receive the information and notify the user through the mobile application.

[0044] The entire mobile architecture is designed to ensure that users can receive important information about equipment status and potential failures in a timely manner through their mobile devices, and can perform necessary interactions, such as confirming warning information or sending maintenance requests. This architecture supports remote monitoring and equipment management functions, allowing users to effectively participate in equipment maintenance and fault handling even when they are not on site in the workshop.

[0045] The innovation points of the present invention focus on the following two points:

[0046] (1) Dual-driven feature extraction method: The present invention not only simply applies traditional data analysis technology, but also innovatively combines expert opinions and practical experience with data analysis to form a feature extraction method based on dual drive of mechanism and data. Through this method, we can more comprehensively capture the key features of the equipment operation status, thereby improving the understanding and prediction capabilities of the equipment status. Compared with the traditional single feature extraction method, the dual-driven feature extraction method of the present invention not only improves the accuracy, but also enhances the interpretability of the features, providing more accurate and reliable input for the early warning system.

[0047] (2) Fault prediction and explainability: The present invention not only achieves timely prediction of industrial equipment failures, but also adds visual displays such as characteristic trend change diagrams with high explainability to the early warning system. These visual displays can not only help operators quickly understand the equipment status and abnormal conditions, but also provide more intuitive and specific information, allowing operators to more accurately judge and respond to different types of failures. Compared with traditional fault prediction systems, the early warning system of the present invention is more intelligent and user-friendly, which can help enterprises to timely discover and solve potential equipment problems, and improve the efficiency and reliability of industrial equipment monitoring and maintenance.

[0048] The positive effects of the present invention are as follows:

[0049] (1) Dual-driven feature extraction method: Traditional methods usually only use a single feature extraction method, such as a statistical method or a machine learning method. The present invention introduces a feature extraction method based on dual-driven mechanism and data, combining expert opinions and practical experience with data analysis, which can more comprehensively capture the characteristics of the equipment operation status and improve the accuracy and interpretability of feature extraction.

[0050] (2) Fault prediction and explainability: This invention introduces highly explainable visual displays such as feature trend change graphs, which enable operators to intuitively understand the equipment status and abnormal conditions, helping them to respond quickly and accurately. Traditional methods usually lack explanations for fault prediction results, making it difficult for operators to understand the meaning of warning information.

[0051] (3) Improve efficiency and reduce costs: The present invention can help enterprises to timely discover and solve equipment failures and reduce downtime losses and maintenance costs by realizing real-time monitoring of equipment status and fault prediction. Compared with traditional manual inspections and basic instrument readings, the present invention can improve monitoring efficiency and reduce labor costs, thus achieving low-carbon production. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is the overall system architecture diagram of the present invention.

[0053] Figure 2 This is a diagram of the architecture of the central server in the present invention.

[0054] Figure 3 This is the architectural diagram of the production workshop in the present invention.

[0055] Figure 4 This is an architectural diagram of a mobile user in the present invention.

[0056] Figure 5 It is a flow chart of the fault detection and early warning function in the present invention.

[0057] Figure 6 This is a flow chart of the feature visualization push function in the present invention.

[0058] Among them, the figure markings are: 1. Central server; 2. Production workshop; 3. Mobile user. DETAILED DESCRIPTION

[0059] In order to more clearly illustrate the technical features of this solution, this solution is described below through specific implementation methods.

[0060] See also Figure 1-Figure 4, a cylinder fault early warning system based on dual-drive of mechanism and data, including an architecture module consisting of a production workshop, a central server and mobile users, from the sensor components in the production workshop to the mobile users, realizing the data collection, storage, analysis and early warning functions;

[0061] The central server is responsible for receiving and storing data from sensor components in the production workshop. The stored data is used for subsequent fault detection and analysis, as well as feedback on workshop working conditions.

[0062] The production workshop is installed with any one of a temperature sensor, a humidity sensor, and a pressure sensor, or a combination of at least two or more of them, for sensing the environmental data of the workshop and transmitting the real-time collected data to a central server via a network.

[0063] A fault detection and analysis module is deployed at the end of the central server, and the fault detection and analysis module identifies whether there is a device fault by analyzing the stored data;

[0064] When the fault detection and analysis module detects a device fault, it triggers a fault warning sending module for mobile users, and the fault warning sending module sends fault warning information to the mobile user;

[0065] At the same time, the characteristic trend change graph sending module on the central server screens out the characteristic trend change graph and sends it to the mobile user. The characteristic trend change graph intuitively displays the abnormal trend of the data, helping the operator to better understand the equipment status and make corresponding adjustments.

[0066] The mobile user uses a mobile application to communicate with the central server through the mobile device. The mobile application receives the fault warning information sent by the central server and displays the fault warning information to the operator of the mobile user in real time. The operator views the fault details through the mobile application and takes corresponding measures in time.

[0067] The server includes a data layer, a processing layer and an application layer;

[0068] The data layer includes a data acquisition and cleaning component and a database module. The data acquisition and cleaning component is responsible for collecting raw data from devices or sensors, and performing necessary cleaning and preprocessing to ensure data quality and availability. The database module is where the cleaned and processed data is stored, providing support for the model analysis of the processing layer and retaining historical data for future analysis and learning.

[0069] The processing layer includes a feature recognition module, a fault detection module, and a feature visualization module. The feature recognition module recognizes the features of the cleaned data based on the dual drive of mechanism and data;

[0070] Based on the mechanism drive, the cleaned data is identified, and according to the expert opinions and practical experience, the features that are strongly related to the machine operation are defined to obtain the feature set X m ,The engine cylinder block casting is an extremely complex process. From the perspective of industrial manufacturing principles, the cylinder block defects involve multiple factors. It is necessary to comprehensively consider the sand core preparation, mold design, gating system design, gating temperature control and metal composition uniformity, which are represented as A1…A5;

[0071] Specifically, they include: (1) Sand core preparation: If the hardness and viscosity of the sand mold and other parameters are not strictly controlled, the liquid metal will not be able to evenly fill the model, resulting in a cold shut. (2) Mold design: Poor internal structure design, such as uneven surfaces, lack of guide structures, etc., will prevent the liquid metal from fully filling the gaps in the mold, resulting in a cold shut. (3) Pouring system: An improperly designed pouring system may cause poor liquid metal flow and easily cause a cold shut. (4) Pouring temperature: Both too low and too high temperatures can easily cause a cold shut. When the temperature is too low, the metal has poor fluidity and cannot fill the voids; when the temperature is too high, the liquid metal is difficult to wet the mold surface. (5) Metal composition: If the metal composition in the casting material is uneven, the liquid metal will not be able to completely fill the small voids, resulting in a cold shut.

[0072] Based on data-driven, the features of the cleaned data are identified. From the perspective of data, the feature set X1 is obtained by using the lasso technology. Y is whether the cylinder body is faulty, X is all the features obtained from the sensors in the production workshop, β is the regression coefficient of the feature and whether it is faulty, λ is the penalty parameter, ||·||1 represents the L1 norm, that is, the sum of the absolute values ​​of the coefficients. The size of λ determines the sparsity of the regression coefficient. When λ is larger, the regression coefficient is sparser, that is, more features are excluded. The L1 regularization term makes part of the parameters in the parameter vector of the model become zero, thereby realizing the function of feature selection.

[0073] min imize||Y-Xβ|| 2 +λ||β||1 (1)

[0074] Finally, the features obtained in the above two steps are combined to obtain the feature set X that is most conducive to feedback on the machine operation status and accurate fault prediction. f ;

[0075]

[0076] X1=x i ∈X|Lasso(X,Y,λ) i≠0 (3)

[0077] X f =X m ∪X1 (4)

[0078] Among them, X m is a feature set obtained based on the mechanism, Ai refers to five mechanism levels, and F is a function summary that represents the process of feature screening; specifically, we start from five aspects, screen out features that meet the mechanism in the original data set X, and take the union to obtain the final X m .

[0079] X1 is the feature set obtained by lasso, where lasso is the abbreviation of the function, Y is the fault index, X is all feature sets, and λ is the penalty parameter of the lasso method; X f is the final feature set, taking X1 and X m The union of;

[0080] The fault detection module targets the fusion feature dataset X f , using deep learning models for feature learning and predicting failures in production in the next time period;

[0081] The feature visualization module is used to fusion feature dataset X f Conduct data statistics and visualization, and generate relevant reports to help staff gain an intuitive understanding of the machine's operating status.

[0082] Lasso (Least Absolute Shrinkage and Selection Operator) is a commonly used statistical method that is widely used in feature selection and variable screening, especially in high-dimensional data analysis. The introduction of the lasso method solves the problems of overfitting and multicollinearity that are prone to occur in traditional linear regression when dealing with high-dimensional data, providing researchers with an effective tool that can mine the most important features while maintaining the simplicity of the model.

[0083] The core idea of ​​the lasso method is to combine the least squares method and the L1 regularization term, and to achieve variable screening and model compression by adding penalty terms to the model parameters. This feature gives lasso a unique advantage when processing data sets with a large number of features. Its objective function is as follows:

[0084]

[0085] Where N is the number of samples, p is the number of features, and y i is the dependent variable, x ij is the independent variable matrix, β jis the model parameter, and λ is the regularization parameter. The key to lasso is to control the sparsity of the model by adjusting the regularization parameter, λ, thereby achieving automatic feature selection.

[0086] The application layer includes a feature push module and a fault warning module;

[0087] The feature push module visualizes and summarizes the working status of the machine and sends it to the mobile terminal user, helping the staff to intuitively view the changing trends and abnormal points of the workshop data;

[0088] When the detection module at the processing layer finds a fault, the fault warning module triggers a warning and sends a warning signal to the mobile terminal user, prompting the staff or maintenance team to take measures in advance.

[0089] The production workshop includes a data layer and an application layer;

[0090] The data layer includes a workshop sensor module, a data collection module and a data transmission module; the workshop sensor is responsible for real-time monitoring of the operating status of the equipment in the workshop and sending the collected data to the data collection component; the data collection module is responsible for receiving the raw data from the sensor and performing preprocessing; the data transmission module sends the collected data to the central server for further analysis and processing to ensure timely and accurate data transmission;

[0091] The application layer includes a feature display module and a fault warning module. The feature display module is responsible for displaying the data status and change trends of machine features to users. These data include key parameters such as temperature, vibration, and pressure. The fault warning module receives analysis results from the data layer. If a production failure occurs, the module sends a fault warning signal to the staff to remind them to respond to potential problems in a timely manner.

[0092] The mobile user includes a data layer and an application layer;

[0093] The data layer includes a data receiving module and a data sending module; the data receiving module is responsible for receiving data from the central server, including real-time feature data and fault warning information of the equipment. Data reception ensures that users can obtain the latest equipment status information on the mobile terminal; the data sending module is used to send the user's operations and feedback in the mobile terminal application to the central server, including the user's response to the warning, maintenance request or other key information upload;

[0094] The application layer includes a feature display module and a fault warning module; the feature display module will display the received feature data information and visualization diagrams to the staff for viewing, thereby assisting the staff to quickly understand the production situation in the workshop; the fault warning module is used to provide users with real-time fault or abnormality warnings. When the server analyzes potential problems and sends warning information, the module will receive the information and notify the user through the mobile application.

[0095] Example 1

[0096] See also Figure 5 , is a flow chart of fault detection and early warning functions. The specific process is as follows:

[0097] (1) The system in the production workshop automatically collects data sent back by sensors. This data includes the current working status and performance indicators of the equipment. The parameter data of the production equipment is also collected at the same time.

[0098] (2) The collected data needs to be sent to a central processing unit for further analysis and processing.

[0099] (3) The central processing unit of the central server pre-processes the data, including data cleaning, normalization and other operations, to provide accurate input for fault diagnosis.

[0100] (4) Feature identification of the cleaned data. Mechanism features are mainly features that are strongly correlated with machine operation from expert opinions and actual operations. Data features are mainly extracted from the data based on the lasso method. Finally, the features obtained by the two feature identification methods are merged to obtain the feature set that is most conducive to feedback on machine operation status and accurate fault prediction.

[0101] (5) The Imbalance-XGBoost model is then used to learn the filtered and aggregated data. After the model is trained, failure prediction is performed on production at future time points.

[0102] (6) When the model predicts that a failure will occur in the future, the central server sends a fault warning to mobile users (such as mobile phones or tablets).

[0103] (7) After receiving the warning, the operator of the mobile terminal needs to investigate the cause of the fault. If the fault is judged to be an isolated incident, the operator chooses to close the warning, the workshop continues to produce normally, and the sensor continues to collect data for prediction.

[0104] (8) If the fault is judged to be caused by improper workshop environment and parameters, the staff will take measures to deal with it. If the fault is solved, the warning is closed and feedback is given to the workshop. If the fault is not solved, the timer will be used to perform fault detection and warning again.

[0105] Example 2

[0106] See also Figure 6 , which is a flowchart of the feature visualization push function. The specific process is as follows:

[0107] (1) Sensors in the production workshop collect raw data in real time, including physical sensors collecting environmental data and parameter data preset by the workshop production machines.

[0108] (2) The collected data will be sent from the production workshop to the central server for subsequent processing.

[0109] (3) After receiving the raw data from the workshop, the central server will store and clean the data, including formatting and missing value processing.

[0110] (4) The cleaned data is used for feature identification, and dual-driven identification of mechanism and data engineering is performed simultaneously to obtain feature parameters with high correlation with the fault, and the features screened out by the two methods are summarized to obtain the final data set.

[0111] (5) For the feature data set, select the most appropriate visualization method to draw a visualization graph for each feature, and mark the abnormal fluctuation points or intervals in the graph to capture the feature anomalies.

[0112] (6) The server pushes a visualization of the operating status of all equipment to mobile users and production workshops.

[0113] (7) After receiving the equipment operation status diagram, the staff of the mobile terminal user can visually observe the workshop environment and production equipment parameters. If there are obvious and unreasonable fluctuations, they can respond and correct them in time to avoid frequent failures in subsequent production.

[0114] Example 3

[0115] This embodiment 3 is illustrated by a practical case. For example, the manufacture of automobile engine cylinder blocks involves multiple processes, wherein the sensors of each process have corresponding index data uploaded to the server for processing. The statistical description of some indicators is listed below to understand the practical application scenarios of the present invention.

[0116] This application takes the aluminum liquid pouring process of cylinder production as an example to illustrate how this application achieves fault detection and early warning based on the following data. The aluminum liquid pouring process involves the transportation and injection of aluminum liquid, and each step will generate corresponding parameter indicators, such as the temperature of the aluminum liquid and the injection height. After the sensor uploads this data to the server at a certain time interval, the server will perform fault detection based on the following data.

[0117] If a production failure is found, in addition to sending an early warning to the mobile terminal and the workshop, a corresponding report of all the screened features will be generated and sent to the staff, so that the staff can intuitively find which indicator is abnormal. For example, at the time of the failure, it is found that the temperature of the aluminum liquid has fluctuated violently or is too high or too low. The staff can investigate the cause of the failure according to the relevant situation and quickly adjust the factors related to the failure.

[0118] Table 1 Index data table corresponding to automobile engine cylinder manufacturing.

[0119]

[0120] Table 1 mainly explains the controllable parameters involved in the aluminum infusion process of the production process. The table simply lists the data characteristics of each parameter within the time window of fault detection, which can help us better understand the distribution and characteristics of the data set. For example, the mean can reflect the concentration of the data set, and the variance can reflect the dispersion of the data, which is conducive to subsequent fault detection and model application.

[0121] Technical features not described in the present invention can be achieved through or by adopting existing technologies and will not be described in detail here. Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A cylinder fault warning system based on dual drive of mechanism and data, characterized in that: The architecture module consists of production workshops, central servers and mobile users. From sensor components in the production workshop to mobile users, it realizes data collection, storage, analysis and early warning functions; The central server is responsible for receiving and storing data from sensor components in the production workshop. The stored data is used for subsequent fault detection and analysis, as well as feedback on workshop working conditions.

2. The cylinder fault warning system based on dual drive of mechanism and data according to claim 1 is characterized in that: The production workshop is installed with any one of a temperature sensor, a humidity sensor, and a pressure sensor, or a combination of at least two or more of them, for sensing the environmental data of the workshop and transmitting the real-time collected data to a central server via a network.

3. The cylinder fault warning system based on dual drive of mechanism and data according to claim 1 is characterized in that: A fault detection and analysis module is deployed at the end of the central server, and the fault detection and analysis module identifies whether there is a device fault by analyzing the stored data; When the fault detection and analysis module detects a device fault, it triggers a fault warning sending module for the mobile terminal user, and the fault warning sending module sends fault warning information to the mobile terminal user; At the same time, the characteristic trend change graph sending module on the central server screens out the characteristic trend change graph and sends it to the mobile user. The characteristic trend change graph intuitively displays the abnormal trend of the data, helping the operator to better understand the equipment status and make corresponding adjustments.

4. The cylinder fault warning system based on dual drive of mechanism and data according to claim 1 is characterized in that: The mobile user uses a mobile application to communicate with the central server through the mobile device. The mobile application receives the fault warning information sent by the central server and displays the fault warning information to the operator of the mobile user in real time. The operator views the fault details through the mobile application.

5. The cylinder fault warning system based on dual drive of mechanism and data according to claim 3 is characterized in that: The server includes a data layer, a processing layer and an application layer; The data layer includes a data acquisition unit and a cleaning component, and a database module. The data acquisition unit and the cleaning component are responsible for collecting raw data from devices or sensors, and performing necessary cleaning and preprocessing to ensure data quality and availability; the database module is a place to store cleaned and processed data, providing support for model analysis at the processing layer, and retaining historical data for future analysis and learning.

6. The cylinder fault warning system based on dual drive of mechanism and data according to claim 5 is characterized in that: The processing layer includes a feature recognition module, a fault detection module, and a feature visualization module. The feature recognition module recognizes the features of the cleaned data based on the dual drive of mechanism and data; Based on the mechanism drive, the cleaned data is identified, and according to the expert opinions and practical experience, the features that are strongly related to the machine operation are defined to obtain the feature set X m ,Engine cylinder casting involves five factors, namely sand core preparation, mold design, pouring system design, pouring temperature control and metal composition uniformity, which are represented as A1...A5; Based on data-driven, the features of the cleaned data are identified. From the perspective of data, the feature set X1 is obtained by using the lasso technology. Y is whether the cylinder body is faulty, X is all the features obtained from the sensors in the production workshop, β is the regression coefficient of the feature and whether it is faulty, λ is the penalty parameter, ||·||1 represents the L1 norm, that is, the sum of the absolute values ​​of the coefficients. The size of λ determines the sparsity of the regression coefficient. When λ is larger, the regression coefficient is sparser, that is, more features are excluded. The L1 regularization term makes part of the parameters in the parameter vector of the model become zero, thereby realizing the function of feature selection. minimize||Y-Xβ|| 2 +λ||β||1 (1) Finally, the features obtained in the above two steps are combined to obtain the feature set X that is most conducive to feedback on the machine operation status and accurate fault prediction. f ; X1=x i ∈X|Lasso(X,Y,λ) i≠0 (3) X f =X m ∪X1 (4) Among them, X m It is a feature set obtained based on the mechanism, Ai refers to five mechanism levels respectively, and F is a function summary that represents the process of feature screening; X1 is the feature set obtained by lasso, where lasso is the abbreviation of the function, Y is the fault index, X is all the features obtained from the sensors in the production workshop, and λ is the penalty parameter of the lasso method; X f is the final feature set, taking X1 and X m The union of; The fault detection module is based on the fusion feature dataset X f , using deep learning models for feature learning and predicting failures in production in the next time period; The feature visualization module integrates the feature dataset X f Conduct data statistics and visualization, and generate relevant reports to help staff gain an intuitive understanding of the machine's operating status.

7. The cylinder fault warning system based on dual drive of mechanism and data according to claim 5 is characterized in that: The application layer includes a feature push module and a fault warning module; The feature push module visualizes and summarizes the working status of the machine and sends it to the mobile terminal user, helping the staff to intuitively view the changing trends and abnormal points of the workshop data; When the detection module of the processing layer finds a fault, the fault warning module triggers a warning and sends a warning signal to the mobile terminal user.

8. The cylinder failure warning system based on dual drive of mechanism and data according to claim 2 is characterized in that: The production workshop includes a data layer and an application layer; The data layer includes a workshop sensor module, a data collection module and a data transmission module; the workshop sensor is responsible for real-time monitoring of the operating status of the equipment in the workshop and sending the collected data to the data collection component; the data collection module is responsible for receiving the raw data from the sensor and performing preprocessing; the data transmission module sends the collected data to the central server for further analysis and processing to ensure timely and accurate data transmission; The application layer includes a feature display module and a fault warning module. The feature display module is responsible for displaying the data status and change trends of machine features to users. These data include key parameters such as temperature, vibration, and pressure. The fault warning module receives analysis results from the data layer. If a production failure occurs, the module sends a fault warning signal to the staff to remind them to respond to potential problems in a timely manner.

9. The cylinder fault warning system based on dual drive of mechanism and data according to claim 4 is characterized in that: The mobile user includes a data layer and an application layer; The data layer includes a data receiving module and a data sending module; the data receiving module is responsible for receiving data from the central server, including real-time feature data and fault warning information of the equipment, and the data receiving module ensures that the user can obtain the latest equipment status information on the mobile terminal; the data sending module is used to send the user's operations and feedback in the mobile terminal application to the central server, including the user's response to the warning, maintenance request or other key information upload; The application layer includes a feature display module and a fault warning module; the feature display module will display the received feature data information and visualization diagrams to the staff for viewing, thereby assisting the staff to quickly understand the production situation in the workshop; the fault warning module is used to provide users with real-time fault or abnormality warnings. When the server analyzes potential problems and sends warning information, the module will receive the information and notify the user through the mobile application.

10. A method according to any one of claims 1 to 9, characterized in that: The following steps are involved: (1) The system in the production workshop automatically collects data sent back by sensors; (2) The collected data needs to be sent to a central processing unit for further analysis and processing; (3) The central processing unit of the central server pre-processes the data, including data cleaning and normalization, to provide accurate input for fault diagnosis; (4) Feature identification of the cleaned data. Among them, the mechanism feature is mainly the feature that has a strong correlation with the machine operation from the expert opinions and actual operations. The data feature is mainly extracted from the data perspective based on the lasso method. Finally, the features obtained by the two feature recognition methods are merged to obtain the feature set that is most conducive to feedback on the machine operation status and accurate fault prediction; (5) Use the Imbalance-XGBoost model to learn the filtered and aggregated data. After the model is trained, it can predict production failures at future time points. (6) When the model predicts that a failure will occur in the future, the central server sends a failure warning to the mobile user; (7) After receiving the warning, the operator of the mobile user needs to investigate the cause of the fault; (8) If it is determined that the fault is caused by improper workshop environment and parameters, the staff will take measures to deal with it on their own; if the fault is solved, the warning is closed and feedback is given to the workshop; if the fault cannot be solved, the timer will be used to perform fault detection and warning again.