Industrial big data-oriented online real-time fault diagnosis and prediction maintenance platform

Abstract

The invention discloses an industrial big data-oriented online real-time fault diagnosis and prediction maintenance platform, particularly relates to the technical field of fault monitoring, and is used for solving the phenomenon that prediction output is possibly inconsistent in the actual industrial application process of an existing machine learning model. The platform collects and records real-time monitoring data from multi-source industrial equipment, predicts the data by using various pre-trained machine learning models after preprocessing, and performs conflict detection on model prediction output; and when it is confirmed that prediction is inconsistent, the platform recognizes an abnormal phenomenon caused by insufficient model adaptability based on nonlinear comparison of a historical operation scene and current working condition data in information entropy and phase characteristics, and dynamically adjusts the weight of each model accordingly, so that accurate recognition of the equipment fault type and prediction of the remaining available life are realized. And the accuracy of fault prediction and the reliability of system operation are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault monitoring, and more specifically, to an online real-time fault diagnosis and predictive maintenance platform for industrial big data. Background Art

[0002] At present, online real-time fault diagnosis and predictive maintenance of industrial data mainly adopt technical means such as data acquisition, data preprocessing, model training, and fault prediction to monitor and analyze the operating status of industrial equipment; to improve the accuracy of fault prediction, various machine learning models are usually used, including deep learning networks, neural networks, and multi-task learning models, to model and predict the real-time monitoring data of industrial equipment.

[0003] However, since each machine learning model is trained separately on different data sets or specific working conditions, there may be a phenomenon of inconsistent prediction outputs in the actual application process, that is, some machine learning models can accurately reflect the equipment status under certain working conditions, but are misleading under other working conditions, resulting in the overall prediction result deviating from the actual situation. This phenomenon can be called negative transfer, which will make it difficult to ensure the consistency and reliability of the prediction result when multiple models are fused. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an online real-time fault diagnosis and predictive maintenance platform for industrial big data to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An online real-time fault diagnosis and predictive maintenance platform for industrial big data, comprising:

[0007] A data acquisition module: receiving and recording real-time monitoring data from multi-source industrial equipment;

[0008] A multi-model management module: storing and scheduling multiple machine learning models to process the real-time monitoring data of industrial equipment;

[0009] A model conflict detection module: analyzing the prediction outputs of each machine learning model for the real-time monitoring data of the same industrial equipment and identifying prediction result conflicts that occur under different working conditions;

[0010] A negative transfer identification module: after confirming the existence of prediction result conflicts, determining whether there is negative transfer caused by the wrong adaptation of the machine learning model to the working conditions based on the analysis of the differences between the historical operation scenarios and the current working conditions of the industrial equipment;

[0011] Dynamic weight adjustment module: When there is negative transfer, reassign different weight parameters to each machine learning model so that the machine learning model more suitable for the current working condition obtains a larger proportion;

[0012] Fault diagnosis and prognosis module: Identify the types of equipment faults based on the adjusted weight parameters and predict the remaining available life of the equipment.

[0013] In a preferred embodiment, receive and record the real-time monitoring data from multi-source industrial equipment, including:

[0014] Receive the real-time monitoring data from multiple industrial equipment. The real-time monitoring data includes temperature, pressure, vibration, current, voltage, and equipment status;

[0015] Preprocess the received real-time monitoring data of each industrial equipment and then record it.

[0016] In a preferred embodiment, store and schedule multiple machine learning models to process the real-time monitoring data of industrial equipment, including:

[0017] Store multiple machine learning models used to process the real-time monitoring data of industrial equipment. The machine learning models are pre-trained before storage to determine the corresponding structural parameters and weight parameters. The storage step saves each machine learning model and its parameters in a stable storage medium;

[0018] According to the preset scheduling criteria, schedule the stored multiple machine learning models. The scheduling criteria include sorting and selecting multiple machine learning models based on the characteristics of the real-time monitoring data of industrial equipment, processing delay requirements, and model prediction performance. The scheduling step sequentially calls the selected machine learning models to process the preprocessed and recorded real-time monitoring data of industrial equipment.

[0019] In a preferred embodiment, analyze the prediction outputs of each machine learning model for the real-time monitoring data of the same industrial equipment and identify the prediction result conflicts that occur under different working conditions, including:

[0020] Obtain the prediction outputs of multiple pre-trained machine learning models. The prediction outputs are generated based on the preprocessed and recorded real-time monitoring data of industrial equipment;

[0021] Conduct a systematic comparison and differential analysis of the prediction data output by each machine learning model within the same time period; identify the prediction result conflicts that occur under different industrial equipment working conditions according to the preset threshold and consistency determination criteria;

[0022] Among them, different industrial equipment working conditions include changes in equipment load, environmental temperature fluctuations, and changes in equipment operating status.

[0023] In a preferred embodiment, after confirming the existence of a prediction result conflict, it is determined whether there is negative transfer caused by the machine learning model's incorrect adaptation to the working conditions based on the analysis of the differences between the historical operation scenarios and the current industrial equipment working conditions, including:

[0024] Perform signal processing on the historical operation scenario data and the current industrial equipment working condition data respectively to obtain their multi-scale information entropy distributions, which reflect the complexity characteristics of the data at multiple time scales; then, calculate the dynamic entropy coupling coefficient by calculating the mutual information difference between the information entropy distributions of the historical operation scenario data and the current industrial equipment working condition data at each scale, and the dynamic entropy coupling coefficient indicates whether there is a significant deviation in information transfer between the two sets of data;

[0025] Perform signal decomposition and instantaneous phase extraction operations on the historical operation scenario data and the current industrial equipment working condition data respectively to obtain their respective instantaneous phase sequences, compare the instantaneous phase sequences obtained at each time scale, and calculate the multi-scale phase consistency index, which reflects the degree of consistency between the two sets of data in terms of instantaneous phase changes;

[0026] Set the lower limit of the dynamic entropy coupling coefficient and the threshold of the multi-scale phase consistency index. When the dynamic entropy coupling coefficient is lower than the lower limit of the dynamic entropy coupling coefficient and the multi-scale phase consistency index is lower than the threshold of the multi-scale phase consistency index, it is determined that there is a negative transfer phenomenon caused by the machine learning model's failure to adapt to the current industrial equipment working conditions; otherwise, it is determined that there is no negative transfer phenomenon caused by the machine learning model's failure to adapt to the current industrial equipment working conditions.

[0027] In a preferred embodiment, the calculation steps of the dynamic entropy coupling coefficient are as follows:

[0028] Obtain two sets of data, including historical operation scenario data and current industrial equipment working condition data, and obtain the Shannon entropy value sequences of the two sets of data at multiple time scales;

[0029] Determine the dynamic entropy coupling coefficient by calculating the mutual information difference between the entropy distributions of the two sets of data at each time scale:

[0030] The total mutual information between the two sets of data is as follows: Among them, represents the total mutual information of the two sets of data at all M time scales, I m represents the mutual information between the two sets of data at the m-th time scale, M represents the total number of time scales, and m represents the index of different time scales;

[0031] Calculate the average Shannon entropy values of the two sets of data respectively, denoted as <H (X) > and <H(Y) >;

[0032] Define the dynamic entropy coupling coefficient as: where Θ represents the dynamic entropy coupling coefficient, and max(<H (X) >, <H (Y) ) represents the larger value of the average Shannon entropy values in two groups of data.

[0033] In a preferred embodiment, the calculation steps of the multi-scale phase consistency index are as follows:

[0034] Obtain the intrinsic mode function sequence decomposed from historical operation scenario data and the intrinsic mode function sequence decomposed from current industrial equipment condition data;

[0035] For each intrinsic mode function, extract its instantaneous phase information through Hilbert transform;

[0036] Calculate the instantaneous phase difference under each intrinsic mode function, where the instantaneous phase difference is the absolute difference between the instantaneous phases of historical operation scenario data and current industrial equipment condition data;

[0037] Calculate the multi-scale phase consistency index using a non-linear similarity function.

[0038] In a preferred embodiment, when there is negative transfer, reassign different weight parameters to each machine learning model so that the machine learning model more suitable for the current working condition obtains a larger proportion, including:

[0039] After the occurrence of negative transfer due to the machine learning model's incorrect adaptation to the working condition, compare the matching degree between the prediction output generated by each pre-trained machine learning model and the industrial equipment real-time monitoring data recorded in the preprocessing, and readjust the weight parameters of each machine learning model in the overall prediction. Among them, the machine learning model with a higher matching degree between the prediction output and the actual data obtains a higher weight parameter, while the machine learning model with a lower matching degree obtains a lower weight parameter.

[0040] In a preferred embodiment, identify the equipment fault type based on the adjusted weight parameters and predict the remaining available life of the equipment, including:

[0041] Perform weighted fusion on the prediction outputs generated by each machine learning model based on the re-adjusted weight parameters to obtain a comprehensive prediction result; among them, the weighted fusion is realized by the cumulative addition of the product of each prediction output and its corresponding weight parameter;

[0042] Analyze the comprehensive prediction result to identify the equipment fault type and clarify the equipment operation state and the cause of the fault;

[0043] Based on historical device operation data, predict the remaining useful life of the device, and determine the future operation cycle and maintenance plan of the device.

[0044] The technical effects and advantages of an online real-time fault diagnosis and predictive maintenance platform for industrial big data according to the present invention:

[0045] 1. By using the non-linear quantitative comparison of historical operation scenario data and current industrial device working condition data in terms of multi-scale information entropy distribution and instantaneous phase characteristics, it is possible to more accurately judge whether there is a prediction deviation caused by insufficient adaptability of the machine learning model under actual working conditions, so as to achieve precise identification of abnormal phenomena; by deeply analyzing the complexity of data on multiple time scales and using signal processing and statistical analysis methods to extract information entropy and phase consistency indicators, the obtained evaluation results can fully reflect the subtle differences in data characteristics between different working conditions, thus providing a strong basis for the reliability of the prediction results.

[0046] 2. Form a strict closed-loop process for device operation status monitoring, prediction output analysis, data comparison, etc., so that after detecting inconsistent prediction output, targeted analysis can be quickly started, and the weights of each machine learning model can be dynamically adjusted according to the evaluation indicators. This adjustment process fully considers the applicability of each machine learning model under the current working conditions, so that the machine learning model with a higher degree of matching with the actual operation data in the prediction output obtains a larger proportion, while those machine learning models with poor adaptability correspondingly reduce their influence, thereby optimizing the overall prediction result; through this multi-link, non-linear information fusion and weight allocation scheme, each link forms a tight coupling, and the synergistic effect significantly improves the accuracy and stability of fault type identification and remaining useful life prediction, effectively reducing the prediction error caused by data feature mismatch, and ensuring the real-time operation and efficient maintenance of the system in a complex industrial environment. Description of the Drawings

[0047] Figure 1 It is a schematic structural diagram of an online real-time fault diagnosis and predictive maintenance platform for industrial big data according to the present invention. Detailed Embodiments

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] Embodiment: Figure 1The present invention provides an online real-time fault diagnosis and predictive maintenance platform for industrial big data, including:

[0050] Data acquisition module: Receives and records real-time monitoring data from multi-source industrial equipment.

[0051] Multi-model management module: Stores and schedules multiple machine learning models to process real-time monitoring data of industrial equipment.

[0052] Model conflict detection module: Analyzes the prediction outputs of each machine learning model for the real-time monitoring data of the same industrial equipment and identifies prediction result conflicts that occur under different working conditions.

[0053] Negative transfer identification module: After confirming the existence of prediction result conflicts, determines whether there is negative transfer caused by the incorrect adaptation of the machine learning model to the working conditions based on the analysis of the differences between the historical operation scenarios and the current working conditions of the industrial equipment.

[0054] Dynamic weight adjustment module: When there is negative transfer, reassigns different weight parameters to each machine learning model so that the machine learning model more suitable for the current working conditions obtains a greater proportion.

[0055] Fault diagnosis and predictive maintenance module: Identifies the types of equipment faults based on the adjusted weight parameters and predicts the remaining available life of the equipment.

[0056] Receives and records real-time monitoring data from multi-source industrial equipment, including:

[0057] Receives real-time monitoring data from multiple industrial equipment, and the real-time monitoring data includes temperature, pressure, vibration, current, voltage, and equipment status.

[0058] Industrial equipment includes but is not limited to equipment operating in industrial production processes such as machine tools, fans, pumps, compressors, transformers, boilers, turbines, etc., and the real-time monitoring data includes but is not limited to data such as temperature, pressure, vibration, current, voltage, rotational speed, flow rate, load, and equipment operating status.

[0059] Industrial equipment transmits real-time monitoring data through industrial field buses, wireless transmission networks, industrial Ethernet, or other communication methods, and the data uses different communication protocols, including but not limited to Modbus, PROF INET, EtherCAT, CAN, OPC UA, etc. After receiving the data, it parses different protocol formats to convert them into a unified structured data format.

[0060] Records the real-time monitoring data of each industrial equipment received after preprocessing.

[0061] Preprocessing includes but is not limited to the following steps: eliminating outliers generated by transmission interference during communication and filtering out short-term impulse interference data through the time window method; since the sampling frequencies of different industrial devices may vary, aligning data from different sources according to timestamps to ensure data synchronization; when data from a certain sensor is missing due to network loss or signal interference, using interpolation, time series regression, or prediction methods based on historical data to fill in the missing values to maintain data integrity; converting some analog signal data into digital signals and performing unit conversion according to device calibration parameters to ensure that all data is in a unified unit system.

[0062] Recording methods include but are not limited to: storing according to the time series to ensure the time continuity of data; storing according to device categories to ensure subsequent query and analysis based on the device dimension; using an industrial database or a distributed storage system for data storage to meet the requirements of high-concurrency reading and long-term storage; setting a data storage strategy to archive or compress and store data that has not been used for a long time to optimize storage space.

[0063] Set access permissions for the recorded real-time monitoring data and provide a standardized interface for subsequent fault diagnosis, predictive maintenance, and data analysis calls. The interface supports multiple data query methods, including retrieval by device, by time, and by parameter type, to ensure data availability.

[0064] Store and schedule multiple machine learning models to process the real-time monitoring data of industrial devices, including:

[0065] Store multiple machine learning models used to process the real-time monitoring data of industrial devices. The machine learning models are pre-trained before storage to determine the corresponding structural parameters and weight parameters. The storage step saves each machine learning model and its parameters in a stable storage medium.

[0066] First, pre-train multiple machine learning models used to process the real-time monitoring data of industrial devices. During the pre-training process, each machine learning model is trained based on the preprocessed real-time monitoring data of industrial devices to determine the corresponding structural parameters and weight parameters.

[0067] Among them, the structural parameters reflect the hierarchical structure, node configuration, and connection method of the internal network of the model, while the weight parameters reflect the connection strength between nodes.

[0068] After the pre-training is completed, save each machine learning model and the determined structural parameters and weight parameters to a stable storage medium. The stable storage medium includes but is not limited to databases, file storage systems, or other storage devices with long-term data storage functions.

[0069] During the saving process, the parameter data of each machine learning model is verified, and the corresponding training version, timestamp, and other necessary information are recorded to ensure the integrity and consistency of the stored data.

[0070] According to the preset scheduling criteria, multiple stored machine learning models are scheduled. The scheduling criteria include sorting and selecting multiple machine learning models based on the characteristics of the real-time monitoring data of industrial equipment, the processing delay requirements, and the model prediction performance. The scheduling steps sequentially call the selected machine learning models to process the preprocessed and recorded real-time monitoring data of industrial equipment.

[0071] After the storage of the machine learning models is completed, multiple stored machine learning models are called and processed according to the preset scheduling criteria.

[0072] The preset scheduling criteria include, but are not limited to, sorting and selecting multiple machine learning models based on the characteristics of the real-time monitoring data of industrial equipment, the data processing delay requirements, and the prediction performance shown by each machine learning model during the pre-training stage.

[0073] Specifically, the scheduling criteria evaluate indicators such as the accuracy rate, response time, and stability obtained by each machine learning model during the pre-training process to determine the priority call order of the models. Subsequently, according to the determined priority order, each selected machine learning model is called in sequence. Each called machine learning model receives the preprocessed and recorded real-time monitoring data of industrial equipment in the previous step for processing and generates corresponding prediction outputs.

[0074] During the scheduling process, ensure that each machine learning model uses the same data format and time identifier as in the preprocessing stage when processing data to ensure the consistency of the output data in terms of format, time continuity, and data synchronization.

[0075] Analyze the prediction outputs of each machine learning model for the same real-time monitoring data of industrial equipment and identify prediction result conflicts that occur under different working conditions, including:

[0076] Obtain the prediction outputs of multiple pre-trained machine learning models, which are generated based on the preprocessed and recorded real-time monitoring data of industrial equipment.

[0077] During the pre-training phase, multiple machine learning models are trained respectively using the real-time monitoring data of industrial equipment obtained after preprocessing steps such as data cleaning, data alignment, data completion, and data transformation, so as to determine their respective structural parameters and weight parameters. After the pre-training is completed, each machine learning model generates corresponding prediction outputs based on the preprocessed and recorded real-time monitoring data of industrial equipment. Here, the real-time monitoring data of industrial equipment includes parameters such as temperature, pressure, vibration, current, voltage, and equipment operating status, and its data format, time series consistency, and numerical accuracy have been ensured during the data preprocessing phase.

[0078] Systematically compare and analyze the difference in the prediction data output by each machine learning model within the same time period; according to the preset threshold and consistency determination criteria, identify the prediction result conflicts that occur under different industrial equipment operating conditions.

[0079] Among them, different industrial equipment operating conditions include changes in equipment load, fluctuations in ambient temperature, and changes in equipment operating status.

[0080] After obtaining the prediction outputs generated by each machine learning model for the real-time monitoring data of the same industrial equipment within the same time period, systematically compare and analyze the difference in these prediction output data. This comparative analysis uses a quantitative comparison method, and its basic idea is to calculate the difference value between the prediction outputs of any two machine learning models, introducing the following formula: ΔP (i，j) =|P i -P j |; where, ΔP (i，j) represents the absolute difference between the prediction outputs generated by machine learning model serial number i and machine learning model serial number j for the real-time monitoring data of industrial equipment within the same time period; P i represents the prediction output generated by the i-th machine learning model for this data; P j represents the prediction output generated by the j-th machine learning model for this data.

[0081] In the above formula, i and j are respectively the indexes representing different machine learning models, and i and j are not equal; at the same time, there is a preset threshold corresponding to ΔP (i，j) , and this threshold is the upper limit of the allowable error of the prediction output obtained by statistical analysis of historical data. If for any pair of machine learning models, the calculated ΔP (i，j) exceeds its corresponding preset threshold, it is considered that within this time period, the prediction outputs of different machine learning models for the real-time monitoring data of the same industrial equipment are significantly inconsistent.

[0082] Among them, the consistency ratio is used to measure the consistency of the model prediction results: Among them, R (i，j)represents the consistency ratio of the prediction results of machine learning model serial number i and machine learning model serial number j in the current time period; max(P i , P j ) represents taking the larger value of the prediction results of the two machine learning models to avoid the denominator approaching zero during normalization calculation.

[0083] If ΔP (i，j) exceeds its corresponding preset threshold, and R (i，j) is lower than the consistency judgment criterion R th , it is determined that the prediction results of the machine learning model conflict during this time period.

[0084] Here, the setting basis of the preset threshold corresponding to ΔP (i，j) includes but is not limited to the average deviation of historical operation data, the range of equipment load changes, the amplitude of environmental temperature fluctuations, and the changes in the operating state of the equipment; R th is the lowest threshold of the consistency ratio, usually set between 0.85 - 0.95, and adjusted according to historical data statistics.

[0085] After confirming that there is a conflict in the prediction results, based on the analysis of the differences between the historical operation scenarios and the current industrial equipment working conditions, it is determined whether there is a negative transfer caused by the machine learning model's incorrect adaptation to the working conditions, including:

[0086] Signal processing is respectively performed on the historical operation scenario data and the current industrial equipment working condition data to obtain their respective multi-scale information entropy distributions. The multi-scale information entropy distribution reflects the complexity characteristics of the data at multiple time scales; then, the dynamic entropy coupling coefficient is calculated by calculating the mutual information difference between the information entropy distributions of the historical operation scenario data and the current industrial equipment working condition data at each scale. The dynamic entropy coupling coefficient indicates whether there is a significant deviation in information transfer between the two groups of data.

[0087] The signal processing operations include noise reduction, smoothing, and multi-scale decomposition processing, aiming to extract the complexity characteristics of the data at different time scales. After the above processing, the Shannon entropy value sequences of the two groups of data at multiple time scales are respectively obtained, which reflect the uncertainty of the data at each scale. For clarity, define:

[0088] Let represent the Shannon entropy distribution sequence of the historical operation scenario data after signal processing at M different time scales; let represent the Shannon entropy distribution sequence of the current industrial equipment working condition data after signal processing at the same M time scales.

[0089] In the above definition, represents the Shannon entropy value of the historical operation scenario data at the m-th time scale; Denotes the Shannon entropy value of the current industrial equipment operating condition data at the m-th time scale; M represents the total number of time scales, which depends on the number of scales selected in the multi-scale decomposition process; m represents the index of different time scales.

[0090] The Shannon entropy is calculated here according to the formula where E represents the Shannon entropy value; p r denotes the probability that the r-th state appears in the data distribution at a certain time scale; R represents the total number of states at that time scale; r represents the index of a state in the data distribution at a certain time scale.

[0091] After obtaining the multi-scale Shannon entropy distribution, next, by calculating the mutual information difference between the entropy distributions of the two groups of data at each time scale, the dynamic entropy coupling coefficient is determined to quantify the coupling degree of the two groups of data in terms of information transfer. For this purpose, first, the total mutual information between the two groups of data is defined as follows: where, denotes the total mutual information between the historical operation scenario data and the current industrial equipment operating condition data at all M time scales; I m denotes the mutual information between the two groups of data at the m-th time scale.

[0092] I m The calculation formula of is: where, denotes the probability that the u-th state and the v-th state appear simultaneously in the joint probability distribution of the historical operation scenario data and the current industrial equipment operating condition data at the m-th time scale; denotes the marginal probability of the u-th state of the historical operation scenario data at the m-th time scale; denotes the marginal probability of the v-th state of the current industrial equipment operating condition data at the m-th time scale; U and V respectively represent the total number of possible states of the historical operation scenario data and the current industrial equipment operating condition data at the m-th time scale; u represents the index of a state of the historical operation scenario data at a certain time scale; v represents the index of a state of the current industrial equipment operating condition data at a certain time scale.

[0093] Subsequently, to normalize the mutual information, the average Shannon entropy values of the historical operation scenario data and the current industrial equipment operating condition data are calculated respectively, denoted as <H (X) > and <H (Y) >.

[0094] The dynamic entropy coupling coefficient is defined as: where Θ represents the dynamic entropy coupling coefficient, used to quantify whether there is a significant deviation in the information entropy distribution between the two groups of data; max(<H (X) >, <H (Y)(>) represents the larger value of the average Shannon entropy values of the historical operation scenario data and the current industrial equipment working condition data, which is used for normalizing mutual information; the lower the value of the dynamic entropy coupling coefficient, the greater the deviation between the two sets of data in terms of information transfer.

[0095] Perform signal decomposition and instantaneous phase extraction operations on the historical operation scenario data and the current industrial equipment working condition data respectively to obtain their respective instantaneous phase sequences, compare the instantaneous phase sequences obtained at each time scale, and calculate the multi-scale phase consistency index. The multi-scale phase consistency index reflects the degree of consistency between the two sets of data in terms of instantaneous phase changes.

[0096] To further reflect the differences in dynamic characteristics between the two sets of data, in this embodiment, signal decomposition operations are performed on the historical operation scenario data and the current industrial equipment working condition data respectively. This operation uses the empirical mode decomposition method to decompose the original data into several intrinsic mode functions. Let the sequence of intrinsic mode functions obtained by decomposing the historical operation scenario data be denoted as And the sequence of intrinsic mode functions obtained by decomposing the current industrial equipment working condition data be denoted as Among them, represents the k-th intrinsic mode function obtained after signal decomposition of the historical operation scenario data, and the function varies with time t; represents the k-th intrinsic mode function obtained after signal decomposition of the current industrial equipment working condition data; K represents the total number of intrinsic mode functions, which reflects the inherent vibration mode of the data; k represents the index of different intrinsic mode functions.

[0097] For each intrinsic mode function, extract its instantaneous phase information through Hilbert transform. Denote the instantaneous phase of the historical operation scenario data under the k-th intrinsic mode function as Denote the instantaneous phase of the current industrial equipment working condition data under the k-th intrinsic mode function as

[0098] Among them, is the instantaneous phase of the k-th intrinsic mode function of the historical operation scenario data at time t; is the instantaneous phase of the k-th intrinsic mode function of the current industrial equipment working condition data at time t.

[0099] Under each intrinsic mode function, define the instantaneous phase difference as: Among them, ΔΦ k (t) represents the absolute difference between the instantaneous phases of the historical operation scenario data and the current industrial equipment working condition data under the k-th intrinsic mode function;

[0100] To quantitatively describe the phase consistency of two sets of data in each modality, a non-linear similarity function is used to calculate the multi-scale phase consistency index. The multi-scale phase consistency index is defined as: where Λ represents the multi-scale phase consistency index, which is used to measure the consistency of historical operation scenario data and current industrial equipment condition data in terms of instantaneous phase; β is a positive adjustment constant used to control the sensitivity of the exponential function, and its unit is the inverse of the phase difference unit; <ΔΦ k (t)>T means that within the predetermined time interval T, the average value of ΔΦ k (t) under the k-th intrinsic mode function.

[0101] Set the lower limit of the dynamic entropy coupling coefficient and the threshold of the multi-scale phase consistency index. When the dynamic entropy coupling coefficient is lower than the lower limit of the dynamic entropy coupling coefficient and the multi-scale phase consistency index is lower than the threshold of the multi-scale phase consistency index, it is determined that there is a negative transfer phenomenon caused by the failure of the machine learning model to adapt to the current industrial equipment condition; otherwise, it is determined that there is no negative transfer phenomenon caused by the failure of the machine learning model to adapt to the current industrial equipment condition.

[0102] Preset a lower limit Θ 0 (where Θ 0 is a positive real number) as one of the judgment criteria. When the calculated Θ is less than Θ 0 , it preliminarily indicates that there is a significant deviation in the information distribution between the two sets of data.

[0103] Preset a threshold Λ 0 (where Λ 0 is a positive real number). When Λ is less than Λ 0 , it indicates that there is a significant difference in the phase characteristics between the historical operation scenario data and the current industrial equipment condition data.

[0104] When and only when Θ < Θ 0 and Λ < Λ 0 , it is determined that there is a negative transfer caused by the machine learning model's incorrect adaptation to the working condition; otherwise, it is determined that there is no negative transfer caused by the machine learning model's incorrect adaptation to the working condition.

[0105] Θ 0 is the preset lower limit of the dynamic entropy coupling coefficient, which is used to represent the minimum allowable coupling degree of information transfer between two sets of data; Λ 0 is the preset threshold of the multi-scale phase consistency index, which is used to represent the minimum allowable consistency level of two sets of data in terms of instantaneous phase.

[0106] Each step operates based on the pre - processed and recorded real - time monitoring data of industrial equipment, ensuring unified data sources and continuous and consistent processing methods; through the above - mentioned multi - angle parameter analysis, the subtle differences in information entropy and instantaneous phase characteristics between historical operation scenarios and current industrial equipment condition data can be fully revealed, and based on this, it can accurately determine whether there is a negative transfer phenomenon caused by the failure of the machine learning model to adapt to the current industrial equipment condition.

[0107] When there is negative transfer, re - assign different weight parameters to each machine learning model so that the machine learning model more suitable for the current working condition obtains a larger proportion, including:

[0108] After there is a negative transfer phenomenon caused by the machine learning model's incorrect adaptation to the working condition, compare the matching degree between the prediction output generated by each pre - trained machine learning model and the pre - processed and recorded real - time monitoring data of industrial equipment according to the dynamic entropy coupling coefficient and multi - scale phase consistency index, and re - adjust the weight parameters of each machine learning model in the overall prediction. Among them, the machine learning model with a higher matching degree between the prediction output and the actual data obtains a higher weight parameter, while the machine learning model with a lower matching degree obtains a lower weight parameter.

[0109] For each machine learning model, calculate the matching score between its prediction output and the pre - processed recorded data. The formula is: where, Q Ω represents the matching score; Ω represents the number of the machine learning model; Ξ represents the index of the data acquisition time point in the time series; N′ represents the total number of time points; Y Ω (Ξ) represents the prediction output of the Ω - th machine learning model at the time point Ξ; Z Ω (Ξ) represents the pre - processed recorded data corresponding to the time point Ξ; Λ Q represents a positive factor for controlling sensitivity, a positive parameter used to adjust the influence degree of the error between the prediction output and the pre - processed recorded data.

[0110] The weight parameters of each machine learning model can be determined by the following formula: where, P Ω represents the weight parameter of the machine learning model; Ψ represents a positive emphasis factor for adjusting the influence degree of the matching score, a positive parameter used to enhance the effect of the difference in the matching scores of each machine learning model on weight allocation; N″ represents the total number of machine learning models participating in the calculation.

[0111] When and only when the negative transfer recognition module determines that there is negative transfer, update the weight parameters of each machine learning model according to the above formula, so that the model with a higher matching score obtains a larger weight, and the model with a lower matching score obtains a smaller weight.

[0112] Identifying the types of equipment failures based on the adjusted weight parameters and predicting the remaining useful life of the equipment, including:

[0113] Performing weighted fusion on the prediction outputs generated by each machine learning model based on the re-adjusted weight parameters to obtain a comprehensive prediction result; among them, the weighted fusion is achieved by accumulating the products of each prediction output and its corresponding weight parameter.

[0114] Based on the re-adjusted weight parameters, perform weighted fusion on the prediction outputs generated by each pre-trained machine learning model within the same time period. Specifically, each machine learning model generates a prediction output for the same batch of pre-processed and recorded real-time monitoring data of industrial equipment. After multiplying each model's prediction output by its corresponding re-adjusted weight parameter, the product results are accumulated in sequence to form a comprehensive prediction result.

[0115] For example, if there are several machine learning models participating in the prediction, the prediction results generated by each machine learning model at a certain moment, after being multiplied by their pre-adjusted weights, are added together to obtain an overall prediction index.

[0116] Analyze the comprehensive prediction result to identify the types of equipment failures and clarify the equipment operating status and the causes of failures.

[0117] After obtaining the comprehensive prediction result, first conduct a detailed data analysis on the comprehensive prediction result to identify the types of equipment failures and clarify the equipment operating status and the causes of failures. Specifically, in this embodiment, methods such as data trend analysis, outlier detection, and historical data comparison are used to compare the changes in the comprehensive prediction result within each time period.

[0118] For example, when the comprehensive prediction result shows a sudden deviation from the normal fluctuation range within a certain time period, by comparing with the actual operating data of the equipment recorded during pre-processing, it can be preliminarily determined that the equipment corresponding to this deviation phenomenon has a failure risk. Further, by comparing the failure records corresponding to similar deviation phenomena in the historical operating scenarios, this embodiment can identify specific failure types, such as mechanical component wear, overload, or temperature abnormality, etc., and at the same time clarify the current operating status of the equipment and the possible causes of failures.

[0119] Predict the remaining useful life of the equipment based on the historical equipment operating data, and determine the future operating cycle and maintenance plan of the equipment.

[0120] After the fault type is identified, the remaining useful life of the equipment is predicted based on historical equipment operation data. The specific steps include: First, collect and organize the operation time and maintenance cycle data of the equipment in different fault states in history; Second, establish the corresponding relationship between the occurrence of faults and the remaining life of the equipment by analyzing the changes in operation indicators before and after the occurrence of equipment faults; Third, predict the current equipment status based on this corresponding relationship to determine the future operation cycle of the equipment; Finally, combine the actual operation conditions of the equipment to formulate a reasonable maintenance plan and overhaul cycle.

[0121] It should be noted that "fault diagnosis and prediction maintenance" in the "fault diagnosis, prediction and maintenance module" is the abbreviation of "fault diagnosis and predictive maintenance".

[0122] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0123] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0124] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here.

[0125] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0126] The modules described as separate components may or may not be physically separated. The components displayed as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0127] In addition, in each embodiment of the present application, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0128] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0129] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0130] Finally, the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An online real-time fault diagnosis and predictive maintenance platform for industrial big data, characterized in that: include: Data acquisition module: receives and records real-time monitoring data from multi-source industrial equipment; Multi-model management module: stores and schedules multiple machine learning models to process real-time monitoring data of industrial equipment; Model conflict detection module: Analyzes the prediction output of each machine learning model for the same industrial equipment real-time monitoring data, and identifies conflicts in prediction results under different working conditions; Negative transfer identification module: After confirming that there is a conflict in the prediction results, it determines whether there is a negative transfer caused by the machine learning model's error in adapting to the working conditions based on the difference analysis between the historical operating scenarios and the current industrial equipment working conditions; Dynamic weight adjustment module: when there is negative migration, different weight parameters are reallocated to each machine learning model, so that the machine learning model that is more suitable for the current working conditions obtains a larger proportion; Fault diagnosis and maintenance module: Identifies equipment failure types based on adjusted weight parameters and predicts the remaining useful life of the equipment.

2. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1 is characterized in that: Receive and record real-time monitoring data from multiple sources of industrial equipment, including: Receive real-time monitoring data from multiple industrial devices, including temperature, pressure, vibration, current, voltage, and equipment status; The received real-time monitoring data of each industrial equipment is pre-processed and recorded.

3. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1 is characterized in that: Store and schedule multiple machine learning models to process real-time monitoring data of industrial equipment, including: Storing multiple machine learning models for processing real-time monitoring data of industrial equipment, the machine learning models are pre-trained before storage to determine corresponding structural parameters and weight parameters, and the storage step saves each machine learning model and its parameters in a stable storage medium; According to the pre-set scheduling criteria, multiple stored machine learning models are scheduled. The scheduling criteria include sorting and selecting multiple machine learning models according to the characteristics of the real-time monitoring data of industrial equipment, processing delay requirements and model prediction performance. The scheduling step calls the selected machine learning models in sequence to process the pre-processed and recorded real-time monitoring data of industrial equipment.

4. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1 is characterized in that: Analyze the prediction output of each machine learning model for the same industrial equipment real-time monitoring data and identify conflicts in prediction results under different operating conditions, including: Obtaining prediction outputs from multiple pre-trained machine learning models, where the prediction outputs are generated based on pre-processed and recorded real-time monitoring data of industrial equipment; Systematically compare and analyze the prediction data output by each machine learning model in the same time period; identify conflicts in prediction results under different industrial equipment conditions based on preset thresholds and consistency criteria; Among them, different industrial equipment operating conditions include changes in equipment load, ambient temperature fluctuations, and changes in equipment operating status.

5. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1 is characterized in that: After confirming that there is a conflict in the prediction results, based on the difference analysis between the historical operation scenarios and the current industrial equipment conditions, determine whether there is a negative migration caused by the machine learning model adapting to the conditions incorrectly, including: Signal processing is performed on the historical operation scenario data and the current industrial equipment operating condition data respectively to obtain their respective multi-scale information entropy distributions, which reflect the complexity characteristics of the data at multiple time scales. Then, the dynamic entropy coupling coefficient is calculated by calculating the mutual information difference between the information entropy distributions of the historical operation scenario data and the current industrial equipment operating condition data at each scale. The dynamic entropy coupling coefficient indicates whether there is a significant deviation in information transmission between the two sets of data. Signal decomposition and instantaneous phase extraction operations are performed on the historical operation scene data and the current industrial equipment operating condition data to obtain their respective instantaneous phase sequences. The instantaneous phase sequences obtained at each time scale are compared and the multi-scale phase consistency index is calculated. The multi-scale phase consistency index reflects the consistency of the two sets of data in terms of instantaneous phase changes. The lower limit of the dynamic entropy coupling coefficient and the threshold of the multi-scale phase consistency index are set. When the dynamic entropy coupling coefficient is lower than the lower limit of the dynamic entropy coupling coefficient and the multi-scale phase consistency index is lower than the threshold of the multi-scale phase consistency index, it is determined that there is a negative migration phenomenon caused by the failure of the machine learning model to adapt to the current operating conditions of the industrial equipment; otherwise, it is determined that there is no negative migration phenomenon caused by the failure of the machine learning model to adapt to the current operating conditions of the industrial equipment.

6. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 5 is characterized in that: The calculation steps of the dynamic entropy coupling coefficient are: Obtain two sets of data, including historical operation scenario data and current industrial equipment operating condition data, and obtain Shannon entropy value sequences of the two sets of data at multiple time scales; The dynamic entropy coupling coefficient is determined by calculating the mutual information difference between the entropy distributions of the two sets of data at each time scale: The total amount of mutual information between the two sets of data is as follows: in, represents the total mutual information of two sets of data at all M time scales, I m represents the mutual information between two sets of data at the mth time scale, M represents the total number of time scales, and m represents the index of different time scales; Calculate the average Shannon entropy of the two sets of data respectively, and record them as <H (X) > and <H (Y) > The dynamic entropy coupling coefficient is defined as: Among them, Θ represents the dynamic entropy coupling coefficient, max( <H (X) >, <H (Y) >) indicates the larger value of the average Shannon entropy in the two sets of data.

7. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 5 is characterized in that: The calculation steps of the multi-scale phase consistency index are: Obtaining an intrinsic modal function sequence obtained by decomposing historical operation scenario data and an intrinsic modal function sequence obtained by decomposing current industrial equipment operating condition data; For each intrinsic mode function, its instantaneous phase information is extracted through Hilbert transform; The instantaneous phase difference is calculated under each intrinsic mode function. The instantaneous phase difference is the absolute difference between the instantaneous phase of the historical operation scenario data and the current industrial equipment operating condition data. A nonlinear similarity function was used to calculate the multiscale phase consistency index.

8. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1 is characterized in that: When negative transfer exists, different weight parameters are reallocated to each machine learning model so that the machine learning model that is more suitable for the current working condition gets a larger proportion, including: When there is a negative migration phenomenon caused by the error in the machine learning model adapting to the working conditions, the matching degree between the predicted output generated by each pre-trained machine learning model and the pre-processed recorded real-time monitoring data of industrial equipment is compared according to the dynamic entropy coupling coefficient and the multi-scale phase consistency index, and the weight parameters of each machine learning model in the overall prediction are readjusted. Among them, the machine learning model with a higher degree of matching between the predicted output and the actual data obtains a higher weight parameter, while the machine learning model with a lower degree of matching obtains a lower weight parameter.

9. The online real-time fault diagnosis and predictive maintenance platform for industrial big data according to claim 1, characterized in that: Identify equipment failure types based on adjusted weight parameters and predict the remaining useful life of the equipment, including: Based on the readjusted weight parameters, the prediction outputs generated by each machine learning model are weighted and fused to obtain a comprehensive prediction result; wherein the weighted fusion is achieved by multiplying and accumulating each prediction output and its corresponding weight parameter; Analyze the comprehensive prediction results to identify the type of equipment failure, and clarify the equipment operating status and the cause of the failure; Based on historical equipment operation data, the remaining useful life of the equipment is predicted to determine the future operation cycle and maintenance plan of the equipment.

Images