Self-adaptive cascade evaluation and prediction system for lubricating oil quality

Abstract

The invention discloses a self-adaptive cascade evaluation and prediction system for lubricating oil quality, and relates to the related field of analysis materials, and the system is characterized in that a cascade evaluation matrix is determined through quality evaluation standards of interactive lubricating oil; multi-source sample data is called through big data, preprocessing and time phase integration are carried out, and time phase sample data are determined. And traversing the time phase sample data, performing spatial-temporal scale processing, performing data-driven modeling in combination with the cascade evaluation matrix, and determining a quality decision module. Multi-source heterogeneous data of target lubricating oil is monitored and obtained and preprocessed, and a prediction task is determined. And transmitting the multi-source heterogeneous data to a quality decision module, and determining a target output matrix. And determining a quality evaluation mode by taking the demand as the guide. And in combination with a quality evaluation mode, analyzing the target output matrix, and determining a quality evaluation result. The technical problems that in the prior art, a lubricating oil quality evaluation method is complex in data processing, long in evaluation period and not high in evaluation result accuracy are solved.

Description

Technical Field

[0001] This application relates to the field of analytical materials, and in particular to an adaptive cascade evaluation and prediction system for lubricating oil quality. Background Technology

[0002] As an indispensable key medium in the operation of mechanical equipment, the quality of lubricating oil directly affects the service life and operating efficiency of the equipment. Existing lubricating oil quality assessment methods mainly rely on laboratory testing and human experience, which suffers from problems such as complex data processing, long assessment cycles, and low accuracy of assessment results. With the development of industrial big data and artificial intelligence technologies, intelligent assessment of lubricating oil quality through big data and deep learning technologies has become a trend.

[0003] Therefore, existing lubricating oil quality assessment methods suffer from technical problems such as complex data processing, long assessment cycles, and low accuracy of assessment results. Summary of the Invention

[0004] This application provides an adaptive cascaded evaluation and prediction system for lubricating oil quality, solving the technical problems of complex data processing, long evaluation cycles, and low accuracy of evaluation results in existing lubricating oil quality evaluation methods. By establishing a lubricating oil quality evaluation and prediction system based on big data and a grey prediction model, efficient and accurate evaluation of lubricating oil quality can be achieved. Furthermore, the model can be dynamically optimized through an intelligent optimizer, improving prediction performance and system adaptability.

[0005] This application provides an adaptive cascaded evaluation and prediction system for lubricating oil quality. The system includes: an evaluation matrix acquisition module, used to interact with lubricating oil quality evaluation standards, integrate and perform equivalent transformations to determine a cascaded evaluation matrix, wherein the cascaded evaluation matrix includes an index layer and a feature layer, and there is an inter-layer mapping relationship; the cascaded evaluation matrix includes physicochemical properties and mechanical properties. A sample data acquisition module is used to perform industrial big data retrieval, call multi-source sample data, and perform preprocessing and temporal integration to determine temporal sample data. A decision module is used to traverse the temporal sample data and perform spatiotemporal scale processing, combine the cascaded evaluation matrix to perform data-driven modeling, and determine a quality decision module, wherein the spatiotemporal scale processing includes spatial scale dimension processing and multi-temporal scale processing. A prediction task acquisition module is used to monitor and acquire multi-source heterogeneous data of the target lubricating oil, perform preprocessing, and determine the prediction task. An analysis module is used to transmit the multi-source heterogeneous data to the quality decision module, combine the prediction task, perform feature extraction and predictive analysis, and determine a target output matrix, wherein the target output matrix includes an evaluation matrix and a prediction matrix. The evaluation mode determination module is used to determine the quality evaluation mode based on demand, wherein the quality evaluation mode includes an assembly evaluation method and rigid limit standards. The evaluation result acquisition module is used to analyze the target output matrix in conjunction with the quality evaluation mode to determine the quality evaluation result.

[0006] In a possible implementation, the evaluation matrix acquisition module is further configured to: determine an indicator matrix; traverse the indicator matrix, determine indicator mapping features and filter them based on indicator correlation, constructing a feature matrix, wherein the indicator mapping features are one-to-one, one-to-many, or many-to-one mapping relationships; and cascade the indicator matrix and the feature matrix to generate the cascaded evaluation matrix.

[0007] In a possible implementation, the evaluation matrix acquisition module is further configured to: cascade the index matrix and the feature matrix to determine an initialization matrix; traverse the initialization matrix, perform dimensionality reduction of matrix elements based on evaluation requirements, and determine first processing data, wherein the element dimensionality reduction method includes scaling transformation and simplification processing; based on the recognition difficulty, filter pre-adjusted matrix elements, determine lower-level equivalent elements, and perform equivalence transformation based on element equivalence relations to determine second processing data; and determine the cascaded evaluation matrix based on the first processing data and the second processing data.

[0008] In a possible implementation, the decision module is further configured to: traverse the temporal sample data to determine a temporal operator sequence; establish a multi-level accumulation principle, traverse the temporal operator sequence for accumulation and reconstruction, and determine N sets of reconstructed operator sequences; based on the N sets of reconstructed operator sequences, establish a gray-level differential equation and perform data-driven modeling to determine N gray-level prediction branches; integrate the N gray-level prediction branches to generate a gray-level prediction unit, wherein the branch prediction mean is used as the prediction output.

[0009] In a possible implementation, the decision module is further configured to: establish a loss function and a trigger function; construct an optimizer, which is built into the quality decision module; perform learning constraints and model application loss measurement based on the loss function; combine the trigger function to make optimization trigger decisions; and trigger the optimizer to perform optimization learning on the quality decision module. The optimization learning of the quality decision module includes model hyperparameter learning optimization and model performance learning optimization.

[0010] In a possible implementation, the analysis module is further configured to: The quality decision module includes a quality assessment unit and a grayscale prediction unit. Based on the quality assessment unit, matrix element-based identification analysis is performed on the multi-source heterogeneous data to determine the evaluation matrix. Using the evaluation matrix as the initial time phase, task time zone prediction is performed to determine the prediction matrix.

[0011] In a possible implementation, the evaluation result acquisition module is further configured to: determine the assembly evaluation method, wherein the assembly evaluation method is a matrix-based hierarchical weighted calculation; traverse the cascaded evaluation matrix, and, guided by requirements, determine strongly correlated matrix elements that satisfy a preset correlation degree; apply threshold limits to the strongly correlated matrix elements to determine rigid limit standards.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application provides an adaptive cascaded evaluation and prediction system for lubricating oil quality, comprising: an evaluation matrix acquisition module, used to interact with lubricating oil quality evaluation standards, integrate and perform equivalent transformations to determine a cascaded evaluation matrix, wherein the cascaded evaluation matrix includes an index layer and a feature layer, and there is an inter-layer mapping relationship; the cascaded evaluation matrix includes physicochemical properties and mechanical properties. A sample data acquisition module, used to perform industrial big data retrieval, call multi-source sample data, and perform preprocessing and temporal integration to determine temporal sample data. A decision module, used to traverse the temporal sample data and perform spatiotemporal scale processing, combine the cascaded evaluation matrix to perform data-driven modeling, and determine a quality decision module, wherein the spatiotemporal scale processing includes spatial scale dimension processing and multi-temporal scale processing. A prediction task acquisition module, used to monitor and acquire multi-source heterogeneous data of the target lubricating oil, perform preprocessing, and determine the prediction task. An analysis module, used to transmit the multi-source heterogeneous data to the quality decision module, combine the prediction task, perform feature extraction and predictive analysis, and determine a target output matrix, wherein the target output matrix includes an evaluation matrix and a prediction matrix. The evaluation mode determination module is used to determine the quality evaluation mode based on demand. This quality evaluation mode includes an assembly evaluation method and rigid limit standards. The evaluation result acquisition module is used to analyze the target output matrix in conjunction with the quality evaluation mode to determine the quality evaluation result. This solves the technical problems of complex data processing, long evaluation cycles, and low accuracy of evaluation results in existing lubricating oil quality assessment methods. By establishing a lubricating oil quality assessment and prediction system based on big data and a grey prediction model, efficient and accurate assessment of lubricating oil quality can be achieved. Furthermore, the model can be dynamically optimized through an intelligent optimizer to improve prediction performance and system adaptability. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0014] Figure 1 A schematic diagram of the structure of an adaptive cascaded evaluation and prediction system for lubricating oil quality provided in an embodiment of this application; Figure 2 This is a schematic diagram illustrating the process of generating a cascaded evaluation matrix in the adaptive cascaded evaluation and prediction system for lubricating oil quality according to this application.

[0015] Figure labeling: Evaluation matrix acquisition module 11, sample data acquisition module 12, decision-making module 13, prediction task acquisition module 14, analysis module 15, evaluation mode determination module 16, evaluation result acquisition module 17. Detailed Implementation

[0016] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, systems, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0019] This application provides an adaptive cascaded evaluation and prediction system for lubricating oil quality, such as... Figure 1 As shown, the system includes: The evaluation matrix acquisition module is used to integrate and perform equivalent transformations of interactive lubricant quality evaluation standards to determine a cascaded evaluation matrix. This cascaded evaluation matrix includes an index layer and a feature layer, with inter-layer mapping relationships. The cascaded evaluation matrix includes physicochemical properties and mechanical performance. The sample data acquisition module is used for industrial big data retrieval, calling multi-source sample data and performing preprocessing and temporal integration to determine temporal sample data. The decision module is used to traverse the temporal sample data and perform spatiotemporal scale processing, combining the cascaded evaluation matrix to perform data-driven modeling and determine the quality decision module. The spatiotemporal scale processing includes spatial scale dimension processing and multi-temporal scale processing.

[0020] Existing lubricant quality assessment standards are obtained through big data interaction, and a cascaded assessment matrix is ​​determined using an equivalent transformation method. This cascaded assessment matrix includes an index layer and a feature layer, with an inter-layer mapping relationship between the index and feature layers. The matrix encompasses physicochemical properties such as viscosity, acid value, and water content, and mechanical properties such as friction coefficient and wear rate. Indicators correspond to viscosity, acid value, and water content, while features are specific characteristics of the assessment indicators; for example, viscosity assessment can be specified as dynamic viscosity and kinematic viscosity. Further, industrial big data retrieval is performed to retrieve multi-source sample data for lubricant quality assessment, including various assessment features and results. Data preprocessing includes denoising and standardization, and temporal integration of data from different time points. During temporal integration, data from different time points are sorted and integrated according to chronological order to determine the temporal sample data. The process involves traversing the temporal sample data and performing spatiotemporal scaling. Data-driven modeling is then performed using the cascaded evaluation matrix to determine the quality decision module. The spatiotemporal scaling includes spatial scale dimension processing and multi-temporal scale processing. Spatial scale dimension processing involves extracting data from the temporal sample data according to the cascaded evaluation matrix to determine the data corresponding to each sample's cascaded evaluation matrix. Multi-temporal scale processing involves grouping the time-series samples into multiple levels according to a predetermined number of groups. Temporal integration weakens data fluctuations and randomness, thereby enabling subsequent trend measurement and data prediction. The quality decision module is then constructed based on the spatiotemporally scaled temporal sample data.

[0021] Furthermore, such as Figure 2 As shown, the evaluation matrix acquisition module 11 is further configured to: determine the indicator matrix; traverse the indicator matrix, determine the indicator mapping features and filter them based on the indicator correlation, and construct a feature matrix, wherein the indicator mapping features are one-to-one, one-to-many, or many-to-one mapping relationships; cascade the indicator matrix and the feature matrix to generate the cascaded evaluation matrix.

[0022] A matrix of indicators is determined, containing all evaluation metrics related to lubricant quality. This matrix includes specific indicators of physicochemical properties and mechanical properties. For example, physicochemical properties include viscosity, flash point, and acid value, while mechanical properties include coefficient of friction and wear rate. Subsequently, the matrix is ​​iterated through, and the mapping characteristics and corresponding correlations of each indicator are determined using big data analysis. These mapping characteristics are specific features that map physicochemical properties to mechanical performance indicators, used to evaluate the relationship between these indicators; they can be one-to-one, one-to-many, or many-to-one. Then, based on the correlation between features and indicators, the mapping characteristics corresponding to indicators whose correlations meet a preset threshold are selected, and a feature matrix is ​​constructed, corresponding to each indicator. The indicator matrix and the feature matrix are cascaded to generate the cascaded evaluation matrix.

[0023] Furthermore, the evaluation matrix acquisition module 11 is also used to: cascade the indicator matrix and the feature matrix to determine an initialization matrix; traverse the initialization matrix, and, guided by evaluation requirements, perform dimensionality reduction of matrix elements to determine first processing data, wherein the element dimensionality reduction method includes scaling transformation and simplification processing; based on the recognition difficulty, filter pre-adjusted matrix elements, determine lower-level equivalent elements, and perform equivalence transformation based on element equivalence relations to determine second processing data; and determine the cascaded evaluation matrix based on the first processing data and the second processing data.

[0024] The cascaded evaluation matrix is ​​generated by concatenating the indicator matrix and the feature matrix, including: mapping the indicator matrix and the feature matrix to determine the initialization matrix; traversing the initialization matrix and performing dimensionality reduction of matrix elements based on evaluation requirements (specific indicators); determining the first processing data, which includes dimensionality-reduced features and the dimensionality-reduced features to reduce data processing volume; the dimensionality reduction method includes scaling transformation and simplification processing; scaling transformation is the normalization of feature parameters; simplification processing is used when multiple features exist, and some features have high similarity (e.g., viscosity includes dynamic viscosity and kinematic viscosity, and the two data have high similarity), so one of the data can be used as the evaluation feature; based on the difficulty of identification, pre-adjusted matrix elements are screened to determine lower-level equivalent elements, and equivalence transformation is performed based on element equivalence relations to determine the second processing data, which includes lower-level equivalent elements and corresponding pre-adjusted matrix elements. The feature recognition difficulty parameter is pre-set by professional technicians. When the recognition difficulty exceeds a certain set parameter, the corresponding feature is obtained, resulting in a pre-tuned matrix element. Subsequently, lower-level equivalent elements are determined based on the pre-tuned matrix element. These lower-level equivalent elements are the lower-level features of the pre-tuned matrix element. For example, taking kinematic viscosity as an example of a viscosity index, since kinematic viscosity requires viscosity measurements under different temperature environments, and lubricating oil commonly operates in high-temperature environments, high-temperature kinematic viscosity can be used as the lower-level equivalent element of the kinematic viscosity feature to reduce data processing volume. Finally, based on the first and second processed data, feature replacement is performed on the initialization matrix to determine the cascaded evaluation matrix.

[0025] Furthermore, the decision module 13 is also used to: traverse the temporal sample data to determine the temporal operator sequence; set a multi-level accumulation principle, traverse the temporal operator sequence for accumulation and reconstruction, and determine N sets of reconstructed operator sequences; based on the N sets of reconstructed operator sequences, establish a gray-level differential equation and perform data-driven modeling to determine N gray-level prediction branches; integrate the N gray-level prediction branches to generate a gray-level prediction unit, wherein the branch prediction mean is used as the prediction output.

[0026] The quality decision-making module includes: traversing the temporal sample data to determine the temporal operator sequence, where the temporal operator sequence is the data sequence of each time series in the temporal sample data. Subsequently, a multi-level accumulation principle is set, which includes a predetermined number of groups. Based on the multi-level accumulation principle, the temporal operator sequence is traversed and reconstructed according to the predetermined number of groups to determine N sets of reconstructed operator sequences. For example, the multi-temporal scale processing involves multi-level accumulation and grouping of the temporal operator sequences according to the predetermined number of groups. For instance, for the first time point, the second time point, the third time point, and the fourth time point, if the predetermined number of groups is 2, then the groups are: the first time point and the second time point as one group, the second time point and the third time point as another group, and the third time point and the fourth time point as yet another group, totaling three sets of reconstructed operator sequences. Based on the N sets of reconstructed operator sequences, a gray-level differential equation is established and data-driven modeling is performed based on a gray prediction model construction method. Each set of reconstructed operator sequences constructs a gray-level prediction branch, determining N gray-level prediction branches. When constructing the gray-level prediction branches, a first-order linear differential equation is first established, which describes the dynamic change process of the system. The reconstruction operator sequence is input into the first-order linear differential equation, and the parameters in the gray-level differential equation are estimated using methods such as least squares to construct a data-driven model. Finally, the N gray-level prediction branches are integrated to generate a gray-level prediction unit, where the average of the branch predictions is used as the prediction output; that is, the average of the outputs of all gray-level prediction branches is used as the final prediction output.

[0027] Furthermore, the decision module 13 is also used to: establish a loss function and a trigger function, construct an optimizer, which is built into the quality decision module. Based on the loss function, learning constraints and model application loss measurement are applied; combined with the trigger function, optimization trigger decisions are made to trigger the optimizer to perform optimization learning on the quality decision module. The optimization learning of the quality decision module includes model hyperparameter learning optimization and model performance learning optimization.

[0028] After determining the quality decision module, the process includes: establishing a loss function and a trigger function, and constructing an optimizer, which is built into the quality decision module. The loss function measures the difference between the model's predicted result and the true value. In lubricating oil quality assessment and prediction systems, commonly used loss functions include mean squared error (MSE) and mean absolute error (MAE). The trigger function determines when to perform optimization learning. The trigger function can judge whether the model needs optimization based on the value or rate of change of the loss function. For example, the mean squared error (MSE) is used to measure the error of the lubricating oil quality prediction result. When the error does not decrease significantly in several consecutive iterations, the optimizer is triggered to perform optimization learning. The optimizer is used to adjust the model parameters to minimize the loss function. The Adam optimizer is used because it performs well in handling sparse and high-dimensional data and has the advantage of adaptive learning rate adjustment. The optimizer is built into the quality decision module and is responsible for adjusting the model parameters under the guidance of the trigger function. During model training, the loss function is used for learning constraints, i.e., the model is optimized by minimizing the loss function. In each iteration, the MSE between the predicted result and the true value is calculated, and the MSE is minimized through the backpropagation algorithm. The current loss value is recorded for subsequent trigger function evaluation. If the trigger condition is met, the optimizer is activated to adjust model parameters and optimize the quality decision module. The optimization of the quality decision module includes model hyperparameter learning and optimization, and model performance learning and optimization. Hyperparameters include learning rate, batch size, regularization coefficient, etc. Optimizing these hyperparameters can improve the model's convergence speed and prediction performance. Optimizing model parameters (such as background value and gray number in a gray prediction model) improves the model's prediction accuracy and generalization ability on test data.

[0029] The prediction task acquisition module monitors and acquires multi-source heterogeneous data of the target lubricating oil, preprocesses it, and determines the prediction task. The analysis module transmits the multi-source heterogeneous data to the quality decision module, performs feature extraction and predictive analysis in conjunction with the prediction task, and determines the target output matrix, which includes an evaluation matrix and a prediction matrix. The evaluation mode determination module determines a quality evaluation mode based on demand, wherein the quality evaluation mode includes an assembly evaluation method and rigid limit standards. The evaluation result acquisition module analyzes the target output matrix in conjunction with the quality evaluation mode to determine the quality evaluation result.

[0030] Real-time acquisition and monitoring of multi-source heterogeneous data of the target lubricating oil is performed and preprocessed, including noise reduction and standardization. A prediction task is then determined, which is the input prediction task and includes the prediction object, prediction time point, and prediction requirements such as viscosity. The multi-source heterogeneous data is transmitted to the quality decision module. Based on the prediction task, feature extraction and predictive analysis are performed on the multi-source heterogeneous data to determine the target output matrix. The target output matrix includes an evaluation matrix and a prediction matrix. The evaluation matrix is ​​the feature parameter data of the current time point extracted from the multi-source heterogeneous data according to a cascaded evaluation matrix structure. The prediction matrix is ​​the prediction matrix for the prediction time point output after predictive analysis of the evaluation matrix. Furthermore, a demand-oriented quality evaluation model is determined. This model includes an assembly evaluation method and a rigid limit standard. The assembly evaluation method involves overall scoring of the matrix; if the overall weighted calculation result is less than or equal to a preset judgment threshold, the corresponding lubricating oil quality meets the requirements. The rigid limit standard involves judging the more important data in the matrix using rigid criteria; if the rigid standard limit is not met, the corresponding result is unqualified. Finally, the target output matrix is ​​analyzed based on the quality evaluation model to determine the quality evaluation result. This solves the technical problems of complex data processing, long evaluation cycle, and low accuracy of evaluation results in existing lubricating oil quality assessment methods. By establishing a lubricating oil quality assessment and prediction system based on big data and a grey prediction model, efficient and accurate assessment of lubricating oil quality can be achieved. Furthermore, the model can be dynamically optimized through an intelligent optimizer to improve prediction performance and system adaptability.

[0031] Furthermore, the analysis module 15 is also used for: the quality decision module includes a quality assessment unit and a gray-scale prediction unit. Based on the quality assessment unit, the multi-source heterogeneous data is subjected to matrix element-based identification analysis to determine the evaluation matrix. Using the evaluation matrix as the initial time phase, task time zone prediction is performed to determine the prediction matrix.

[0032] The determination of the target output matrix includes: the quality decision module comprises a quality assessment unit and a gray-scale prediction unit. Based on the quality assessment unit, the multi-source heterogeneous data is analyzed based on matrix elements according to the cascaded assessment matrix, and feature parameters corresponding to the cascaded assessment matrix in the multi-source heterogeneous data are extracted to determine the evaluation matrix. Further, using the evaluation matrix as the initial time phase, the gray-scale prediction unit is input to perform task time zone prediction for the prediction task, and the prediction matrix for the corresponding time node is determined.

[0033] Furthermore, the evaluation result acquisition module 17 is also used to: determine the assembly evaluation method, wherein the assembly evaluation method is a matrix-based hierarchical weighted calculation; traverse the cascaded evaluation matrix, and, guided by requirements, determine strongly correlated matrix elements that meet a preset correlation degree; apply threshold limits to the strongly correlated matrix elements to determine rigid limit standards.

[0034] The quality evaluation model includes an assembly evaluation method and rigid limit standards, comprising: determining the assembly evaluation method, wherein the assembly evaluation method is a matrix-based hierarchical weighted calculation to ensure the scientificity and accuracy of the evaluation results. Before performing the matrix-based hierarchical weighted calculation, an evaluation matrix is ​​pre-constructed, which includes the weights of each element in the cascaded evaluation matrix; the greater the importance of an element, the higher its corresponding weight parameter. Based on the evaluation matrix, a standard data deviation is calculated for the target output matrix; the standard data is the standard parameter of the lubricant category cascaded evaluation matrix at the corresponding time node. The data deviation and the corresponding weights of the evaluation matrix are weighted and summed to obtain the weighted calculation result. The weighted calculation result is compared with a preset judgment threshold to determine whether the weighted calculation result is less than or equal to the preset judgment threshold. When it is less than or equal to the preset judgment threshold, the corresponding lubricant quality is better; otherwise, the lubricant quality is worse. The preset judgment threshold is the minimum weighted calculation result that pre-sets to meet the lubricant quality standard. Furthermore, the cascaded evaluation matrix is ​​traversed, guided by demand. This means that, based on actual needs, such as the requirement for lubricating oil to meet specific mechanical equipment viscosity requirements, viscosity indicators in the cascaded evaluation matrix are selected. Then, by using the corresponding weights in the evaluation matrix, strongly correlated matrix elements that meet a preset correlation degree are determined. Specifically, the preset correlation degree is used to determine whether the weights meet the requirements. When a weight is greater than or equal to the preset correlation degree, the corresponding feature element is a strongly correlated matrix element. The preset correlation degree is a pre-set minimum weight parameter for meeting strong correlation. Finally, threshold limits are applied to the strongly correlated matrix elements. This involves judging the parameters of the strongly correlated matrix elements based on pre-set threshold limits, determining whether the strongly correlated matrix elements meet the threshold limits. For example, if some parameters are greater than or equal to, the threshold limit requirement is met; if some parameters are less than or equal to, the threshold limit requirement is met. Different judgment methods exist for each parameter, establishing a rigid limit standard. If any parameter fails to meet the rigid standard, the corresponding lubricating oil quality is deemed unqualified.

[0035] This application embodiment integrates and performs equivalent transformations on interactive lubricant quality assessment standards to determine a cascaded assessment matrix. This cascaded assessment matrix includes an index layer and a feature layer, with inter-layer mapping relationships. The matrix includes physicochemical properties and mechanical performance. Industrial big data retrieval is performed, calling multi-source sample data and preprocessing and temporal integration to determine temporal sample data. The temporal sample data is traversed and spatiotemporal scale processing is performed. Data-driven modeling is then combined with the cascaded assessment matrix to determine a quality decision module. Spatiotemporal scale processing includes spatial scale dimension processing and multi-temporal scale processing. Multi-source heterogeneous data of the target lubricant is monitored and preprocessed to determine the prediction task. The multi-source heterogeneous data is transmitted to the quality decision module. Combined with the prediction task, feature extraction and predictive analysis are performed to determine the target output matrix, which includes an evaluation matrix and a prediction matrix. A demand-oriented quality evaluation mode is determined, including an assembly evaluation method and rigid limit standards. The target output matrix is ​​analyzed based on the quality evaluation mode to determine the quality evaluation result. This invention addresses the technical problems of complex data processing, long evaluation cycles, and low accuracy in existing lubricant quality assessment methods. By establishing a lubricant quality assessment and prediction system based on big data and a grey prediction model, efficient and accurate assessment of lubricant quality can be achieved. Furthermore, the model can be dynamically optimized using an intelligent optimizer to improve predictive performance and system adaptability.

[0036] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An adaptive cascaded evaluation and prediction system for lubricating oil quality, characterized in that, The system includes: The evaluation matrix acquisition module is used to integrate and perform equivalent transformations of the quality evaluation standards for interactive lubricating oils to determine the cascade evaluation matrix. The cascade evaluation matrix includes an index layer and a feature layer, and there is an inter-layer mapping relationship. The cascade evaluation matrix includes physicochemical properties and mechanical properties. The sample data acquisition module is used for industrial big data retrieval, calling up multi-source sample data and performing preprocessing and temporal integration to determine the temporal sample data; The decision module is used to traverse the temporal sample data and perform spatiotemporal scale processing, combine the cascaded evaluation matrix to perform data-driven modeling, and determine the quality decision module. The spatiotemporal scale processing includes spatial scale dimension processing and multi-temporal scale processing. The prediction task acquisition module is used to monitor and acquire multi-source heterogeneous data of the target lubricating oil, perform preprocessing, and determine the prediction task. The analysis module is used to transmit the multi-source heterogeneous data to the quality decision module, and combine it with the prediction task to perform feature extraction and predictive analysis to determine the target output matrix, which includes an evaluation matrix and a prediction matrix. The evaluation mode determination module is used to determine the quality evaluation mode based on demand. The quality evaluation mode includes the assembly evaluation method and the rigid limit standard. The evaluation result acquisition module is used to analyze the target output matrix in conjunction with the quality evaluation mode to determine the quality evaluation result.

2. The adaptive cascade evaluation and prediction system for lubricating oil quality as described in claim 1, characterized in that, The evaluation matrix acquisition module is also used for: Determine the indicator matrix; Traverse the indicator matrix, determine the indicator mapping features and filter them based on the indicator correlation degree to construct a feature matrix, wherein the indicator mapping features are one-to-one, one-to-many, or many-to-one mapping relationships. The index matrix and the feature matrix are cascaded to generate the cascaded evaluation matrix.

3. The adaptive cascade evaluation and prediction system for lubricating oil quality as described in claim 2, characterized in that, The evaluation matrix acquisition module is also used for: The index matrix and the feature matrix are concatenated to determine the initialization matrix; Traverse the initialization matrix and, guided by the evaluation requirements, perform dimensionality reduction of matrix elements to determine the first processing data. The element dimensionality reduction method includes scaling transformation and simplification processing. Based on the difficulty of identification, the elements of the pre-adjusted matrix are screened, the lower equivalent elements are determined, and the equivalent transformation is performed based on the element equivalence relationship to determine the second processing data; The cascaded evaluation matrix is ​​determined based on the first processed data and the second processed data.

4. The adaptive cascaded evaluation and prediction system for lubricating oil quality as described in claim 1, characterized in that, The decision module is also used for: The temporal sample data is traversed to determine the temporal operator sequence; A multi-level accumulation principle is set, and the time phase operator sequence is traversed for accumulation and reconstruction to determine N sets of reconstruction operator sequences; Based on the N sets of reconstruction operator sequences, a gray-level differential equation is established and data-driven modeling is performed to determine N gray-level prediction branches. The N grayscale prediction branches are integrated to generate a grayscale prediction unit, wherein the average of the branch predictions is used as the prediction output.

5. The adaptive cascaded evaluation and prediction system for lubricating oil quality as described in claim 1, characterized in that, The decision module is also used for: Establish a loss function and a trigger function, and construct an optimizer, which is built into the quality decision module; Based on the loss function, learning constraints and model application loss measurement are performed. Combined with the trigger function, optimization trigger decisions are made to trigger the optimizer to optimize the quality decision module.

6. The adaptive cascade evaluation and prediction system for lubricating oil quality as described in claim 5, characterized in that, The optimization learning of the quality decision module includes model hyperparameter learning optimization and model performance learning optimization.

7. The adaptive cascaded evaluation and prediction system for lubricating oil quality as described in claim 1, characterized in that, The analysis module is also used for: The quality decision module includes a quality assessment unit and a gray-scale prediction unit; Based on the quality assessment unit, the multi-source heterogeneous data is subjected to matrix element-based identification and analysis to determine the evaluation matrix; Using the evaluation matrix as the initial time phase, the task time zone is predicted to determine the prediction matrix.

8. The adaptive cascade evaluation and prediction system for lubricating oil quality as described in claim 1, characterized in that, The evaluation result acquisition module is also used for: The evaluation method for the assembly is determined, wherein the evaluation method for the assembly is a matrix-based hierarchical weighted calculation; Traverse the cascaded evaluation matrix and, based on demand, determine the strongly correlated matrix elements that meet the preset correlation degree. Threshold limits are applied to the elements of the strongly correlated matrix to determine rigid limit standards.

Images