Intelligent prediction and data traceability management method for lubrication system

Abstract

This invention provides an intelligent prediction and data traceability management method for lubrication systems. It acquires multi-source operational data throughout the entire lifecycle of the lubricating medium, including its input, operation, and disposal stages, and establishes cross-stage data associations based on unique batch identifiers and spatiotemporal metadata, forming a traceable operational data chain. The method further quantifies the dynamic correlation characteristics between equipment operating efficiency, lubricating medium status, and environmental conditions through a multi-dimensional correlation analysis module, and generates a state evolution decision model using a knowledge rule base. Based on this model, the system calls a risk assessment knowledge base to generate oil replacement prediction suggestions, component risk priority lists, and maintenance strategy reports. Maintenance results are fed back to the operational data chain, and the knowledge base's adaptive calibration mechanism drives decision model updates and strategy optimization, achieving closed-loop optimization of intelligent lubrication management and effectively improving fault tracing efficiency, prediction reliability, and maintenance cost-effectiveness.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment lubrication system management technology, and in particular to an intelligent prediction and data traceability management method for lubrication systems. Background Technology

[0002] In industrial production lines, ensuring the efficient operation of lubrication systems is a core management objective for maintaining the stability and extending the service life of critical moving components. Achieving accurate predictive maintenance planning and comprehensive lifecycle data traceability management constitutes a key path to improving system reliability and optimizing management decision-making efficiency. By integrating multi-source operational data, a solid basis can be provided for developing effective predictive maintenance plans, improving the reliability of condition prediction and data traceability, optimizing supply chain and inventory management efficiency, and ultimately driving overall production and operational efficiency improvements.

[0003] However, existing methods primarily focus on localized monitoring during the lubricant usage phase, lacking comprehensive lifecycle data management for lubricants from input and use to disposal. When a failure occurs, root cause analysis becomes difficult, and cross-stage impact analysis is lacking. Furthermore, existing methods generally rely on pre-set and rigid management rules. The system struggles to dynamically adjust and continuously optimize maintenance strategies based on equipment status evolution, new operational data, and dynamically changing operating conditions.

[0004] To address this, a method for intelligent prediction and data traceability management of lubrication systems is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent prediction and data traceability management method for lubrication systems, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent prediction and data traceability management method for lubrication systems, the method comprising:

[0007] Acquire multi-source operational data of the lubricating medium throughout its entire life cycle, and establish cross-stage data associations based on unique batch identifiers and spatiotemporal metadata to form a traceable operational data chain;

[0008] Based on the multi-source operation data, the equipment operation efficiency index, lubrication medium status index and environmental condition index are extracted. Combined with oil chemical deterioration parameters and real-time contamination data, the dynamic correlation characteristics between them are quantified through a multi-dimensional correlation analysis module. Combining the dynamic correlation characteristics with the system health assessment results, a state evolution decision model is generated based on a preset knowledge rule base.

[0009] Based on the state evolution decision model, a preset risk assessment knowledge base is invoked to generate oil replacement prediction suggestions, component risk priority lists, and maintenance strategy reports. Operations are performed according to the maintenance strategy reports, and the maintenance results are fed back to the operation data chain. The decision model is updated and the strategy is optimized through the knowledge base adaptive calibration mechanism.

[0010] Preferably, the full lifecycle multi-source operational data includes:

[0011] During the deployment phase, obtain the unique batch identifier of the lubricating medium, the process parameters at the time of actual filling, and the initial performance indicators of the lubricating medium;

[0012] During operation, a network of multiple types of sensors deployed on key components and data acquisition points of the equipment is used to collect equipment operating efficiency indicators, lubrication medium status indicators, and environmental condition indicators in near real-time. The deployment of the sensors should follow the principles of key points, fault-prone points, and points that are difficult to detect manually. For example, for rotating machinery, vibration sensors can be installed in three directions: vertical, horizontal, and axial, while temperature sensors can be installed near bearing seats and data acquisition points.

[0013] During the disposal phase, key deterioration parameters of the waste oil and wear status data of components are recorded.

[0014] Preferably, the steps for establishing cross-stage data relationships to form a traceable operational data chain include:

[0015] Using the unique batch identifier of the lubricating medium as the core traceability key, the system logically correlates the operational performance of the batch of lubricating medium in different equipment and time periods, all related abnormal events, and even the final scrap analysis and the resulting component wear data, ensuring the vertical continuity of the data flow. By adding spatiotemporal metadata to all multi-source operational data, a correlation index is established between environmental impact intensity data and oil deterioration characteristic data, thereby analyzing the specific impact of environmental factors on the deterioration rate of lubricating medium in time periods and spatial regions, in order to identify the impact mechanism of the external environment on the health of the lubrication system.

[0016] Preferably, the construction of the state evolution decision model includes the following steps:

[0017] The generation of update rules for the decision model specifically includes matching the dynamic correlation features identified by the multi-dimensional correlation analysis module based on the physical connection topology of the equipment with the failure correlation rules stored in the preset knowledge rule base, and thereby activating and updating the weights and activation conditions of the corresponding paths in the model. The triggering of the update of the decision model specifically includes immediately prompting maintenance actions and simultaneously initiating the knowledge rule base calibration process when the oil contamination level exceeds the preset warning threshold. The calibration process involves fine-tuning the rule weights and re-evaluating the triggering conditions, and is subject to data verification and manual confirmation.

[0018] Preferably, the risk assessment knowledge base is executed, and the specific steps include:

[0019] Based on nodes that meet preset failure characteristic conditions in the state evolution decision model, risk sources are identified; according to the physical connection topology of the equipment, a transmission path template is retrieved from the association knowledge rule base. The template describes in detail how a fault propagates from a risk source to other related components through physical connections and predicts its potential impact range; path association weight factors, node impact factors, and environmental adaptation correction factors are loaded; based on the factor superposition value, a risk level assessment is generated through a comprehensive risk assessment algorithm. The multi-model combination method includes, but is not limited to, weighted summation, fuzzy comprehensive evaluation, and Bayesian network inference, and a detailed risk report is output. The report includes the distribution of risk sources, high-risk path identification, and a maintenance priority list sorted according to risk level and impact degree.

[0020] Preferably, the maintenance strategy report generation includes:

[0021] The oil replacement prediction suggestion is based on the chemical degradation parameters and real-time contamination data of the lubricating medium. It calls a preset replacement knowledge rule base to match the equipment load conditions, combining the actual degradation degree of the lubricating medium with the actual operating load of the equipment to predict the remaining effective life of the lubricating medium and provide optimized replacement timing suggestions. The component risk priority list clearly marks the paths with risk levels exceeding preset thresholds and high-risk indicators in a visual form. The maintenance strategy report also includes specific maintenance suggestions, covering recommended lubricating medium models, detailed replacement and maintenance steps, and preventive measures.

[0022] Preferably, the decision model update and strategy optimization include:

[0023] The system collects equipment status recovery data within a preset period after maintenance, including whether the data collection point and component operating performance indicators have recovered to the healthy baseline. It compares the differences in data characteristics before and after maintenance to identify the actual impact of maintenance operations on equipment status and lubrication medium performance, and dynamically adjusts the weights of correlation factors in the knowledge rule base using an iterative optimization algorithm based on error feedback. This iterative optimization algorithm can employ, but is not limited to, gradient descent, genetic algorithms, and reinforcement learning algorithms to more accurately reflect actual causal relationships. When the prediction errors for lubrication medium replacement cycle and component failure exceed allowable deviations, a knowledge rule base parameter revision process is triggered. This process includes adjusting the parameters of the multi-dimensional correlation analysis module, updating the thresholds of failure correlation rules, and reconstructing model parameters to achieve continuous optimization of the entire intelligent lubrication management system.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] This invention uses a unique batch number throughout the entire lifecycle of the lubricating medium as the core traceability key. By linking the lubricating medium's multi-source operational data during the input, operation, and disposal stages through this unique batch number, a traceable data chain covering the entire lifecycle is constructed. When equipment malfunctions, maintenance personnel can quickly trace the historical usage trajectory of the problematic batch of lubricating medium, accurately obtaining its operating status, anomaly records, and deterioration process in multiple equipment environments.

[0026] Compared to traditional methods that rely on static rules, the state evolution decision model uses a multi-dimensional analysis module to deeply mine the correlation features between multi-source operating data. By integrating dynamic correlation features into the equipment state model, the system can accurately characterize the real-time health status and degradation trend of the equipment. This invention significantly improves the adaptability to complex production environments, effectively reduces the false alarm rate caused by environmental fluctuations, and enhances the practical value of state prediction.

[0027] The system integrates a state model and a risk assessment knowledge base to accurately identify high-risk components and potential fault propagation paths; it generates detailed risk reports and outputs customized maintenance strategies through a multi-factor risk quantification assessment mechanism; and it optimizes the priority of maintenance resource allocation based on risk levels, enabling maintenance teams to focus on high-consequence risk points and avoid unnecessary inspections of healthy components; thereby reducing labor costs, spare parts consumption, and downtime losses, and improving overall maintenance efficiency.

[0028] By comparing and analyzing the restored equipment data with historical data after maintenance, the system continuously evaluates the maintenance effect. When the predicted results deviate from the actual state, the system dynamically adjusts the parameters of the analysis module, updates the fault judgment rules, and optimizes the model structure based on the adaptive algorithm. This closed-loop feedback mechanism ensures that the prediction model and knowledge rule base can continuously iterate and evolve, constantly improving the prediction accuracy, optimizing the maintenance strategy, and maintaining the efficient operation of the intelligent lubrication management system. Attached Figure Description

[0029] Figure 1 A flowchart of an intelligent prediction and data traceability management method for lubrication systems is provided as an embodiment of this invention.

[0030] Figure 2 This is a schematic diagram illustrating the construction of the runtime data chain proposed in an embodiment of this invention application;

[0031] Figure 3 This is a schematic diagram of the state evolution decision model proposed in an embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Please see Figures 1-3 The intelligent prediction and data traceability management method for lubrication systems involved in this invention has the following specific implementation steps:

[0034] Acquire multi-source operational data of the lubricating medium throughout its entire life cycle, and establish cross-stage data associations based on unique batch identifiers and spatiotemporal metadata to form a traceable operational data chain;

[0035] Based on the multi-source operational data, the equipment operating efficiency index, lubricating medium status index, and environmental condition index are extracted, combined with oil chemical degradation parameters and real-time contamination data. The dynamic correlation characteristics among the three are quantified through a multi-dimensional correlation analysis module. Combining the dynamic correlation characteristics with the system health assessment results, a state evolution decision model is generated based on a preset knowledge rule base.

[0036] Based on the state evolution decision model, a preset risk assessment knowledge base is invoked to generate oil replacement prediction suggestions, component risk priority lists, and maintenance strategy reports; operations are performed according to the maintenance strategy reports, and the maintenance results are fed back to the operation data chain, driving decision model updates and strategy optimization through the knowledge base adaptive calibration mechanism.

[0037] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0038] Example 1:

[0039] This embodiment describes a method for obtaining multi-source operational data of a lubricating medium throughout its entire life cycle, and establishes cross-stage data associations based on unique batch identifiers and spatiotemporal metadata to form a traceable operational data chain.

[0040] Specifically, during the lubricant introduction stage, the system collects the unique batch identifier of the lubricant and records the process parameters at the time of addition. These parameters include the addition volume, addition pressure, addition time, and addition point. At the same time, the initial performance indicators of the lubricant are recorded, including the viscosity, acid value, cleanliness level, and main chemical components of the new oil. When the lubricant is put into storage, the above information is bound to the batch number and recorded. During actual addition, the system records the environmental conditions of addition, such as temperature and humidity, and adds them to the initial information. To achieve the correlation of multi-source operating data, this method uses a relational database for storage. By establishing core data tables for batch information, addition events, and initial performance data, and using the batch number as the primary key for correlation, the traceability of the data is ensured.

[0041] During the operation phase, this method uses a multi-type sensor network deployed on key components and data acquisition points of the equipment to collect equipment operating performance indicators, lubrication medium status indicators, and environmental condition indicators in near real-time. Key points refer to the core transmission components or high-load-bearing components of the equipment, such as main bearings, gearboxes, and pumps. Fault-prone points refer to locations where historical fault records frequently occur or known stress concentration areas. Points that are difficult to detect manually refer to dangerous areas such as complex structures, confined spaces, or high temperatures and pressures.

[0042] Specifically, the sensor network includes an online laser particle counter, an online capacitive moisture sensor, a triaxial accelerometer, a resistance temperature detector, and a temperature and humidity composite sensor, etc. All sensor data transmission and edge computing mainly adopt the Modbus TCP / IP protocol. For vibration data that requires high-speed transmission, the system considers using the EtherCAT or OPCUA protocol.

[0043] The system employs an industrial-grade edge computing gateway equipped with a multi-core processor, at least 8GB of RAM, and 128GB of SSD storage, running an embedded Linux operating system, to effectively transmit data from the sensor network. The gateway's built-in data preprocessing module performs timestamp calibration, unit conversion, preliminary filtering, and outlier detection on the raw data. The processed data is then uploaded to the cloud platform in JSON or Protobuf format, ensuring standardized and universal data transmission. Data acquisition frequency varies depending on the data type: oil parameters are typically acquired once per minute; temperature, humidity, and power consumption data are acquired every 5 to 10 seconds; vibration data is sampled at high frequencies as needed, and feature extraction is performed at the edge before uploading the feature values ​​at a lower frequency. Specific transmission protocol configurations include Modbus TCP / IP with a polling period of 1 second in master-slave mode; EtherCAT for microsecond-level synchronization of vibration data; and OPCUA for a unified data model and secure communication.

[0044] The gateway's built-in data preprocessing module uses a moving average filtering algorithm with a window size of 5 for initial filtering. Outlier detection is performed using a Z-score-based method with a threshold set to 3. After ensuring stable data transmission, the system acquires vibration characteristic spectra (such as root mean square value, peak value, and kurtosis) via accelerometers; operating temperature is acquired via thermocouples and infrared sensors; and power consumption data is obtained via current transformers and power meters. Lubricating medium state parameters are acquired through methods including laser photoresist technology for particle counting; capacitance and infrared spectroscopy absorption methods for moisture content; and a dielectric constant sensor for dielectric constant. Environmental monitoring sensors collect environmental condition indicators including ambient temperature, humidity, corrosive gas concentration, and dust concentration.

[0045] The system records key degradation parameters and component wear status data during the waste oil treatment phase. Offline laboratory spectral analysis is used to determine the concentration of wear elements and additive depletion. Waste oil degradation parameters include changes in acid value and viscosity. Component wear status data includes non-invasive testing results, wear type identification after disassembly of faulty components, and macroscopic wear assessment. Multi-source operational data includes precise spatiotemporal metadata, such as timestamps, equipment unique identifiers, component unique identifiers, and spatial location information of data acquisition points; rich and multi-source operational data ensures data traceability and relevance.

[0046] This method uses the unique batch identifier of the lubricating medium as the core traceability key. Based on this batch identifier, the operational performance of the batch of lubricating medium in different equipment and time periods, all related abnormal events, and the final scrapping analysis are logically correlated. The correlation of multi-source operational data is stored in a relational database, with the batch number as the primary key for association.

[0047] Based on spatiotemporal metadata of multi-source operational data, a correlation index is established between environmental impact intensity data and oil deterioration characteristic data. The system identifies the potential impact mechanisms of the external environment on lubrication health by analyzing the specific impact of environmental factors on the deterioration rate of lubricating media under specific time periods and spatial regions, i.e., under specific equipment and data collection points.

[0048] Based on the traceable operational data chain, this method enters the intelligent analysis and prediction stage, and the specific implementation is as follows:

[0049] First, the system extracts equipment operating efficiency indicators, lubricating medium status indicators, and environmental condition indicators from the data chain, as well as oil chemical degradation parameters and real-time contamination data from the treatment stage. Then, the multi-dimensional correlation analysis module deeply mines the correlation between these multi-source data to quantify the dynamic correlation characteristics between equipment operating efficiency indicators, lubricating medium status indicators, and environmental condition indicators.

[0050] The multi-dimensional correlation analysis module employs a Long Short-Term Memory (LSTM) network to perform time-series analysis on sensor data within a sliding time window, predicting future component health indices or the rate of change of key lubrication medium parameters. It applies SMOTE technology to handle unbalanced fault samples; combines Bayesian networks for structure learning, incorporating known physicochemical mechanisms as prior knowledge through a PC algorithm to identify causal relationships between variables; and uses a Variational Autoencoder (VAE) to perform feature fusion and dimensionality reduction on multi-source operational data, mapping high-dimensional data to a low-dimensional latent space to provide optimized input for the LSTM model. A ridge regression model is used to quantify the degree of influence of equipment load on specific parameters such as bearing temperature. The state evolution decision model improves the model's predictive accuracy and fault analysis capabilities by inputting quantified dynamic correlation features.

[0051] Furthermore, the multi-dimensional correlation analysis module employs a time series analysis model based on long short-term memory networks to analyze and quantify the temporal relationships between multi-source operational data. Its input features include a sliding time window, such as raw sensor data from the past 24 hours, statistical features like mean and variance, and trend features like slope. The model's output layer predicts the component health index for the next 72 hours, which is a continuous value from 0 to 100, or the rate of change of key parameters of the lubrication medium. To address the imbalance problem of fault samples in historical data, the model training uses SMOTE oversampling technology to expand the minority class samples.

[0052] The model employs a Bayesian network-based structural learning algorithm, such as the PC algorithm, which combines known physicochemical mechanisms of lubrication system wear, oxidation, and contamination as prior knowledge to construct a causal graph. This graph is used to identify the causal tendencies between variables, such as the strength of the causal relationship between high temperature and the oxidation rate of the lubricating medium.

[0053] The model employs a variational autoencoder (VAE) to fuse features and reduce dimensionality from multi-source running data. The original high-dimensional features are mapped to a low-dimensional latent space, preserving key information. The dimensionality-reduced features can then be used as input to a long short-term memory (LSTM) network model.

[0054] The model uses a ridge regression model to represent the quantitative value of the effect of equipment load on bearing temperature by the magnitude of the coefficients, thus quantifying the degree of its influence.

[0055] Preferably, the Long Short-Term Memory (LSTM) network model in this embodiment typically contains 2 to 3 LSTM layers, each with 256 units; it employs the ReLU activation function and sets a Dropout rate of 0.2 to prevent overfitting. The model uses the Adam optimizer during training. The learning rate during training is 0.001; the batch size is 64; and the training lasts for 200 epochs. In data preprocessing, the model uses Min-Max normalization to scale the data to the range of 0 to 1. The SMOTE oversampling technique is used with a k value of 5 and a sampling ratio of 200% to effectively balance faulty samples. In the causal inference auxiliary model, the PC algorithm is implemented using a relevant open-source library with a significance level set to 0.05. The Variational Autoencoder (VAE) uses a neural network with two fully connected layers for both the encoder and decoder; the number of neurons in each layer is 128 and 64, respectively; the loss function combines reconstruction loss and a KL divergence term, where the KL divergence weight is 0.001. The regularization parameter for ridge regression is determined to be a preset value of 1.0 through cross-validation.

[0056] For example, when the equipment load continuously exceeds a certain threshold X% and the moisture content of the lubricating medium exceeds a certain threshold Yppm, the wear rate of specific components may show an accelerated trend; a certain increase in ambient temperature may lead to a certain increase in the oxidation rate of the lubricating medium. The multi-dimensional correlation analysis module integrates these quantified dynamic correlation characteristics with the preset equipment baseline and historical fault mode library to achieve real-time and refined assessment of the current health status of each component and the overall lubrication of the equipment.

[0057] The system collects long-term stable data of the equipment under normal operating conditions as the equipment baseline and automatically learns its normal parameter range using statistical analysis and anomaly detection algorithms to establish a preset equipment baseline. It also constructs a historical fault mode library by summarizing historical maintenance records, expert experience, and fault tree analysis. This library stores typical fault types and associated abnormal data characteristics and thresholds. Evaluation results are typically presented as a numerical health index and discrete health levels, including four levels: extremely high, high, medium, and low. These evaluation results are dynamically updated based on the latest data and changes in associated characteristics.

[0058] The state evolution decision model includes a system health assessment module. This module collects long-term stable data of the equipment under normal operating conditions and uses this data as the equipment baseline. The module applies statistical analysis and anomaly detection algorithms to construct a normal parameter range model. Through historical maintenance records, expert experience, and fault tree analysis, a historical fault mode library is built. This library stores typical fault types, along with associated abnormal data characteristics and thresholds. The system health assessment module combines quantified dynamic correlation characteristics, a preset equipment baseline, and the historical fault mode library. Ultimately, it achieves real-time and refined assessment of the current health status of each component and the overall lubrication of the equipment.

[0059] Furthermore, the system health assessment module, combining quantified dynamic correlation features, preset equipment baselines (including various parameter standards for new equipment operation), and a historical fault mode library, enables real-time and refined assessment of the current health status of each component and the overall lubrication of the equipment. The health index is generated using a support vector machine (SVM)-based classifier. This classifier takes multi-dimensional correlation features as input, such as the principal components of vibration signals, the rate of change in contamination, and the cumulative value of temperature anomalies. Simultaneously, it outputs a continuous health index from 0 to 100.

[0060] For example, the features reduced by PCA are input into a pre-trained SVM model, and the model's output value is the health index. The discrete health levels of the health index are divided as follows: 0 to 25 is extremely high risk (red warning); 26 to 50 is high risk (orange warning); 51 to 75 is medium risk (yellow warning); and 76 to 100 is low risk (green normal). The initial values ​​of the above thresholds are set based on historical failure data analysis and expert experience. The SVM classifier uses the RBF kernel function, with a regularization parameter C of 1.0 and a Gamma value of 0.1. The model is trained through supervised learning on historical data containing various states such as normal, slight wear, moderate contamination, and early failure, with the health index labeled 0 to 100. The training dataset extracts 30 key features, including vibration, temperature, and oil analysis, through feature engineering. The number of principal components in the PCA dimensionality reduction is selected to retain 95% of the variance. The threshold setting for health level is based on cluster analysis and statistical distribution of a large amount of historical fault data. For example, the health index is divided into four clusters through K-Means clustering. The threshold is determined by combining the experience judgment of domain experts on the actual equipment condition represented by different health index intervals, such as wear and tear, pollution level and risk tolerance. For example, a health index below 25 usually corresponds to the end of the component's life or a sign of an impending serious failure, requiring immediate dispatch of maintenance personnel to maintain the equipment.

[0061] This method combines quantitative dynamic correlation characteristics with health assessment results to generate a dynamic state evolution decision model based on a pre-set knowledge rule base. Specific steps include:

[0062] The preset knowledge rule base includes databases of failure association rules, risk assessment knowledge base, and replacement knowledge rule base. The state evolution decision model activates the failure rule matching mechanism based on conditions such as lubricating medium contamination in the knowledge rule base to achieve risk source identification and fault propagation path prediction. By combining rules in the knowledge rule base and real-time data, the state evolution decision model generates oil replacement prediction suggestions, component risk priority lists, and detailed maintenance strategy reports to achieve intelligent lubrication management and improve maintenance efficiency.

[0063] Furthermore, the model is determined to be a directed graph model. The nodes of the model represent equipment components such as bearings, gearboxes, and pumps. The quantification of node states is achieved through fuzzy logic reasoning, including the use of triangular membership functions; fuzzy rules based on the IF-THEN form; and the centroid method for defuzzification. Edges represent transition paths between states, with quantitative transition probabilities or influence strengths attached, such as the probability of transitioning from moderate contamination to early failure. The lubrication system exhibits multi-level health and abnormal states, such as normal operation, minor wear, moderate contamination, overheating warning, early failure, and severe failure. Thresholds can be adaptively adjusted through historical data analysis and expert experience, including dynamic calibration using moving averages or adaptive algorithms based on model prediction errors, ensuring the initial threshold settings are reasonable and effective.

[0064] After determining the model framework, new dynamic correlation features are identified through the model's multi-dimensional correlation analysis module. Simultaneously, these features are precisely matched with preset failure correlation rules using a knowledge rule base. The model's weights and activation conditions are updated using an iterative optimization algorithm based on Bayesian updates. The loss function is defined as the mean squared error between the predicted state and the actual post-maintenance state. The model calculates the gradient of the loss function with respect to the model weights and activation conditions and iteratively updates the parameters along the gradient direction to more accurately reflect the actual causal relationships. The failure correlation rules are mined using the Apriori algorithm, with a minimum support set to 0.1 and a minimum confidence set to 0.7. The Bayesian update algorithm updates the transition probability by calculating the posterior probability. Reinforcement learning can employ the Q-learning algorithm, defining the state as equipment health and lubrication medium status; the action as the weight adjustment magnitude; the reward function as the negative value of the prediction error; the learning rate as 0.01; and the discount factor as 0.9. The iterative adjustment algorithm based on historical data statistical frequency calculates the occurrence frequency of each rule using an exponentially weighted moving average and adjusts the weights accordingly.

[0065] The initial structure of the model is constructed based on expert knowledge, equipment fault tree analysis, and historical failure modes. Initial weights or probabilities are set according to expert experience or historical failure frequencies. The equipment physical connection topology refers to the actual mechanical, hydraulic, and lubrication pipeline connections and functional dependencies between the internal components of the equipment. Through the above operations, the initial state evolution decision model structure is ensured to be realistic and effective.

[0066] The knowledge rule base stores failure association rules, such as a lubrication pump malfunction causing distributor blockage, which in turn leads to insufficient oil supply to the data collection point, ultimately resulting in component wear. These rules are constructed through domain expert experience, historical fault data analysis, and combined with the equipment's physical connection topology model, using association rule mining algorithms or decision tree algorithms; the rules can be stored in the form of production rules or semantic web.

[0067] Failure characteristic conditions are defined as a combination of threshold values ​​or time series patterns of a series of indicators, such as oil contamination exceeding a certain threshold and vibration root mean square value exceeding another threshold, or temperature continuously rising by a specific degree over a specific number of consecutive days. When the multi-dimensional correlation analysis module identifies new and stronger dynamic correlation characteristics, the model matches these characteristics with failure correlation rules stored in the knowledge rule base. For example, if a new strong correlation is found between the continuous abnormal output pressure of the lubrication pump and the continuous zero reading of the flow sensor of a downstream distributor, the weights and activation conditions of the corresponding paths in the model will be updated after data verification and expert confirmation to better reflect the current actual operating state of the equipment. The weights are quantified as probability values ​​between 0 and 1, representing the possibility of state transition. Activation conditions are defined as a combination of threshold values ​​for a set of indicators. The updating of weights and activation conditions adopts Bayesian updates, reinforcement learning, and iterative adjustment algorithms based on historical data statistical frequencies.

[0068] This method employs a robust model update trigger mechanism. When the oil contamination level exceeds a preset warning threshold, maintenance actions are prompted. For example, when the particle count in the lubricating medium exceeds a specific level of ISO 4406 (e.g., 17 / 15 / 12) or the moisture content exceeds 200 ppm, the system immediately prompts maintenance and simultaneously initiates the model's knowledge rule base calibration process. The system considers abnormalities in key lubricating medium parameters in the equipment not only as an alarm signal but also as a strong model feedback learning signal. This signal prompts the evaluation and targeted adjustment of relevant rules in the knowledge rule base to optimize the model's real-time performance and predictive accuracy. Simultaneously, the knowledge rule base calibration includes fine-tuning rule weights and re-evaluating trigger conditions to prevent overlearning and model instability.

[0069] This method utilizes a pre-defined risk assessment knowledge base and generates a risk report based on a dynamically updated state evolution decision model. The report includes oil replacement prediction suggestions, a list of component risk priorities, and a maintenance strategy report. During the execution of the risk assessment knowledge base, risk sources are first identified based on nodes in the model that meet pre-defined failure characteristic conditions. These conditions are defined using historical data and domain knowledge to identify which components and data collection points have entered a high-risk state.

[0070] The risk assessment knowledge base retrieves transmission path templates from the associated knowledge rule base based on the physical connection topology of the equipment. Each path template details how a fault propagates from a risk source to other related components through physical connections. Based on these templates, the model can predict the potential impact range of a risk and its transmission path. The transmission path templates are presented in the form of a weighted directed graph; nodes represent components, edges represent physical connections and fault propagation paths, and the weights of the edges represent propagation probability or speed.

[0071] A weighted directed graph template for fault propagation is adopted; in the graph, nodes represent components, edges represent physical connections and fault propagation paths, and edge weights represent propagation probability or speed. After obtaining the template from the risk assessment knowledge base, the model loads path association weight factors, node influence factors, and environmental adaptation correction factors to conduct multi-dimensional risk overlay assessment, ensuring the accuracy and sufficiency of the risk assessment results.

[0072] Furthermore, the path association weighting factor directly adopts the probability value of state transition in the state evolution model, such as the transition probability from wear state to early failure state; the node influence factor is determined by combining the analytic hierarchy process (AHP) with expert scoring, and the evaluation criteria include downtime loss, maintenance cost, safety risk, and production impact, which are obtained based on equipment criticality assessment and expert scoring; the environmental adaptation correction factor is dynamically adjusted using a sigmoid nonlinear function. The environmental adaptation correction factor reflects the acceleration or delaying effect of environmental factors on the occurrence and development of failures, and is determined by fitting historical data. The comprehensive risk assessment algorithm adopts a weighted summation method. The assessment results are finally mapped to discrete risk levels, and the final risk level mapping rule is as follows: total risk value 0 to 0.2 is low risk; 0.2 to 0.5 is medium risk; 0.5 to 0.8 is high risk; and 0.8 to 1.0 is extremely high risk.

[0073] The model invokes a pre-defined replacement knowledge rule base and predicts the oil replacement cycle based on the chemical degradation parameters of the lubricating medium, such as acid value, viscosity change rate, and real-time contamination data. This replacement knowledge rule base includes, but is not limited to, rules recommending replacement within the next 30 days when the lubricating medium's acid value exceeds 0.4 mg KOH / g and its operating time exceeds 800 hours, while the equipment is under medium or high load conditions. The system dynamically matches these rules to the equipment's load conditions, predicts the remaining effective lifespan of the lubricating medium, and provides optimized replacement timing recommendations.

[0074] The model generates maintenance strategy reports based on assessments. When the risk assessment identifies a specific component, such as the main bearing, as having an extremely high risk and a wear-related failure type, the model retrieves and recommends the appropriate lubricant type, such as ISO VG68 synthetic lubricant, from a predefined maintenance knowledge base; it also provides detailed replacement and repair procedures. For example, Step 1: Drain the old oil; Step 2: Flush the system; Step 3: Add new oil to the specified level, accompanied by illustrated instructions. Furthermore, the maintenance strategy report provides preventative measures such as increasing the frequency of monthly oil analysis and checking seal wear. These recommendations are compiled and stored from expert experience, equipment manufacturer manuals, and historical maintenance records, and the model makes recommendations using a semantic matching algorithm.

[0075] Fault propagation paths obtained from equipment topology analysis can predict and determine the scope of risk impact, and then multi-factor analysis can be used to further analyze these paths.

[0076] The system uses superposition to quantify the risk level of sudden failures, sets maintenance priorities based on the risk level, and generates a visual report that intuitively displays key hidden dangers in the equipment. This allows for the elimination of important failures first, greatly shortening the time required for decision-making.

[0077] The state evolution decision model generates and outputs a detailed risk report. The report includes a risk source distribution list comprising all risky components and data collection points; high-risk path identifiers clearly indicating critical paths through which the fault may propagate; and a maintenance priority list of maintenance tasks ordered by risk level and impact. A maintenance strategy report is then generated based on the risk report.

[0078] The risk report clearly displays a list of component risk priorities, visually indicating paths where risk levels exceed preset thresholds and highlighting high-risk components using color coding and flashing icons to show potential fault propagation paths and the most dangerous components. This provides maintenance personnel with an intuitive basis for maintenance decisions.

[0079] The maintenance strategy report in the risk report includes specific maintenance recommendations. These recommendations cover the recommended lubricant type, detailed replacement and repair procedures, and other necessary preventative measures. After maintenance is performed according to the risk report, the system feeds the maintenance results back to the operational data chain.

[0080] Simultaneously, the model optimization is triggered by a knowledge base calibration and iterative optimization mechanism based on error feedback, which drives the decision model update and strategy optimization. When the prediction error of the lubricating medium replacement cycle and component failure exceeds the allowable deviation (defined as an absolute error of less than or equal to 7 days or a relative error of less than or equal to 10% between the predicted and actual occurrence times), the knowledge rule base parameter revision process is triggered. Specifically, the update of the state evolution decision model employs an iterative optimization algorithm based on error feedback, including a time-series model algorithm based on policy gradients. Based on this algorithm, the weights of the associated factors are the policy parameters in the model, with the goal of maximizing long-term rewards, i.e., minimizing prediction errors and improving maintenance efficiency. The state space is defined as the equipment health, lubricating medium status, and prediction error, while the action space is defined as the adjustment magnitude of the weights of specific associated factors in the knowledge rule base. The algorithm uses the Monte Carlo policy gradient algorithm, with the policy network being a neural network containing two fully connected layers, and the output layer representing the probability distribution of the adjustment magnitude. The loss function is defined as the negative logarithmic policy gradient multiplied by the reward value; the optimizer is Adam with a learning rate of 0.0001; and the policy parameters are the weights of each associated factor. The adjustment range in the action space can be defined as a floating-point number in the range of [-0.1, 0.1]. The weights of specific association factors in the knowledge rule base are modified by multiplying this floating-point number by the current weight and then summing them up.

[0081] By interacting with actual maintenance results, the model employs reinforcement learning to iteratively update maintenance strategies. This allows the model to more accurately reflect actual state evolution and achieves closed-loop optimization of intelligent lubrication management. The calibration and revision process of the knowledge rule base, in addition to adjusting weights and thresholds, also assists in optimizing the model structure. Through continuous validation and iteration, the system's prediction and data management methods can continuously improve the accuracy of data traceability and the effectiveness of maintenance strategies, ultimately achieving intelligent management of the entire lubrication system.

[0082] The system is fully validated in a controlled environment using historical data, and automatically re-evaluates, calibrates, and even adjusts some parameters of the prediction model and knowledge rule base under expert confirmation, including adjusting the parameters of the multi-dimensional association analysis module, updating the thresholds of failed association rules, and model parameters.

[0083] This invention constructs an end-to-end, intelligent solution encompassing multi-source data acquisition of lubrication media and closed-loop optimization of maintenance strategies. This system overcomes the shortcomings of traditional lubrication management methods, such as scattered data, insufficient real-time monitoring, low prediction accuracy, and wasted maintenance. It also addresses issues in initial data-driven management, such as weak data correlation, inaccurate fault propagation prediction, and neglect of environmental factors. By enabling traceable data chains, quantifiable dynamic feature correlations, updatable decision models, and multi-dimensional risk assessment, this system achieves real-time monitoring and accurate prediction of the lubrication system's health status. Ultimately, it provides data traceability capabilities for intelligent prediction and safe operation and maintenance of industrial equipment.

[0084] Example 2:

[0085] This embodiment will explain in detail how the lubrication system of a bottling machine on a beer production line can be intelligently predicted using the method proposed in this invention. This method aims to solve problems such as prediction blind spots, response lag, and low maintenance efficiency in traditional lubrication management.

[0086] Taking a beer bottling machine on a production line as an example, its 26 data acquisition points are distributed as follows: The main bearing area has one triaxial vibration sensor and one temperature sensor, totaling 16 acquisition points across 4 bearings; the gearbox input and output shafts each have one vibration sensor and one temperature sensor, totaling 4 acquisition points; the capping machine's key lubrication points each have one flow sensor and one pressure sensor, totaling 4 acquisition points; and the hydraulic pump station includes one vibration sensor, one temperature sensor, and one pressure sensor, totaling 2 acquisition points. This distribution of 26 data acquisition points ensures coverage of the bottling machine's critical force-bearing, friction-prone, and transmission components.

[0087] First, this embodiment focuses on how the model traces the required data based on a traceable operational data chain, and details how to obtain multi-source operational data throughout the entire lifecycle of the lubricating medium. Simultaneously, it establishes cross-stage data associations based on unique batch identifiers and spatiotemporal metadata to form a traceable operational data chain, such as in bottle washing machines, labeling machines, and rinsing machines in packaging equipment. These devices have numerous data collection points and often operate in special environments such as humidity, high temperatures, and confined spaces. Furthermore, these packaging devices commonly employ automated point-to-point lubrication systems, such as the German Bekaert multi-point electric pump with integrated flow sensors, to monitor oil flow and provide low-level alarm functions.

[0088] The system obtains the unique batch number of each batch of lubricating medium by scanning the unique QR code at the input stage; and extracts multi-source operational data strongly correlated with that batch number. For example, a batch of grease used for lubricating the bearings of a bottle washing machine has the batch number 20250101-001 and is manufactured by brand X; at the same time, it records the initial performance data of this batch of grease, such as dropping point 280 degrees Celsius and cone penetration 265; and the amount and time of each addition to the specific data collection points of the bottle washing machine, labeling machine, and rinsing machine.

[0089] The system continuously collects multi-dimensional data from the bottle washing machine, labeling machine, and rinsing machine during actual production, including equipment operating efficiency indicators. For example, it collects real-time temperature data for the left bearing component IDXPJ-AXC-L005 of the bottle washing machine; and noise levels in decibels for the gear meshing component ITBJ-GRL-002 of the labeling machine. Regarding lubricant condition indicators, online sensors are deployed in the oil tank and pipelines, including pipeline flow sensors to detect blockages, and to monitor the particle count of the lubricant or grease in real time, such as meeting ISO 4406 standards (levels 17 / 15 / 12), as well as moisture content, such as the moisture content in the return oil from the bottle washing machine reaching 800 ppm, higher than that of the rinsing machine. Regarding environmental conditions, environmental sensors are deployed near the equipment to collect real-time ambient temperature and humidity data for the area where the equipment is located; for example, the humidity in the bottle washing machine area is consistently maintained at 85%, significantly higher than that in the labeling machine area, and there may even be corrosive gas concentrations. All these multi-source operating data are accurately timestamped and associated with the equipment and component IDs, and logically linked to the batch number of the currently used lubricant.

[0090] During the disposal phase, when the lubricating medium reaches its replacement cycle or equipment components fail due to lubrication issues and are replaced, the system records an analysis report of the waste oil. For example, the acid value of the waste oil from the bottle washing machine may spike to 3 mg / g potassium hydroxide, and the iron content of the wear metal element may reach 150 ppm, as determined by offline spectral analysis. The system also records detailed information about the failed components, such as severe pitting corrosion on the large gear ring of the bottle rinsing machine. This scrap data is also linked to the batch number of the lubricating medium and the ID of the failed component.

[0091] Through this mechanism, all collected data—from the production, warehousing, and dispensing of lubricating media, to changes in equipment operation status, and finally to lubricating media scrapping and component failure—are tightly linked to form a traceable operational data chain. For example, by tracing the batch number of grease 20250101-001, the system can clearly see that this batch of grease was dispensed into the XPJ-AXC-L005 bearing of the bottle washing machine on January 10, 2015. Over the next six months, the bearing's operating environment humidity consistently exceeded 85% relative humidity, causing the grease's moisture content to rise continuously. Finally, in early July, the bearing temperature abnormally increased, accompanied by slight abnormal noises. Based on waste oil analysis, the system confirmed that the grease failed due to hydrolysis, leading to premature bearing wear. This end-to-end traceability of the data chain greatly enhances the understanding of the complex causal relationships between lubricating media performance, equipment health, and environmental impact.

[0092] This embodiment mainly focuses on the intelligent prediction and data traceability functions, explaining in detail how to generate oil replacement prediction suggestions, component risk priority lists, and maintenance strategy reports through data analysis and model building.

[0093] First, the system extracts real-time equipment operating efficiency indicators and lubrication medium condition indicators such as the root mean square vibration value and bearing temperature of the main bearing from the multi-source operational data chain of the filling machine. Simultaneously, the system extracts online particle counts of the gear oil via sensors, which must meet specific requirements of the ISO 4406 standard. Furthermore, sensors also extract lubrication medium condition indicators such as moisture content and dielectric constant, as well as environmental condition indicators. These data are combined with oil chemical degradation parameters such as acid value and viscosity changes from historical waste oil analysis reports, as well as real-time contamination data.

[0094] Before constructing the state evolution decision model, the dynamic correlation characteristics between these indicators are quantified through the multi-dimensional correlation analysis module in the model. For example, when the root mean square vibration value of the main bearing continuously exceeds the warning line and the gear oil particle count increases sharply, the model can identify a strong correlation between bearing wear and oil contamination. Next, the model integrates these dynamic correlation characteristics with the health assessment results of various components of the filling machine, and generates a dynamic state evolution decision model based on a pre-set knowledge rule base. For example, the output of dynamic correlation characteristics could be an evolutionary path where accelerated grease deterioration leads to an abnormal increase in the temperature of the capping machine bearing, which in turn causes increased bearing wear, ultimately increasing the risk of equipment downtime. The model will mark key indicator thresholds and decision points along this path. When the temperature of a bearing in the capping machine exceeds the safety limit for several consecutive days, such as 70 degrees Celsius, and its corresponding grease oxidation index increases significantly in a short period through dielectric constant changes, the system will match the rule that accelerated grease oxidation under high temperature and high load leads to bearing failure, and adjust the weight of this path in the model accordingly, making it more sensitive in subsequent predictions.

[0095] Based on this dynamically updated state evolution decision model, the system invokes a pre-set risk assessment knowledge base. For example, when the vibration index of the main bearing component GZJ-AXC-001 of the filling machine is detected to be in a high-level warning state, and its lubricating medium contamination level has reached the threshold requiring immediate replacement, the model identifies the bearing as a high-risk source. Subsequently, the model loads pre-set transmission path templates, such as the possibility that the failure of the main bearing may be transmitted to the motor through gears, and a multi-factor calculation model, including path association weight factors, node influence factors, and environmental adaptation correction factors, to perform a risk superposition assessment. Finally, the model generates a detailed risk report, clearly indicating that the main bearing of the filling machine is the highest risk, and inferring that it may experience a serious failure within the next one hundred hours under the current operating conditions.

[0096] Based on the risk assessment results output by the model and combined with the risk assessment report and actual management resources, the system further generates a specific maintenance strategy report. This report first provides a prediction of oil replacement, such as predicting, based on real-time deterioration data of the lubricating medium and the current load condition model, that the effective lifespan of the lubricating medium in the filling machine gearbox has been shortened to 60% of its original cycle, i.e., from 2,000 hours to 1,200 hours, and recommending replacement within the next fifteen days. Simultaneously, the report generates a component risk priority list and risk distribution map; the report lists the filling machine main bearing and related gearboxes as the highest maintenance priority and traces their origin; finally, the report provides specific maintenance recommendations, including recommended lubricating medium models, replacement methods, and bearing inspection and replacement plans. The report recommends an emergency inspection of the main bearing and replacement with a specific model of high-performance bearing to avoid potential downtime risks.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent prediction and data traceability management of lubrication systems, characterized in that, include: Acquire multi-source operational data of the lubricating medium throughout its entire life cycle, and establish cross-stage data associations based on unique batch identifiers and spatiotemporal metadata to form a traceable operational data chain; During the lubricating medium introduction stage, the system collects the unique batch identifier of the lubricating medium and records the process parameters at the time of addition; These parameters cover the filling volume, filling pressure, filling time, and filling point; at the same time, the initial performance indicators of the lubricating medium are recorded, including the viscosity, acid value, cleanliness level, and main chemical components of the new oil; when the lubricating medium enters the warehouse, the process parameters and initial performance indicators are bound to the batch number and recorded; during actual filling, the system records the filling environmental conditions and supplements them to the initial information. To achieve the association of multi-source operational data, this method uses a relational database for storage. It establishes core data tables for batch information, annotation events, and initial performance data, and associates them with the batch number as the primary key to ensure data traceability. The system records key degradation parameters and component wear status data during the waste oil disposal stage; it uses offline laboratory spectral analysis to determine the concentration of wear elements and additive losses; waste oil degradation parameters include changes in acid value and viscosity; component wear status data includes non-invasive testing results, wear type identification after disassembly of faulty components, and macroscopic wear degree assessment; all multi-source operational data contain accurate spatiotemporal metadata, including timestamps, unified equipment identifiers, unique component identifiers, and spatial location information of data acquisition points. Based on the aforementioned operational data chain, equipment operating efficiency indicators, lubricating medium status indicators, and environmental condition indicators are extracted. Combined with oil chemical degradation parameters and real-time contamination data, the dynamic correlation characteristics between them are quantified through a multi-dimensional correlation analysis module. Combining the dynamic correlation characteristics with the system health assessment results, a state evolution decision model is generated based on a preset knowledge rule base. The construction of the state evolution decision model includes the following steps: constructing a directed graph model where nodes represent various health and abnormal states of equipment components and lubrication systems, and edges represent transition paths between states; generating update rules for the decision model specifically includes matching failure association rules stored in a preset knowledge rule base with dynamic association features identified by a multi-dimensional association analysis module based on the equipment's physical connection topology, and activating and updating the weights and activation conditions of corresponding paths in the model accordingly; triggering the update of the decision model specifically includes immediately prompting maintenance actions and simultaneously initiating a knowledge rule base calibration process when the oil contamination level exceeds a preset warning threshold. The calibration process involves fine-tuning the rule weights and re-evaluating the triggering conditions, and is subject to data verification and manual confirmation. Based on the state evolution decision model, a preset risk assessment knowledge base is invoked to generate oil replacement prediction suggestions, component risk priority lists, and maintenance strategy reports. Operations are performed according to the maintenance strategy reports, and the maintenance results are fed back to the operation data chain. The decision model is updated and the strategy is optimized through a knowledge base calibration and iterative optimization mechanism based on error feedback.

2. The intelligent prediction and data traceability management method for lubrication systems according to claim 1, characterized in that, The full lifecycle multi-source operational data includes: During the input phase, the unique batch identifier of the lubricating medium, the process parameters at the time of actual filling, and the initial performance indicators of the lubricating medium are obtained. During the operation phase, the equipment operating efficiency indicators, lubricating medium status indicators, and environmental condition indicators are collected in near real-time through a multi-type sensor network deployed on key equipment components and data acquisition points. During the disposal phase, the key deterioration parameters of waste oil and component wear status data are recorded.

3. The intelligent prediction and data traceability management method for lubrication systems according to claim 1, characterized in that, The establishment of cross-stage data associations to form a traceable operational data chain includes the following specific steps: The unique batch identifier of the lubricating medium is used as the core traceability key. Based on this traceability key, the operating performance of the batch of lubricating medium in different equipment and time periods, all related abnormal events, and the final scrap analysis and the resulting component wear data are logically correlated to ensure the vertical continuity of the data flow. By establishing a correlation index between environmental impact intensity data and oil deterioration characteristic data through spatiotemporal metadata associated with all multi-source operational data, we can analyze the specific impact of environmental factors on the deterioration rate of lubricating media in different time periods and spatial regions, and identify the impact mechanism of the external environment on the health of the lubrication system.

4. The intelligent prediction and data traceability management method for lubrication systems according to claim 1, characterized in that, The execution of the risk assessment knowledge base includes the following specific steps: Based on the nodes in the state evolution decision model that meet the preset failure characteristic conditions, the risk sources are identified; Based on the physical connection topology of the equipment, a transmission path template is retrieved from the association knowledge rule base. The template describes in detail how a fault propagates from a risk source to related components through physical connection relationships and predicts its potential impact range. Loading path association weight factor, node influence factor and environment adaptation correction factor; Based on the factor superposition value, a risk level assessment is generated through a comprehensive risk assessment algorithm, and a detailed risk report is output. The report includes the distribution of risk sources, high-risk path identification, and a maintenance priority list sorted according to risk level and impact.

5. The intelligent prediction and data traceability management method for lubrication systems according to claim 1, characterized in that, The maintenance strategy report generation includes: The oil replacement prediction suggestion is based on the chemical degradation parameters and real-time contamination data of the lubricating medium. It calls a preset replacement knowledge rule base to match the equipment load conditions, so as to combine the actual degradation degree of the lubricating medium with the actual operating load of the equipment, predict the remaining effective life of the lubricating medium, and give an optimized replacement timing suggestion. The component risk priority list clearly marks paths with risk levels exceeding preset thresholds and high-risk indicators in a visual format; The maintenance strategy report also includes specific maintenance recommendations, covering recommended lubricant types, detailed replacement and repair procedures, and preventative measures.

6. The intelligent prediction and data traceability management method for lubrication systems according to claim 1, characterized in that, The decision model update and strategy optimization include: Acquire equipment status recovery data within a preset period after maintenance, including whether the data collection point and component operating performance indicators have recovered to the healthy baseline; By comparing the differences in data characteristics before and after maintenance, the actual impact of maintenance operations on equipment status and lubrication medium performance is identified, and the weights of correlation factors in the knowledge rule base are dynamically adjusted through an iterative optimization algorithm based on error feedback to more accurately reflect the actual causal relationship. When the prediction error of lubricant replacement cycle and component failure exceeds the allowable deviation, the knowledge rule base parameter revision process is triggered. This process includes adjusting the parameters of the multi-dimensional correlation analysis module, updating the threshold of failure correlation rules, and reconstructing model parameters to achieve closed-loop optimization of the entire intelligent lubrication management system.

Images