Fault monitoring analysis method and system for lubricating oil production

Abstract

The invention relates to the field of production fault analysis, in particular to a fault monitoring analysis method and system for lubricating oil production. The method comprises the following steps: acquiring a multi-dimensional working monitoring parameter and an omnibearing working image; performing abnormal outlier filtering processing on the multi-dimensional working monitoring parameters, and performing time sequence evolution simulation to generate time sequence state evolution data; performing continuous frame optical flow motion analysis on the omnibearing working image, performing dynamic parameter mapping based on time sequence state evolution data, and constructing a dynamic production twinborn model; obtaining a preset production standard demand log; carrying out dynamic process deviation detection based on a preset production standard demand log dynamic production twinborn model, and carrying out production fault analysis so as to obtain stage production fault data; performing potential fault traceability speculation on the staged production fault data to obtain an abnormal fault potential node; and constructing a potential fault mark twinborn model based on the abnormal fault potential nodes. According to the invention, efficient and accurate fault analysis of lubricating oil production is realized.

Description

Technical Field

[0001] The present invention relates to the field of production fault analysis, and in particular to a fault monitoring and analysis method and system for lubricating oil production. Background Art

[0002] In today's highly automated industrial production, lubricating oil, as a key energy medium, is crucial to the normal operation of equipment due to its quality and stability. However, traditional lubricating oil production and monitoring technologies have many limitations and cannot meet the increasingly stringent requirements of modern factories for lubricating oil quality and safety.

[0003] On the one hand, there are many influencing factors in the existing lubricant production process, such as raw material property deviation, equipment aging, and process parameter out of control. These factors lead to abnormal performance of lubricant products, which in turn cause equipment failure. However, traditional fault diagnosis methods often rely on empirical judgment and manual inspection, which are inefficient and difficult to timely discover and locate the root cause of the fault. On the other hand, due to the complex production environment and huge process data, a single empirical rule is difficult to fully cover various fault modes, and it is easy to have false alarms or omissions, affecting production safety and reliability. At the same time, the lack of visual fault diagnosis and maintenance support means for on-site operators also exacerbates the difficulty of fault response and disposal.

[0004] Therefore, there is an urgent need for an intelligent fault monitoring and analysis method that can realize real-time monitoring and intelligent diagnosis of the entire lubricant production process, greatly improve the accuracy of fault warning and disposal, thereby ensuring the stability of lubricant quality and ensuring the safe and reliable operation of equipment. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a fault monitoring and analysis method and system for lubricating oil production to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a fault monitoring and analysis method for lubricating oil production, comprising the following steps: Step S1: acquiring multi-dimensional working monitoring parameters and omnidirectional working images; performing time-series evolution simulation on the multi-dimensional working monitoring parameters to generate time-series state evolution data; Step S2: Perform continuous frame optical flow motion analysis on the omnidirectional working image, and perform dynamic parameter mapping based on the time series state evolution data to build a dynamic production twin model; Step S3: Perform production failure analysis on the dynamic production twin model to obtain staged production failure data; Step S4: Conduct potential fault tracing and speculation on the staged production fault data, and build a potential fault marking twin model; Step S5: simulating the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; Step S6: Dynamically optimize the production of the lubricating oil production line based on the fault impact range data and build a fault monitoring optimization model.

[0007] The present invention performs abnormal outlier filtering and time series evolution simulation on multi-dimensional work monitoring parameters, identifies abnormal situations and simulates the change process of system status, monitors the production environment in real time, discovers potential problems in advance, and lays the foundation for future fault prediction. Through continuous frame optical flow motion analysis and dynamic parameter mapping, real-time monitoring of the production process and establishment of a dynamic production twin model are achieved, which enables operators to better understand the changes in production status and make timely adjustments and optimizations. The dynamic production twin model is used to effectively detect process deviations and perform production failure analysis, improve production quality, reduce failure rates, and optimize production efficiency. By inferring the source of potential faults and constructing a potential fault marker twin model, it helps to determine the root cause of the problem and provide support for formulating effective fault handling solutions. By simulating the evolution of fault propagation and predicting the scope of fault impact, potential risks are identified in time and necessary measures are taken to reduce the risk of damage to the production line and improve production stability. By iteratively adjusting production parameters and extracting the optimal production parameter regulation strategy, the production process is optimized, production efficiency and product quality are improved, a fault monitoring optimization model is constructed to discover potential faults in advance, and appropriate measures are taken to minimize production interruptions.

[0008] Preferably, step S1 comprises the following steps: Step S11: Performing all-round work monitoring of the lubricating oil production line based on multiple sensors, extracting multi-dimensional work monitoring parameters and all-round work images; Step S12: performing abnormal outlier filtering on the multi-dimensional working monitoring parameters to generate consistent monitoring parameters; Step S13: performing time series parameter change analysis on the consistent monitoring parameters to generate time series monitoring parameter change data; Step S14: perform deep correlation mining on multi-dimensional work monitoring parameters to identify correlation rules between parameters; Step S15: Based on the correlation rules between parameters, a time series evolution simulation is performed on the time series monitoring parameter change data to generate time series state evolution data.

[0009] The present invention realizes all-round monitoring of the lubricating oil production line through multi-sensor monitoring and parameter extraction, obtains multi-dimensional working monitoring parameters and all-round working images, and comprehensively understands various changes and situations in the production process. It eliminates the interference of abnormal data through abnormal outlier filtering processing, thereby generating consistent monitoring parameters, ensuring the accuracy and consistency of the data, and providing a reliable data basis for subsequent analysis. Through time series parameter change analysis, the changing trend of monitoring parameters over time is revealed, and time series monitoring parameter change data is generated to identify periodic changes, abnormal events, and potential problems over time. Through parameter deep correlation mining, the correlation rules between different monitoring parameters are discovered, the mutual influence and dependency between parameters are identified, and the relationship between parameters is more comprehensively understood, providing a deeper foundation for subsequent analysis. The correlation rules between parameters are used to perform time series evolution simulation, and time series state evolution data is generated, thereby identifying the changing patterns and trends of parameters under different states, simulating system state changes, and providing a more accurate reference for fault monitoring and prediction.

[0010] Preferably, the specific steps of step S12 are: Define standard sampling timestamp nodes; Performing time synchronization processing on multi-dimensional work monitoring parameters based on standard sampling timestamp nodes to generate time-synchronized work monitoring parameters; Perform abnormal outlier data detection on time synchronization work monitoring parameters and mark abnormal outlier data points; Performing abnormal outlier filtering based on abnormal outlier data points to generate outlier filtering monitoring parameters; The outlier filtering monitoring parameters are processed by parameter scale consistency to generate consistent monitoring parameters.

[0011] The present invention defines a standard sampling timestamp node to ensure that all data are sampled based on the same time standard, thereby unifying the time dimension and facilitating subsequent time-related processing and analysis. The time synchronization process ensures that the data of each monitoring parameter remains consistent in time, eliminates data dislocation or confusion caused by time asynchrony, and provides an accurate data basis for subsequent analysis. Through abnormal outlier data detection, abnormal points in the monitoring data are timely identified, these abnormal data points are marked, and the abnormal situations are further analyzed and processed. Through abnormal outlier filtering, abnormal data points are eliminated to ensure the accuracy and reliability of the data, and provide a more reliable data basis for subsequent analysis and processing. Through parameter scale consistency processing, it is ensured that the scale of each monitoring parameter is consistent, and the comparison difficulties caused by different scales are eliminated, making it easier to compare and analyze different parameters.

[0012] Preferably, step S2 comprises the following steps: Step S21: performing multi-frame sampling on the omnidirectional working image to obtain multiple image frame sequences; Step S22: visually identifying key points of the production line on multiple image frame sequences and marking characteristic points of the production line; Step S23: Calculating the spatial position change of the feature points of the production line to obtain the spatial position coordinates of the feature points at different times; Step S24: performing optical flow motion analysis of consecutive frames based on the spatial position coordinates of the feature points at different times to generate optical flow motion trajectories of the feature points; Step S25: Dynamically map the optical flow motion trajectory of the feature points based on the time series state evolution data to build a dynamic production twin model.

[0013] The present invention observes the continuous changes in the working process of the production line through multi-frame image sampling, captures more subtle changes and features, and accurately identifies the key feature points in the production line through key point visual recognition. Subsequent analysis and tracking, through spatial position change calculation, tracks the position changes of the feature points of the production line over time, and provides spatial information about the working status of the production line. Through optical flow motion analysis, the motion trajectory of the feature points over time is obtained, revealing the movement conditions and trends of each feature point in the production line. Through dynamic parameter mapping, the motion trajectory of the feature points is combined with the time series state evolution data to construct a dynamic production twin model, so as to realize dynamic modeling and monitoring of the working status of the production line.

[0014] Preferably, the specific steps of step S25 are: Perform dynamic production process behavior analysis on the optical flow motion trajectory of feature points to generate dynamic production process behavior features; Perform multi-stage process recognition on the optical flow motion trajectory of feature points to obtain multi-stage process data; Based on the dynamic production process behavior characteristics, the process data of multiple stages are mined for logical context to generate logical dependencies between the processes of each stage; The time-series state evolution data is used to dynamically map the logical dependencies of the processes at each stage and build a dynamic production twin model.

[0015] The present invention captures the dynamic behavior characteristics in the production process through optical flow motion trajectory analysis, helping to identify changes and abnormal conditions in the process. Through multi-stage process identification, the production process is divided into different stages to better understand the flow and changes of the production process. Through logic mining, the logical dependencies between processes in each stage are identified, revealing the associations and influences between different processes. Through dynamic parameter mapping, the logical dependencies of processes in each stage are combined with the time series state evolution data to establish a dynamic production twin model, thereby realizing dynamic modeling and monitoring of the production process.

[0016] Preferably, step S3 specifically comprises the following steps: Step S31: Obtaining a preset production standard requirement log; Step S32: performing phased demand identification on the preset production standard demand log, and extracting demand feature data of multiple stages; Step S33: Performing dynamic process deviation detection on the dynamic production twin model based on the demand feature data of multiple stages to generate staged production process deviation data; Step S34: Perform production failure analysis on the staged production process deviation data to obtain staged production failure data. The present invention ensures the standardized recording and tracking of production requirements by obtaining preset production standard requirement logs, compares actual production conditions with standard requirements, promptly discovers deviations and problems, subdivides production requirements into different stages through phased requirement identification and feature extraction, and extracts feature data for each stage to understand the demand characteristics of each production stage in more detail, performs process deviation detection through a dynamic production twin model, monitors deviations in the production process in real time, generates phased process deviation data, promptly discovers anomalies and deviations in the production process, performs fault analysis on process deviation data, identifies factors and patterns that lead to faults, and then obtains phased production fault data, predicts and prevents potential faults, and improves production efficiency and product quality.

[0017] Preferably, the specific steps of step S4 are: Step S41: performing fault type analysis on the staged production fault data to extract production fault type data; Step S42: tracing and inferring potential fault sources on the production fault type data to obtain potential nodes of abnormal faults; Step S43: Based on the abnormal fault potential nodes, the dynamic production twin model is marked for potential fault location to obtain a potential fault marking twin model.

[0018] The present invention can accurately identify different types of production failures and extract production failure type data through fault type analysis. Establish targeted fault handling strategies and preventive measures. Through potential fault tracing and speculation, find out the potential root causes and nodes that lead to production failures, and obtain potential nodes of abnormal failures. Deeply understand the causes and mechanisms of failures. Use potential fault nodes to locate and mark the twin model, accurately locate the location and nodes where the failure occurs, and obtain a potential fault marking twin model. Early warning and management of potential failures to reduce the impact of failures on the production process.

[0019] Preferably, the specific steps of step S5 are: Step S51: performing upstream and downstream correlation analysis on the potential nodes of abnormal faults to generate upstream and downstream correlation logic of the fault points; Step S52: Based on the upstream and downstream association logic of the fault point, the potential fault mark twin model is subjected to fault diffusion path evolution, thereby obtaining multiple potential fault diffusion paths; Step S53: simulating the evolution of fault propagation on multiple potential fault diffusion paths to obtain the law of fault propagation evolution; Step S54: predicting the fault impact range of the lubricating oil production line based on the fault propagation evolution law to generate fault impact range data. The present invention establishes the correlation logic between abnormal fault nodes through upstream and downstream correlation analysis, and understands the causal relationship and impact path between faults. Identify potential fault propagation paths and impact ranges. Generate multiple potential fault diffusion paths through the evolution of fault diffusion paths, revealing the propagation paths and impact ranges of different faults in the production system. Predicted fault propagation. Through fault propagation evolution simulation, obtain the propagation law and evolution trend of the fault in the system. In-depth understanding of the fault propagation mechanism provides a basis for preventing and responding to faults. Predict the impact range of the production line through the fault propagation law, generate fault impact range data, and help determine the scope and degree of impact caused by the fault. Take appropriate measures to reduce the losses caused by the fault.

[0020] Preferably, the specific steps of step S6 are: Step S61: Performing a dynamic risk assessment on the fault impact range data to generate a fault risk assessment value; Step S62: performing iterative production parameter regulation based on the fault risk assessment value to generate multiple production parameter regulation strategies; Step S63: performing production simulation evolution on multiple production parameter control strategies to generate multiple production parameter control simulation data; Step S64: performing production optimization evaluation on multiple production parameter control simulation data to generate an evaluation value for each control strategy; Step S65: extracting the optimal production parameter control strategy based on the evaluation value of each control strategy; Step S66: Dynamically optimize the production of the lubricating oil production line according to the optimal production parameter control strategy, and build a fault monitoring optimization model.

[0021] The present invention can quantify the risks brought by failures and generate corresponding risk assessment values ​​through dynamic risk assessment. Potential risk points are identified and corresponding measures are taken to reduce risks. According to the risk assessment value, multiple production parameter control strategies are generated by iteratively adjusting the production parameters. Help optimize the production process and reduce the risk of failure. Through production simulation evolution, simulation data of multiple production parameter controls are generated. Evaluate the effects of various control strategies and production results. Through production optimization evaluation, each control strategy is evaluated to determine which strategies are most effective in practical applications. Select the optimal control scheme. According to the evaluation value, the production parameter control strategy with the best performance is extracted. Determine the most effective production parameter configuration method. Through dynamic production optimization, the production line is adjusted according to the optimal strategy, and a fault monitoring optimization model is established. Improve production efficiency and quality and reduce the probability of failure.

[0022] In this specification, a fault monitoring and analysis system for lubricating oil production is provided, which is used to execute the fault monitoring and analysis method for lubricating oil production as described above, including: The time series evolution module is used to obtain multi-dimensional work monitoring parameters and all-round work images; perform abnormal outlier filtering on the multi-dimensional work monitoring parameters, and perform time series evolution simulation to generate time series state evolution data; The twin model module is used to perform continuous frame optical flow motion analysis on omnidirectional working images, and to perform dynamic parameter mapping based on time-series state evolution data to build a dynamic production twin model; The fault analysis module is used to analyze production faults on the dynamic production twin model to obtain staged production fault data; The fault tracing module is used to tracing and infer potential fault sources based on periodic production fault data and to build a potential fault marking twin model; The fault propagation evolution module is used to simulate the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; The production parameter control module dynamically optimizes the lubricant production line based on the fault impact range data and builds a fault monitoring optimization model.

[0023] The present invention realizes comprehensive monitoring and analysis of working status by acquiring multi-dimensional working monitoring parameters and all-round working images. Abnormal outlier filtering processing and time series evolution simulation help identify abnormal situations in work and generate meaningful time series state evolution data to provide basic information for subsequent analysis. Through continuous frame optical flow motion analysis and dynamic parameter mapping, a dynamic production twin model is constructed to realize real-time monitoring and prediction of the production process. Combined with the time series state evolution data, this module helps simulate the evolution of working status, discovers potential problems in advance and predicts them, thereby improving production efficiency and quality. Dynamic process deviation detection and production fault analysis are performed based on the dynamic production twin model to identify abnormal situations in production. By analyzing the preset production standard requirement log, process deviations are detected in time, fault analysis is performed and phased results are obtained. Production fault data provides a basis for subsequent fault handling. Through potential fault tracing and identification of potential nodes of abnormal faults, potential problems can be quickly located and solved. A potential fault marking twin model is constructed to help predict the occurrence and propagation path of faults, and measures can be taken in advance to reduce the impact of faults on production. Through fault propagation evolution simulation and fault impact range prediction, the production department can be helped to respond to problems in a timely manner. Fault impact range data is generated to guide production parameter regulation, reduce the negative impact of faults on the entire production process, and improve production efficiency and stability. Through iterative production parameter regulation and extraction of optimal production parameter regulation strategies, the production process is optimized. According to the optimal production parameter regulation strategy, the lubricant production line is dynamically optimized, and a fault monitoring optimization model is constructed to sustainably improve production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of a fault monitoring and analysis method for lubricating oil production according to the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0026] The present application example provides a fault monitoring and analysis method and system for lubricating oil production. The execution subject of the fault monitoring and analysis method and system for lubricating oil production includes but is not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of an audio and image management system, an information management system, and a cloud data management system.

[0027] See also Figures 1 to 4 The present invention provides a fault monitoring and analysis method for lubricating oil production, the fault monitoring and analysis method for lubricating oil production comprising the following steps: Step S1: acquiring multi-dimensional working monitoring parameters and omnidirectional working images; performing time-series evolution simulation on the multi-dimensional working monitoring parameters to generate time-series state evolution data; Step S2: Perform continuous frame optical flow motion analysis on the omnidirectional working image, and perform dynamic parameter mapping based on the time series state evolution data to build a dynamic production twin model; Step S3: Perform production failure analysis on the dynamic production twin model to obtain staged production failure data; Step S4: Conduct potential fault tracing and speculation on the staged production fault data, and build a potential fault marking twin model; Step S5: simulating the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; Step S6: Dynamically optimize the production of the lubricating oil production line based on the fault impact range data and build a fault monitoring optimization model.

[0028] The present invention performs abnormal outlier filtering and time series evolution simulation on multi-dimensional work monitoring parameters, identifies abnormal situations and simulates the change process of system status, monitors the production environment in real time, discovers potential problems in advance, and lays the foundation for future fault prediction. Through continuous frame optical flow motion analysis and dynamic parameter mapping, real-time monitoring of the production process and establishment of a dynamic production twin model are achieved, which enables operators to better understand the changes in production status and make timely adjustments and optimizations. The dynamic production twin model is used to effectively detect process deviations and perform production failure analysis, improve production quality, reduce failure rates, and optimize production efficiency. By inferring the source of potential faults and constructing a potential fault marker twin model, it helps to determine the root cause of the problem and provide support for formulating effective fault handling solutions. By simulating the evolution of fault propagation and predicting the scope of fault impact, potential risks are identified in time and necessary measures are taken to reduce the risk of damage to the production line and improve production stability. By iteratively adjusting production parameters and extracting the optimal production parameter regulation strategy, the production process is optimized, production efficiency and product quality are improved, a fault monitoring optimization model is constructed to discover potential faults in advance, and appropriate measures are taken to minimize production interruptions.

[0029] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a fault monitoring and analysis method for lubricating oil production of the present invention. In this example, the steps of the fault monitoring and analysis method for lubricating oil production include: Step S1: acquiring multi-dimensional working monitoring parameters and omnidirectional working images; performing time-series evolution simulation on the multi-dimensional working monitoring parameters to generate time-series state evolution data; In this embodiment, the source of the multi-dimensional working monitoring parameters is determined, such as sensors, monitoring systems or data acquisition devices. The source of the omnidirectional working images is determined to be a monitoring camera or image acquisition device. The data acquisition device is configured, and the acquisition frequency (such as every second, every minute) is set to obtain real-time monitoring parameters and images. It is ensured that each monitoring parameter has a clear description, including temperature, pressure, flow, vibration, etc. The data acquisition system is started to start real-time collection of multi-dimensional working monitoring parameters and omnidirectional working images. The timestamp of the acquisition is recorded for subsequent analysis and synchronization of data. The acquired monitoring parameters and image data are stored in a database or a file of a specified format to ensure the structuring and accessibility of the data. The original data is extracted from the stored multi-dimensional monitoring parameters to ensure that the data is complete and in a consistent format. A suitable anomaly detection method is selected, such as: Z-score method: Calculate the Z-score of each data point, identify outliers that exceed a set threshold (such as 3), IQR (interquartile range) Methods: Identify outliers below Q1-1.5IQR or above Q3+1.5IQR, apply the selected algorithm to perform abnormal outlier detection on multi-dimensional monitoring parameters, mark and filter out outlier data points, store the filtered monitoring data in a new data frame to ensure that subsequent analysis is based on a clean data set, sort the filtered monitoring parameters by timestamp to ensure that the data is suitable for time series analysis, and select a suitable time series evolution simulation model, such as: ARIMA model: used for time series prediction, suitable for processing stationary and non-stationary data, LSTM (Long Short-Term Memory Network): suitable for processing long time series data, able to capture complex time series relationships, input the filtered monitoring data into the selected time series model for training, generate time series state evolution data, generate prediction data of future time series states through the model, identify potential trends and changes, and store the generated time series state evolution data in a data frame, including the predicted value and its confidence interval at each time point, for subsequent analysis and verification.

[0030] Step S2: Perform continuous frame optical flow motion analysis on the omnidirectional working image, and perform dynamic parameter mapping based on the time series state evolution data to build a dynamic production twin model; In this embodiment, the acquired omnidirectional working image is preprocessed, including denoising, contrast enhancement, etc., to improve the accuracy of subsequent optical flow analysis, ensure the correct time sequence between images, and facilitate optical flow calculation. An optical flow algorithm (such as the Lucas-Kanade method) is used to calculate the motion vector between consecutive frames, extract the motion information of the feature points, and obtain the motion trajectory of the object in the image by calculating the optical flow of each feature point. The motion trajectory of each feature point is extracted based on the result of the optical flow calculation, and the information such as the motion direction and speed is recorded. The motion trajectory of the feature points is visualized for easy analysis and understanding. Matplotlib or OpenCV is used to visualize the motion trajectory, and arrows of the optical flow field are drawn to indicate the motion direction and speed. The obtained feature point motion data is integrated with the previously generated time-series state evolution data to form a comprehensive data set to ensure data alignment in the time dimension for subsequent analysis. Based on the motion trajectory and time-series state data, a dynamic parameter mapping relationship is established to identify the association between motion characteristics and production status, and determine how each parameter (such as speed, acceleration, position, etc.) affects the specific state in the production process. Use multivariate regression analysis or machine learning algorithms (such as random forests, support vector machines) to establish a mapping model between motion characteristics and production status. Based on the mapping relationship, a dynamic production twin model is constructed to simulate state changes in the production process. The motion analysis results are combined with the time-series changes of production parameters to form a complete dynamic model. Use system modeling tools (such as MATLAB Simulink, AnyLogic) to construct a dynamic production twin model to ensure that the model can reflect changes in production status in real time.

[0031] Step S3: Perform production failure analysis on the dynamic production twin model to obtain staged production failure data; In this embodiment, the source of the production standard requirement log is determined, such as a production management system, an enterprise resource planning (ERP) system, a manufacturing execution system (MES) or other related software, and a programming tool (such as Python, SQL) is used to connect to the data source. The conditions for data extraction (such as time range, product type, etc.) are set, and the data extraction process is executed to obtain log data containing production standard requirements. The data structure is ensured to be clear, including fields such as timestamp, product type, production quantity, quality standard, etc., to ensure that the previously constructed dynamic production twin model can receive the input data in the standard requirement log and has the ability to monitor in real time, and to select a suitable dynamic process deviation detection method, such as: Control chart: Use control charts to monitor deviations in the production process and identify points that are out of control. Machine learning algorithm: Apply methods such as anomaly detection and regression analysis to identify deviations between the actual production status and the standard requirements, and extract The production standard requirement log data is input into the dynamic production twin model to monitor the production status in real time. According to the set deviation detection method, the process deviation is identified and recorded, and deviation data is generated, including deviation type, deviation amplitude and related timestamps. The appropriate production failure analysis method is selected, such as: Root Cause Analysis (RCA): Identify the root cause of process deviation, Failure Mode and Effect Analysis (FMEA): Evaluate the impact of potential failures on production, extract failure-related information based on dynamic process deviation data, identify deviation patterns and trends, apply the selected failure analysis method, analyze the causes of process deviations, identify production failures and influencing factors, record the occurrence frequency of each failure, the affected production links and potential improvement measures, and store the generated phased production failure data in the data frame to ensure that each failure data is connected to the corresponding deviation feature, including failure type, occurrence time, impact range and other information.

[0032] Step S4: Conduct potential fault tracing and speculation on the staged production fault data, and build a potential fault marking twin model; In this embodiment, periodic production failure data is collected to ensure that the data includes information such as failure type, occurrence time, affected production links, and related process deviations. A framework for tracing and inferring potential failures is determined, including the definition and analysis method of potential failure nodes. A suitable method for tracing and inferring failure sources is selected, such as: Fault Tree Analysis (FTA): Construct a fault tree to identify potential failure causes in a hierarchical manner. Causal analysis: Analyze the causal relationship between failure data and key nodes in the production process. Based on the failure data and the selected analysis method, identify the source of potential failures. Generate a list of potential nodes for abnormal failures. Record each potential node and the corresponding failure type, occurrence conditions, and related parameters. Store the identified potential nodes for abnormal failures in a data frame. Ensure that the information of each node is clearly visible, including the node Description, impact degree, etc., design the framework of the potential fault marking twin model, determine the model input (such as potential fault nodes, fault types, deviation data) and output (such as fault prediction, state monitoring, etc.), use modeling tools (such as MATLAB, Simulink or Python modeling library) to implement the twin model of potential fault marking, embed the information of potential fault nodes in the model, ensure that the model can reflect the status of potential faults in real time, mark the potential fault nodes in the dynamic production twin model, highlight the fault area and its influencing factors, verify the accuracy of potential fault marking through model simulation, ensure that the model can effectively identify the impact of faults, record the constructed potential fault marking twin model and its related parameters in the document, and ensure the reproducibility and verifiability of the model.

[0033] Step S5: simulating the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; In this embodiment, ensure that the potential fault marking twin model has been built and can receive input data (such as potential fault nodes, production status, etc.). Design a framework for fault propagation evolution, and determine the input (fault type, propagation path, influencing factors) and output (fault state evolution, propagation results) of the model. Analyze the relationship between the nodes in the potential fault marking model, identify the path of fault propagation, including the relationship between direct and indirect influences. Select a suitable simulation method, such as: System dynamics: simulate the dynamic changes of various factors in the system and analyze the impact of fault propagation. Discrete event simulation (DES): simulate the order of events in the system to capture the process of fault propagation. Input the state of the potential fault node into the selected simulation method, perform fault propagation evolution simulation, and record the propagation process and state change of the fault in the system. Determine the standards and methods for evaluating the scope of fault impact, such as: Impact scope indicators: including equipment, production lines, personnel safety, etc. affected by the fault. Impact degree classification: divide the impact degree into mild, moderate and severe. Select a suitable impact prediction model, such as: Regression analysis: used to quantify the relationship between fault impact and production parameters. Machine learning models (such as random forests, support vector machines): used to predict the impact of faults on different production links. Combine the results of fault propagation simulation with the impact prediction model, input fault status data, and calculate the impact range of the fault on each production link. Generate fault impact range data, including information such as affected equipment, production process, and potential economic losses. Store the generated fault impact range data in a data frame, and use visualization tools (such as Tableau, Power BI) to display the impact range and prediction results to facilitate relevant personnel to understand and make decisions.

[0034] Step S6: Dynamically optimize the production of the lubricating oil production line based on the fault impact range data and build a fault monitoring optimization model.

[0035] In this embodiment, the fault impact range data is collected to ensure that the data includes the affected equipment, production links, fault types and related production parameters (such as temperature, pressure, flow rate, etc.), and the control objectives are determined, such as reducing the occurrence rate of faults, improving production efficiency, optimizing resource utilization, etc. A preliminary production parameter control strategy is designed, considering the impact of different parameters on production, such as the ratio of lubricating oil, flow rate, temperature, etc., and selecting a suitable control model, for example: optimization algorithms (such as genetic algorithms, particle swarm optimization) are used to find the optimal production parameters, control theory (such as PID control) is used to adjust the production parameters in real time, the fault impact range data is input into the selected control model, and the iterative process of parameter control is executed. According to the model output, the production parameters are adjusted, and the changes in their impact on the fault are monitored, and the effect after each control is recorded. The data in the iterative control process is analyzed, and the effects of different parameter configurations on production efficiency and fault impact are compared. A suitable strategy identification method is selected, for example: multi-objective optimization: an optimization method that considers multiple objectives (such as cost, efficiency, and failure rate) at the same time, decision tree analysis: used to identify the optimal parameter combination, and based on the analysis results, the optimal production that can effectively reduce the impact of faults and improve production efficiency is extracted. Parameter control strategy, and form a strategy document, record the optimal production parameter control strategy in the document to ensure traceability and enforceability, evaluate the operating status of the existing lubricant production line, ensure that the equipment, process and personnel can support the implementation of the new strategy, dynamically adjust the lubricant production line according to the extracted optimal production parameter control strategy, monitor the key parameters in the production process in real time, implement new parameter settings on the production line, and ensure that each link can cooperate with the requirements of the new strategy, monitor the operating status of the adjusted production line, record key indicators such as production efficiency and failure rate to evaluate the effect of the new strategy, design the framework of the fault monitoring optimization model, determine the input (such as production parameters, fault data, operating status) and output (such as fault warning, optimization suggestions) of the model, use modeling tools (such as MATLAB, Simulink or Python modeling library) to implement the fault monitoring optimization model, ensure that the model can process input data in real time and generate optimization suggestions, verify the accuracy of the model through historical data and real-time data, make necessary parameter adjustments to improve the prediction ability of the model, record the constructed fault monitoring optimization model and its related parameters in the document to ensure the reproducibility and verifiability of the model.

[0036] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: Performing all-round work monitoring of the lubricating oil production line based on multiple sensors, extracting multi-dimensional work monitoring parameters and all-round work images; Step S12: performing abnormal outlier filtering on the multi-dimensional working monitoring parameters to generate consistent monitoring parameters; Step S13: performing time series parameter change analysis on the consistent monitoring parameters to generate time series monitoring parameter change data; Step S14: perform deep correlation mining on multi-dimensional work monitoring parameters to identify correlation rules between parameters; Step S15: Based on the correlation rules between parameters, a time series evolution simulation is performed on the time series monitoring parameter change data to generate time series state evolution data.

[0037] In this embodiment, a variety of sensors are installed at key positions of the lubricating oil production line, including temperature sensors, pressure sensors, flow meters, vibration sensors, etc., to achieve comprehensive monitoring. An appropriate data collection frequency (such as every second, every minute) is set and a data recording device is configured to ensure real-time acquisition of multi-dimensional working monitoring parameters. A camera or image sensor is installed to regularly capture all-round working images of the production line to ensure that the image quality meets the analysis requirements. A programming tool (such as Python) is used to perform preliminary cleaning on the collected multi-dimensional monitoring parameters to remove obvious erroneous data (such as negative values ​​and extreme values). Z-score method: calculate the Z-score of each data point and identify outliers that exceed a set threshold (such as 3). IQR (interquartile range) method: identify outliers below Q1-1.5IQR or above Q3+1.5IQR by calculating IQR. Use the selected algorithm to identify and remove outliers. Generate a consistent monitoring parameter set to ensure the integrity and accuracy of the data set. Sort the consistent monitoring parameters by timestamp to ensure that the data set is suitable for time series analysis. Moving average: calculate the moving average under different windows and smooth the data to identify trends. Autocorrelation analysis: Use the autocorrelation function (ACF) and partial autocorrelation function (PACF) to identify the periodicity of time series data, apply the selected method to analyze the changes of consistent monitoring parameters, generate time series monitoring parameter change data, identify trends and periodicity, integrate multi-dimensional working monitoring parameters into a unified data set, and ensure correlation analysis between various parameters. Pearson correlation coefficient: calculate the correlation between parameters, association rule learning: use Apriori algorithm or FP-growth algorithm to mine association rules between parameters, apply the selected algorithm to analyze multi-dimensional monitoring parameters, identify the association rules between parameters, generate association rules and correlation matrix, and select suitable time series evolution simulation methods, such as: time series prediction model: use ARIMA model for time series prediction, state transition model: construct a state transition matrix based on Markov model, input the time series monitoring parameter change data into the selected simulation model, ensure that the data format meets the model requirements, run the selected model, generate time series state evolution data, predict the parameter change trend in the future state, store the generated time series state evolution data in the data frame, and ensure that the prediction results at each time point are clearly recorded.

[0038] In this embodiment, the specific steps of step S12 are: Define standard sampling timestamp nodes; Performing time synchronization processing on multi-dimensional work monitoring parameters based on standard sampling timestamp nodes to generate time-synchronized work monitoring parameters; Perform abnormal outlier data detection on time synchronization work monitoring parameters and mark abnormal outlier data points; Performing abnormal outlier filtering based on abnormal outlier data points to generate outlier filtering monitoring parameters; The outlier filtering monitoring parameters are processed by parameter scale consistency to generate consistent monitoring parameters.

[0039] In this embodiment, according to the characteristics of the production process and the monitoring requirements, a suitable sampling frequency (for example, every second or every minute) is determined, the format of the timestamp is unified (such as ISO 8601), and it is ensured that all collected data are processed according to a unified time standard. Time node generation: based on the set sampling frequency, a list of standard sampling timestamp nodes is generated, raw data is extracted from the multi-dimensional working monitoring parameters, and it is ensured that the data contains timestamps and various monitoring parameters. The raw data is synchronized in time using an interpolation method (such as linear interpolation, spline interpolation) or a resampling method (such as the resample function of Pandas). For each standard sampling timestamp node, the corresponding raw data value is found. If there is a missing value, the missing data is filled using the interpolation method to ensure that all timestamps have corresponding monitoring parameter values. A suitable anomaly detection algorithm is selected, such as the Z-score, IQR method or machine learning method (such as isolation forest), and the selected algorithm is applied to perform anomaly detection on the time-synchronized working monitoring parameters to identify abnormal outliers in the data. Abnormal points are marked: the detected abnormal outlier data points are marked to facilitate subsequent processing. The process is set In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: performing multi-frame sampling on the omnidirectional working image to obtain multiple image frame sequences; Step S22: visually identifying key points of the production line on multiple image frame sequences and marking characteristic points of the production line; Step S23: Calculating the spatial position change of the feature points of the production line to obtain the spatial position coordinates of the feature points at different times; Step S24: performing optical flow motion analysis of consecutive frames based on the spatial position coordinates of the feature points at different times to generate optical flow motion trajectories of the feature points; Step S25: Dynamically map the optical flow motion trajectory of the feature points based on the time series state evolution data to build a dynamic production twin model.

[0040] In this embodiment, a sampling strategy is set from the omnidirectional working image, and an appropriate time interval (such as every second or every minute) is selected for image sampling. According to the set time interval, multiple image frames are extracted from the video stream or image sequence to form an image frame sequence. The multiple image frame sequences obtained by sampling are stored in a specified format (such as a folder or a database), ensuring that the timestamp and sequence number of each frame are clearly visible. import cv2 import time cap = cv2.VideoCapture(0)# Open the camera for i in range(10):# Collect 10 frames ret, frame = cap.read() cv2.imwrite(f'frame_{i+1:02d}.jpg', frame) time.sleep(0.1)# Collect one frame every 0.1 seconds cap.release() Select an appropriate computer vision algorithm, a feature point detection algorithm such as SIFT (Scale Invariant Feature Transform), SURF (Speeded Up Robust Features), or ORB (Oriented FAST and Rotated BRIEF), apply the selected algorithm to multiple image frame sequences, identify and mark the key feature points of the production line (such as machines, conveyor belts, sensors, etc.), mark the feature points in each image frame, and store the recognition results in a data frame. import cv2 for i in range(1, 11): img = cv2.imread(f'frame_{i:02d}.jpg') gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) corners = cv2.cornerHarris(gray, 2, 3, 0.04) img[corners>0.01 * corners.max()] = [0, 0, 255]# Mark feature points in red cv2.imwrite(f'marked_frame_{i:02d}.jpg', img) Including the coordinates of the feature points and the timestamp of the corresponding frame, extract the coordinates of the feature points of each time frame (such as (x, y, z) coordinates) from the marked feature point data, compare the coordinates of the feature points in consecutive time frames, calculate the spatial position changes of each feature point at different times, use simple difference methods or more complex motion analysis algorithms (such as optical flow methods) to calculate the changes, select a suitable optical flow algorithm, for example: Lucas-Kanade optical flow method: for optical flow calculation of small motion scenes, Horn-Schunck method: for estimation of global optical flow field, based on the spatial position coordinates of feature points at different times, apply the selected optical flow algorithm, generate the motion trajectory of the feature points, describe the motion of the feature points in consecutive frames, and store the generated feature point optical flow motion trajectory in the data frame, import pandas as pd feature_points = [] for i in range(1, 11): img = cv2.imread(f'frame_{i:02d}.jpg') gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) corners = cv2.cornerHarris(gray, 2, 3, 0.04) coords = np.argwhere(corners>0.01 * corners.max()) for coord in coords: feature_points.append([i, coord[0], coord[1]])# [frame, x, y] df = pd.DataFrame(feature_points, columns=['Frame', 'X', 'Y']) df.to_csv('feature_points.csv', index=0) Including the start and end coordinates and time information of the motion trajectory, collecting time-series state evolution data to support dynamic parameter mapping, associating the time-series state evolution data with the optical flow motion trajectory of the feature points, mapping the dynamic parameters of the feature points in different states (such as speed, acceleration), and building a dynamic production twin model based on the mapping results. The model should be able to reflect the production line status and feature point movement in real time. Use simulation tools or modeling software (such as MATLAB, Simulink) to build and verify the model, and record the constructed dynamic production twin model and its parameters in the document to ensure the reproducibility and verifiability of the model.

[0041] We have 3 image frames (Frame 1, Frame 2, Frame 3), and the coordinates of the feature points in each image frame are as follows: FramePoint IDXY 11100150 12200250 21105155 22195245 31110160 32190240 For each feature point, calculate the change from Frame 1 to Frame 2, and the change from Frame 2 to Frame 3.

[0042] Point IDΔX (Frame 1 to 2)ΔY (Frame 1 to 2)ΔX (Frame 2 to 3)ΔY(Frame 2 to 3) 1105 - 100 = 5155 - 150 = 5110 - 105 = 5160 - 155 = 5 2195 - 200 = -5245 - 250 = -5190 - 195 = -5240 - 245 = -5 The motion trajectory of each feature point is represented as a vector.

[0043] Point IDStart (X, Y)End (X, Y)ΔXΔY 1(100, 150)(110, 160)1010 2(200, 250)(190, 240)-10-10 For Point ID 1: vector = (10, 10) Unit vector = (10 / √(10² + 10²), 10 / √(10² + 10²)) = (0.707, 0.707) For Point ID 2: Vector = (-10, -10) Unit vector = (-10 / √((-10)² + (-10)²), -10 / √((-10)² + (-10)²)) = (-0.707, -0.707) FrameTemperaturePressure 170°C1.5 MPa 272°C1.6 MPa 375°C1.7 MPa Setting model: Temperature = a * ΔX + b * ΔY + c The data were fitted by the least squares method to obtain the model parameters (a, b, c).

[0044] Suppose we obtain the following parameters through regression analysis: a = 0.1 b = 0.1 c = 70 For the movement of Point ID 1 in Frame 2: ΔX = 5, ΔY = 5 Predicted temperature = 0.1 * 5 + 0.1 * 5 + 70 = 72.0°C For the movement of Point ID 2 in Frame 2: ΔX = -5, ΔY = -5 Predicted temperature = 0.1 * (-5) + 0.1 * (-5) + 70 = 68.0°C In this embodiment, the specific steps of step S25 are: Perform dynamic production process behavior analysis on the optical flow motion trajectory of feature points to generate dynamic production process behavior features; Perform multi-stage process recognition on the optical flow motion trajectory of feature points to obtain multi-stage process data; Based on the dynamic production process behavior characteristics, the process data of multiple stages are mined for logical context to generate logical dependencies between the processes of each stage; The time-series state evolution data is used to dynamically map the logical dependencies of the processes at each stage and build a dynamic production twin model.

[0045] In this embodiment, motion features, including speed, acceleration, and motion direction, are extracted from the optical flow motion trajectory of the feature point. The extracted motion features are used to analyze the dynamic changes of the production process behavior. Statistical analysis methods (such as mean and standard deviation) are used to describe the behavior features. Time series analysis (such as sliding window analysis) is applied to identify the periodicity and trend changes of the process behavior. The generated dynamic production process behavior features are stored in a data frame to ensure that each behavior feature is connected to the corresponding timestamp and process status. Different process stages are divided according to the characteristics of the optical flow motion trajectory, such as mixing, heating, cooling, etc. The key data of each stage is recorded, including the spatial position and motion status of the feature points, so as to form a multi-stage process data set. The motion trajectory is analyzed using a clustering algorithm (such as K-means and DBSCAN). Dynamically identify different process stages, analyze the logical relationships between processes, such as forward and backward dependencies, conditional relationships, etc., based on dynamic production process behavior characteristics and process data of multiple stages, build a logical dependency diagram between processes, clarify the execution order and conditions of each process, apply association rule mining technology (such as Apriori algorithm) or graph theory analysis method, extract the logical dependencies between processes, dynamically map the logical dependencies of processes at each stage based on time series state evolution data, analyze the changes in parameters during process execution, integrate the mapping results, build a dynamic production twin model, reflect the processes in the production process and their relationships, use system modeling tools (such as Simulink) or custom modeling frameworks, combine process logical relationships and dynamic parameter mapping results, and establish a comprehensive dynamic production twin model.

[0046] Calculate the optical flow trajectory of the feature points, the data is as follows: Point IDFrameΔXΔY 1100 1255 1355 2100 22-5-5 23-5-5 Calculate the velocity and acceleration of each feature point.

[0047] Velocity = ΔX / Δt, assuming Δt = 1 second, then acceleration = ΔV / Δt.

[0048] Example calculation: For Point ID 1 (Frame 1 to Frame 2): Speed ​​= √(ΔX² + ΔY²) = √(5² + 5²) = √50 ≈ 7.07 Acceleration = (change in velocity) / Δt = (7.07 - 0) / 1 ≈ 7.07 Combine the velocity and acceleration data of feature points into a new dataset.

[0049] Point IDFrameSpeedAcceleration 127.077.07 137.070 227.077.07 237.070 Define the stages of the process based on dynamic behavior characteristics, for example: Stage 1: Speed ​​less than 5 (stationary or slow moving) Stage 2: Speed ​​between 5 and 10 (normal movement) Stage 3: Speed ​​greater than 10 (fast movement) Assign a process stage to each feature point based on speed data.

[0050] Point IDFrameSpeedStage 127.07Stage 2 137.07Stage 2 227.07Stage 2 237.07Stage 2 If process A (such as Stage 1) usually occurs before process B (such as Stage 2), it is recorded as a logical dependency.

[0051] Collect relevant data, such as temperature, pressure, etc., and record them as follows: TimeTemperaturePressure 170°C1.5 MPa 272°C1.6 MPa 375°C1.7 MPa Map temporal state data to logical dependencies.

[0052] Use a linear regression model or other regression algorithm to combine the process stage data with the time series status data.

[0053] Set up the model: Pressure = a * Stage + b * Temperature + c Fit the model through historical data and obtain the parameters (assuming a=0.5, b=0.3, c=1.0).

[0054] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Step S31: Obtaining a preset production standard requirement log; Step S32: performing phased demand identification on the preset production standard demand log, and extracting demand feature data of multiple stages; Step S33: Performing dynamic process deviation detection on the dynamic production twin model based on the demand feature data of multiple stages to generate staged production process deviation data; Step S34: Perform production failure analysis on the staged production process deviation data to obtain staged production failure data. In this embodiment, preset production standard demand logs are extracted from the production management system or database. These logs usually contain various standard parameters and requirements in the production process. The extracted data format is ensured to be unified, such as timestamp, parameter name, target value, etc., to facilitate subsequent analysis. The demand logs are extracted from the database using a data query tool (such as SQL), or real-time data is obtained through an API interface. The obtained production standard demand logs are analyzed to identify different stages in the production process (such as preparation, production, inspection, etc.). Demand feature data, including key parameters, target values, timestamps, etc., are extracted from each stage to form demand feature data sets for multiple stages. State machine models or logical rules are applied in combination with time series analysis methods to identify and classify demand features of each stage. The extracted demand feature data of multiple stages are compared with the actual operation data in the dynamic production twin model to identify process deviations. The process deviations of each stage, such as the difference between the actual parameters and the target parameters, are calculated to generate staged production process deviation data. Deviation analysis methods, such as control charts, are used. Chart) or statistical process control (SPC), identify and quantify process deviations, conduct in-depth analysis of staged production process deviation data, find out the causes of production failures, such as deviations exceeding the preset range, frequent fluctuations, etc., associate the identified causes of failures with process deviations, generate staged production failure data, record the failure type, occurrence time and impact range, apply fault tree analysis (FTA) or root cause analysis (RCA) methods, conduct systematic analysis of deviation data, and identify potential sources of failures.

[0055] In this embodiment, step S4 includes the following steps: Step S41: performing fault type analysis on the staged production fault data to extract production fault type data; Step S42: tracing and inferring potential fault sources on the production fault type data to obtain potential nodes of abnormal faults; Step S43: Based on the abnormal fault potential nodes, the dynamic production twin model is marked for potential fault location to obtain a potential fault marking twin model.

[0056] In this embodiment, a statistical analysis method is used to summarize and classify the fault data, identify the frequency and characteristics of various faults, extract key fault type data, such as equipment failure, process deviation, human error, etc., and apply data visualization tools (such as Tableau, Power BI) generates a histogram or pie chart of fault types to facilitate identification of major fault types. Use Python for data analysis and the Pandas library for data grouping and statistics. Analyze the relationship between various faults and production parameters through fault type data, identify the influencing factors of potential faults, use correlation analysis to determine which parameter changes lead to specific types of faults, infer potential fault nodes based on the identified fault types and correlations, such as the specific components or production links of equipment failures, and record the characteristics of potential fault nodes for subsequent analysis. Use statistical analysis tools (such as SPSS and R) for correlation and regression analysis, apply machine learning algorithms (such as decision trees and random forests) for fault inference, identify potential fault nodes, and integrate these node information into the dynamic production twin model based on the identified abnormal fault potential nodes. Mark the key parameters in the dynamic model to indicate which parameters are related to potential faults, generate a twin model of potential fault markers, clearly show the relationship between fault nodes and production processes, and record the time series changes of potential fault markers for subsequent monitoring and analysis. Use modeling tools (such as MATLAB) Simulink, AnyLogic) to build and update the dynamic production twin model, apply a graphical interface in the model, clearly mark potential fault nodes, and ensure real-time monitoring and feedback.

[0057] In this embodiment, the specific steps of step S5 are: Step S51: performing upstream and downstream correlation analysis on the potential nodes of abnormal faults to generate upstream and downstream correlation logic of the fault points; Step S52: Based on the upstream and downstream association logic of the fault point, the potential fault mark twin model is subjected to fault diffusion path evolution, thereby obtaining multiple potential fault diffusion paths; Step S53: simulating the evolution of fault propagation on multiple potential fault diffusion paths to obtain the law of fault propagation evolution; Step S54: predicting the fault impact range of the lubricating oil production line based on the fault propagation evolution law to generate fault impact range data. In this embodiment, relevant data of potential nodes of abnormal faults are collected, including production parameters, equipment status and production line configuration when the fault occurs, upstream and downstream equipment and process information related to the fault point are sorted out, the logical relationship between the fault point and its upstream and downstream links is analyzed, which links are directly affected by the fault point, and the upstream and downstream association logic diagram of the fault point is generated. The causal relationship between the fault point and the associated links is clarified, and a graph analysis tool (such as Gephi) is used to construct a fault point association diagram. A system engineering method is adopted to combine expert knowledge and historical data to deduce the upstream and downstream relationship. Based on the upstream and downstream association logic of the fault point, the potential fault diffusion path is identified, that is, the order of production links affected by the fault, and the key links of each path are determined. key nodes, record the propagation direction and impact range of potential faults, integrate the identified potential fault diffusion paths into the potential fault tag twin model, form a dynamic fault diffusion path evolution model, use system dynamic modeling tools (such as Vensim, Stella) to build the path model, combine historical fault cases to verify the path, ensure the accuracy of the model, determine the parameters of fault propagation simulation, including propagation speed, impact intensity and time window, set simulation conditions based on historical fault data and expert experience, run fault propagation evolution simulation, observe the diffusion process of potential faults on the production line, record simulation results, analyze the evolution law of fault propagation, use simulation software (such as MATLAB Simulink, AnyLogic) to simulate fault propagation, generate fault propagation paths and impact data under different situations through Monte Carlo simulation method, analyze the specific impact range of the fault on each link of the production line based on the results of fault propagation evolution simulation, extract impact range data, identify the affected equipment, process and production parameters, generate data report on the fault impact range, clearly display the consequences and impact range caused by the fault, provide corresponding improvement suggestions and preventive measures to facilitate decision-making by production management, use data visualization tools (such as Tableau, Power BI) to generate a visualization report on the impact range, combine with statistical analysis tools to analyze the frequency and severity of the fault impact, and form a quantitative impact assessment.

[0058] According to the production topology, the upstream and downstream equipment of the liquid level sensor include pump A and heater B.

[0059] Starting with P=30%, the downstream propagation probability is given by a heuristic algorithm: AReceives = 60% BReceives=40% Establish the SIR model and iterate 10 steps: A status change: [susceptible->infected] times: 6 B status change: [susceptible->infected] times: 4 Weighted calculation based on the impact coefficient of equipment failure A and B on production capacity (A: 30%, B: 20%): When date T1: P=30%, the estimated impact range is 30%×0.6+20%×0.4=24%.

[0060] In this embodiment, the specific steps of step S6 are: Step S61: Performing a dynamic risk assessment on the fault impact range data to generate a fault risk assessment value; Step S62: performing iterative production parameter regulation based on the fault risk assessment value to generate multiple production parameter regulation strategies; Step S63: performing production simulation evolution on multiple production parameter control strategies to generate multiple production parameter control simulation data; Step S64: performing production optimization evaluation on multiple production parameter control simulation data to generate an evaluation value for each control strategy; Step S65: extracting the optimal production parameter control strategy based on the evaluation value of each control strategy; Step S66: Dynamically optimize the production of the lubricating oil production line according to the optimal production parameter control strategy, and build a fault monitoring optimization model.

[0061] In this embodiment, data on the scope of fault impact is collected, including information such as fault type, impact degree, and impact link, to ensure data integrity, remove redundant and irrelevant data points, use risk assessment models such as fault tree analysis (FTA) or fuzzy logic assessment, analyze the potential risks of faults to the production line, calculate fault risk assessment values, consider the probability of fault occurrence and its impact degree, use statistical analysis tools (such as SciPy in R and Python) to perform risk assessment calculations, combine historical data and expert judgment, apply a weight model to quantify risks, and formulate preliminary production parameter control strategies based on the fault risk assessment values, such as adjusting parameters such as temperature, pressure, and flow rate, and set control strategy goals, such as reducing risks. , improve efficiency, etc., iterate the control strategy through simulation and feedback mechanism, gradually optimize the parameter settings, record the parameter changes of each iteration and their corresponding risk assessment values, use optimization algorithms (such as genetic algorithms, particle swarm optimization) to iterate the strategy, perform automatic calculations of parameter control in Excel or Python, build a production simulation model according to the control strategy, determine the input parameters and operating conditions of the model, ensure that the simulation model can reflect the actual operation of the production line, run simulations of multiple production parameter control strategies, record the production data under each strategy, generate multiple production parameter control simulation data, analyze the effects of different strategies, and use simulation software (such as AnyLogic, MATLAB Simulink) for production simulation, improve the reliability and accuracy of the results through Monte Carlo simulation method, determine the key indicators for evaluating each control strategy, such as production efficiency, resource utilization, failure rate, etc., set the weight of each indicator for comprehensive evaluation, analyze multiple production parameter control simulation data, calculate the evaluation value of each control strategy, generate evaluation report, compare the effects of different strategies, use multi-index decision analysis (such as TOPSIS, AHP) method for comprehensive evaluation, generate visual charts to facilitate intuitive comparison of various strategies, screen out the best performing strategies based on the evaluation value of each control strategy, set the selection criteria for the optimal strategy, such as minimum risk, maximum efficiency, etc., confirm the optimal production parameter control strategy, record its specific parameter settings and expected effects, and use data analysis tools (such as Python Pandas) to evaluate and compare strategies, generate a final report, clarify the parameter settings of the optimal strategy, adjust the strategy according to the optimal production parameters, update the dynamic production optimization model of the lubricant production line, integrate fault monitoring functions, ensure real-time monitoring of the production line status, implement the optimal strategy in the actual production environment, observe its impact on production efficiency and failure rate, continuously monitor and verify the effectiveness of the model, adjust the strategy in time to adapt to changes, use system modeling tools (such as Simulink, Arena) to build the optimization model, and combine it with the real-time monitoring system (such as SCADA) for dynamic adjustment and feedback.

[0062] The correlation coefficient between the production line blockage rate P and the liquid level sensor data H is 0.8 Upstream and downstream equipment: Pump A (influence α=0.3) and Heater B (α=0.2) When P=20%, AReceives=65%, BReceives=35% Establish the SIR model and iterate 10 steps: A state changes: [2,5] times B state changes: [1,3] times When date=D1,P=20%, the impact range is: A=(0.3×0.65×2) / 10+B=(0.2×0.35×1) / 10=0.19 The risk score is R = P × impact range = 20% × 0.19 = 3.8 Design 1: Filter pressure increased by 10% Option 2: Increase pressure by 15% Solution 1 - 5% capacity loss, 2% quality improvement Solution 2 - 8% capacity loss, 3% quality improvement Evaluation function V = production capacity × 0.6 + quality × 0.4 Solution 1 V = 0.95 × 0.6 + 1.02 × 0.4 = 0.976 Solution 2 V = 0.92 × 0.6 + 1.03 × 0.4 = 0.972 This leads to the optimal strategy plan 1.

[0063] In this embodiment, a fault monitoring and analysis system for lubricating oil production is provided, which is used to execute the fault monitoring and analysis method for lubricating oil production as described above, including: The time series evolution module is used to obtain multi-dimensional work monitoring parameters and all-round work images; perform abnormal outlier filtering on the multi-dimensional work monitoring parameters, and perform time series evolution simulation to generate time series state evolution data; The twin model module is used to perform continuous frame optical flow motion analysis on omnidirectional working images, and to perform dynamic parameter mapping based on time-series state evolution data to build a dynamic production twin model; The fault analysis module is used to analyze production faults on the dynamic production twin model to obtain staged production fault data; The fault tracing module is used to tracing and infer potential fault sources based on periodic production fault data and to build a potential fault marking twin model; The fault propagation evolution module is used to simulate the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; The production parameter control module dynamically optimizes the lubricant production line based on the fault impact range data and builds a fault monitoring optimization model.

[0064] The present invention realizes comprehensive monitoring and analysis of working status by acquiring multi-dimensional working monitoring parameters and all-round working images. Abnormal outlier filtering processing and time series evolution simulation help identify abnormal situations in work and generate meaningful time series state evolution data to provide basic information for subsequent analysis. Through continuous frame optical flow motion analysis and dynamic parameter mapping, a dynamic production twin model is constructed to realize real-time monitoring and prediction of the production process. Combined with the time series state evolution data, this module helps simulate the evolution of working status, discovers potential problems in advance and predicts them, thereby improving production efficiency and quality. Dynamic process deviation detection and production fault analysis are performed based on the dynamic production twin model to identify abnormal situations in production. By analyzing the preset production standard requirement log, process deviations are detected in time, fault analysis is performed and phased results are obtained. Production fault data provides a basis for subsequent fault handling. Through potential fault tracing and identification of potential nodes of abnormal faults, potential problems can be quickly located and solved. A potential fault marking twin model is constructed to help predict the occurrence and propagation path of faults, and measures can be taken in advance to reduce the impact of faults on production. Through fault propagation evolution simulation and fault impact range prediction, the production department can be helped to respond to problems in a timely manner. Fault impact range data is generated to guide production parameter regulation, reduce the negative impact of faults on the entire production process, and improve production efficiency and stability. Through iterative production parameter regulation and extraction of optimal production parameter regulation strategies, the production process is optimized. According to the optimal production parameter regulation strategy, the lubricant production line is dynamically optimized, and a fault monitoring optimization model is constructed to sustainably improve production efficiency and product quality.

[0065] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0066] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A fault monitoring and analysis method for lubricating oil production, characterized in that: The following steps are involved: Step S1: acquiring multi-dimensional working monitoring parameters and omnidirectional working images; performing time-series evolution simulation on the multi-dimensional working monitoring parameters to generate time-series state evolution data; Step S2: Perform continuous frame optical flow motion analysis on the omnidirectional working image, and perform dynamic parameter mapping based on the time series state evolution data to build a dynamic production twin model; Step S3: Perform production failure analysis on the dynamic production twin model to obtain staged production failure data; Step S4: Conduct potential fault tracing and speculation on the staged production fault data, and build a potential fault marking twin model; Step S5: simulating the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; Step S6: Dynamically optimize the production of the lubricating oil production line based on the fault impact range data and build a fault monitoring optimization model.

2. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: Performing all-round work monitoring of the lubricating oil production line based on multiple sensors, extracting multi-dimensional work monitoring parameters and all-round work images; Step S12: performing abnormal outlier filtering on the multi-dimensional working monitoring parameters to generate consistent monitoring parameters; Step S13: performing time series parameter change analysis on the consistent monitoring parameters to generate time series monitoring parameter change data; Step S14: perform deep correlation mining on multi-dimensional work monitoring parameters to identify correlation rules between parameters; Step S15: Based on the correlation rules between parameters, a time series evolution simulation is performed on the time series monitoring parameter change data to generate time series state evolution data.

3. The fault monitoring and analysis method for lubricating oil production according to claim 2, characterized in that: The specific steps of step S12 are: Define standard sampling timestamp nodes; Performing time synchronization processing on multi-dimensional work monitoring parameters based on standard sampling timestamp nodes to generate time-synchronized work monitoring parameters; Perform abnormal outlier data detection on time synchronization work monitoring parameters and mark abnormal outlier data points; Performing abnormal outlier filtering based on abnormal outlier data points to generate outlier filtering monitoring parameters; The outlier filtering monitoring parameters are processed by parameter scale consistency to generate consistent monitoring parameters.

4. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: performing multi-frame sampling on the omnidirectional working image to obtain multiple image frame sequences; Step S22: visually identifying key points of the production line on multiple image frame sequences and marking characteristic points of the production line; Step S23: Calculating the spatial position change of the feature points of the production line to obtain the spatial position coordinates of the feature points at different times; Step S24: performing optical flow motion analysis of consecutive frames based on the spatial position coordinates of the feature points at different times to generate optical flow motion trajectories of the feature points; Step S25: Dynamically map the optical flow motion trajectory of the feature points based on the time series state evolution data to build a dynamic production twin model.

5. The fault monitoring and analysis method for lubricating oil production according to claim 4, characterized in that: The specific steps of step S25 are: Perform dynamic production process behavior analysis on the optical flow motion trajectory of feature points to generate dynamic production process behavior features; Perform multi-stage process recognition on the optical flow motion trajectory of feature points to obtain multi-stage process data; Based on the dynamic production process behavior characteristics, the process data of multiple stages are mined for logical context to generate logical dependencies between the processes of each stage; The time-series state evolution data is used to dynamically map the logical dependencies of the processes at each stage and build a dynamic production twin model.

6. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S3 are: Step S31: Obtaining a preset production standard requirement log; Step S32: performing phased demand identification on the preset production standard demand log, and extracting demand feature data of multiple stages; Step S33: Performing dynamic process deviation detection on the dynamic production twin model based on the demand feature data of multiple stages to generate staged production process deviation data; Step S34: Perform production failure analysis on the staged production process deviation data to obtain staged production failure data.

7. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S4 are: Step S41: performing fault type analysis on the staged production fault data to extract production fault type data; Step S42: tracing and inferring potential fault sources on the production fault type data to obtain potential nodes of abnormal faults; Step S43: Based on the abnormal fault potential nodes, the dynamic production twin model is marked for potential fault location to obtain a potential fault marking twin model.

8. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S5 are: Step S51: performing upstream and downstream correlation analysis on the potential nodes of abnormal faults to generate upstream and downstream correlation logic of the fault points; Step S52: Based on the upstream and downstream association logic of the fault point, the potential fault mark twin model is subjected to fault diffusion path evolution, thereby obtaining multiple potential fault diffusion paths; Step S53: simulating the evolution of fault propagation on multiple potential fault diffusion paths to obtain the law of fault propagation evolution; Step S54: predicting the fault impact range of the lubricating oil production line based on the fault propagation evolution law to generate fault impact range data.

9. The fault monitoring and analysis method for lubricating oil production according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: Performing a dynamic risk assessment on the fault impact range data to generate a fault risk assessment value; Step S62: performing iterative production parameter regulation based on the fault risk assessment value to generate multiple production parameter regulation strategies; Step S63: performing production simulation evolution on multiple production parameter control strategies to generate multiple production parameter control simulation data; Step S64: performing production optimization evaluation on multiple production parameter control simulation data to generate an evaluation value for each control strategy; Step S65: extracting the optimal production parameter control strategy based on the evaluation value of each control strategy; Step S66: Dynamically optimize the production of the lubricating oil production line according to the optimal production parameter control strategy, and build a fault monitoring optimization model.

10. A fault monitoring and analysis system for lubricating oil production, characterized in that: The method for performing the fault monitoring and analysis method for lubricating oil production according to claim 1 comprises: The time series evolution module is used to obtain multi-dimensional work monitoring parameters and all-round work images; perform abnormal outlier filtering on the multi-dimensional work monitoring parameters, and perform time series evolution simulation to generate time series state evolution data; The twin model module is used to perform continuous frame optical flow motion analysis on omnidirectional working images, and to perform dynamic parameter mapping based on time-series state evolution data to build a dynamic production twin model; The fault analysis module is used to analyze production faults on the dynamic production twin model to obtain staged production fault data; The fault tracing module is used to tracing and infer potential fault sources based on periodic production fault data and to build a potential fault marking twin model; The fault propagation evolution module is used to simulate the fault propagation evolution of the potential fault marker twin model to generate fault impact range data; The production parameter control module dynamically optimizes the lubricant production line based on the fault impact range data and builds a fault monitoring optimization model.

Images