Fault root cause analysis method based on multi-source equipment data fusion

Abstract

The invention discloses a fault root cause analysis method based on multi-source equipment data fusion. According to the method, multi-source heterogeneous data such as equipment operation data, maintenance records and environment data are collected and efficiently fused, a fault mode is automatically recognized from fused data, and time sequence evolution characteristics related to faults are mined. And based on the established knowledge base of the fault modes and reasons, and in combination with interactive Bayesian network reasoning and manual feedback, the root cause of the fault is deduced. According to the method, innovative technologies such as active learning and transfer learning are introduced, and the intelligent level, accuracy and efficiency of fault diagnosis are improved. And a visual interface visually presents an analysis process, supports man-machine interaction correction, and enhances interpretability. The method is suitable for the fields of industrial equipment, transportation tools, medical equipment and the like, and the system stability and usability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a fault root cause analysis method based on multi-source equipment data fusion. Background Art

[0002] Equipment fault diagnosis is a key link to ensure the normal operation of the system. Once a device fails, whether the root cause can be quickly and accurately located directly affects the timely repair of the fault and the availability of the system.

[0003] At present, equipment fault diagnosis mainly relies on expert experience and rule systems. Experts rely on their rich field experience and information such as maintenance records to gradually narrow down the scope of fault causes through analysis and elimination. Rule-based expert systems attempt to formalize expert knowledge and infer possible fault causes through rule matching. However, these traditional methods face many challenges such as low data utilization, poor analysis efficiency, and poor interpretability.

[0004] With the development of emerging technologies such as the Industrial Internet and the Internet of Things, a large number of sensors have been integrated into equipment, generating massive amounts of operating data. These multi-source heterogeneous data provide new opportunities for fault diagnosis, but also put forward higher requirements for analysis methods. How to efficiently integrate the rich information in multi-source data, automatically identify complex fault modes, and infer accurate root causes has become a key technical problem that needs to be solved urgently.

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention proposes an innovative fault root cause analysis method based on multi-source equipment data fusion, aiming to improve the intelligence level of fault diagnosis and analysis accuracy. Summary of the invention

[0006] In order to solve the problems of the prior art, an embodiment of the present invention provides a method for analyzing the root cause of a fault based on multi-source device data fusion. The technical solution is as follows:

[0007] On the one hand, a fault root cause analysis method based on multi-source equipment data fusion is provided, including:

[0008] Collecting multi-source operation data from the equipment, the multi-source operation data including equipment operation data, maintenance record data and environmental data;

[0009] Preprocessing the multi-source operation data to eliminate noise, outliers and missing values, and standardize data format, dimension and sampling frequency;

[0010] Performing multi-source data fusion on the pre-processed multi-source operation data to fuse key features of different data sources and generate a fused feature vector;

[0011] Based on the fused feature vector, identifying potential failure modes through a pattern recognition technique combining unsupervised learning and active learning;

[0012] Performing time series pattern mining on the identified fault mode to discover time series evolution characteristics related to the fault;

[0013] Establishing a knowledge base between fault modes and fault causes, and matching the identified fault modes and the discovered timing characteristics with the knowledge base to obtain possible fault causes;

[0014] An interactive Bayesian network reasoning framework is introduced to infer the root cause of the fault by taking the possible causes of the fault obtained as a priori and combining it with manual feedback;

[0015] The root cause of the fault obtained by the reasoning is presented in a visual form for human-computer interaction and correction.

[0016] Preferably, the multi-source data fusion includes: extracting features from time series data based on time-frequency analysis techniques such as wavelet transform; extracting features from discrete state data based on embedded coding techniques such as Word2vec; extracting features from structured data based on unsupervised learning techniques such as autoencoder networks;

[0017] The extracted features of each source data are concatenated to form a fused feature vector.

[0018] Preferably, the active learning-based pattern recognition includes: selecting unlabeled samples that exhibit the greatest distinguishing ability; presenting the selected unlabeled samples to manual labeling; and iteratively training a fault mode supervision classifier using the labeled samples as supervision signals.

[0019] Preferably, the time series pattern mining includes: using time-frequency analysis techniques such as wavelet transform to discover local time series patterns in the data; using a sequence model based on an attention mechanism to discover long-range time series evolution characteristics in the data.

[0020] Preferably, the establishment of a knowledge base between fault modes and fault causes includes: marking the corresponding relationship between fault modes and causes of historical fault cases and expert knowledge as knowledge base items;

[0021] Through technologies such as machine learning, the correspondence between potential failure modes and causes is automatically discovered and added to the knowledge base.

[0022] Preferably, the method further comprises: introducing transfer learning and domain adaptation technology, applying the established general fault knowledge base to a new equipment field, so as to improve the applicability of the fault analysis model in the new field.

[0023] Preferably, the visual presentation of the root cause of the fault includes: intuitively displaying the fault analysis process and reasoning results; and visually displaying key evidence and basis supporting the inferred root cause of the fault.

[0024] Preferably, the human-computer interaction correction includes: receiving manual feedback on the inferred root cause of the fault; using the feedback as new evidence, inputting it into the interactive Bayesian network, and re-inferring the root cause of the fault.

[0025] Preferably, the multi-source data fusion is performed at the decision level based on an evidence theory framework, and the fault probability generated by each source data is used as evidence to obtain the overall fault probability distribution through fusion.

[0026] Preferably, the method is used to diagnose faults of industrial equipment, transportation vehicles or medical equipment.

[0027] The technical solution provided by the embodiment of the present invention has the following beneficial effects:

[0028] 1. By integrating multi-source heterogeneous data such as equipment operation data, maintenance records, and environmental data, the information sources and analysis dimensions of fault analysis are comprehensively improved, thereby greatly improving the accuracy and comprehensiveness of fault diagnosis.

[0029] 2. Using advanced multi-source data fusion technology, the key features of heterogeneous data sources are efficiently integrated, reducing data noise and redundancy and improving feature quality.

[0030] 3. The integration of unsupervised learning and active learning techniques can automatically discover potential failure modes, reduce dependence on manual labeling, and reduce the cost of knowledge acquisition.

[0031] 4. The introduction of the time series pattern mining module can automatically discover the time series evolution characteristics related to faults, enhancing the analysis ability of continuous process system faults.

[0032] 5. Establish a knowledge base of fault modes and causes, and automatically discover new mode-cause correspondences through machine learning, continuously expand the knowledge base, and improve the completeness of fault knowledge.

[0033] 6. The use of interactive Bayesian network reasoning can effectively integrate automatic reasoning results and manual experience feedback to form a closed-loop optimization and improve the accuracy and explainability of reasoning.

[0034] 7. Visually and intuitively display the fault analysis process, key evidence and reasoning results, and support human-computer interaction correction, enhancing the system's interpretability and user trust.

[0035] 8. By introducing transfer learning and domain adaptation technology, the general fault knowledge base can be quickly applied to new fields, reducing the cost of knowledge acquisition and model training, and enhancing the versatility and deployment efficiency of the system.

[0036] In short, the present invention comprehensively improves the intelligence level, analysis accuracy and analysis efficiency of fault diagnosis, is applicable to various complex systems such as industrial equipment, transportation vehicles, medical equipment, etc., and can bring important value to improving system stability and availability. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 The present invention is a flowchart of a method for analyzing root causes of faults based on multi-source device data fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] The present invention discloses a fault root cause analysis method based on multi-source equipment data fusion, aiming to improve the accuracy and efficiency of equipment fault diagnosis. Figure 1 The present invention is described in further detail.

[0041] The present invention first collects multi-source operation data from the equipment, including equipment sensor data, control system data, maintenance record data, environmental data, etc. Then, preprocessing is performed on these raw data to eliminate noise, outliers and missing values, and to unify the data format, dimension and sampling frequency, etc., so as to improve the quality of subsequent processing.

[0042] The pre-processed multi-source operation data will be input into the multi-source data fusion module for feature-level fusion. Specifically, for time series data sources, time-frequency analysis techniques such as wavelet transform are used to extract its features; for discrete state data, embedded coding techniques such as Word2vec are used to encode it into feature vectors; for structured data, unsupervised learning techniques such as autoencoder networks are used to extract features. Then, the extracted features of each source data are spliced ​​to form a fused feature vector. At the same time, the present invention will also adopt a decision-level fusion framework based on evidence theory, taking the fault probability generated by each source data as evidence, and fusing it to obtain the overall fault probability distribution.

[0043] Based on the obtained fused feature vector, the present invention will use a pattern recognition technology that combines unsupervised learning and active learning to identify potential fault modes. Specifically, firstly, based on unsupervised learning models such as deep autoencoder networks, the inherent pattern distribution is mined from the fused feature vector, and clustering algorithms are used to group them into preliminary fault mode families. Then, an active learning strategy is adopted to select unlabeled samples that show the greatest distinguishing ability and present them to manual labeling. The labeling results are used as supervision signals to iteratively train the supervised classifier of the fault mode.

[0044] For the identified fault mode, the present invention will further perform time series pattern mining to discover the time series evolution characteristics related to the fault. This includes using time-frequency analysis techniques such as wavelet transform to discover local time series patterns in the data, and using sequence models based on attention mechanisms to discover long-range time series evolution laws.

[0045] The present invention establishes a knowledge base between fault modes and fault causes. The knowledge base items include historical fault cases and the correspondence between fault modes and causes in expert knowledge. At the same time, it also uses machine learning and other technologies to automatically discover the correspondence between potential fault modes and causes and add them to the knowledge base. The identified fault modes and discovered timing features are matched with the knowledge base to obtain the possible causes of the faults.

[0046] In order to further infer the root cause of the fault, the present invention introduces an interactive Bayesian network reasoning framework. Taking the possible causes of the fault as a priori, combined with manual feedback, the new evidence is integrated into the Bayesian network, and the root cause analysis results are dynamically adjusted and inferred.

[0047] The present invention presents the root cause of the fault obtained by reasoning in a visual form, showing the fault analysis process and reasoning results, and visually presents the key evidence and basis supporting the root cause of the fault. Users can give feedback on the results, and these feedbacks will be used as new evidence to input into the interactive Bayesian network to re-reason and correct the root cause of the fault.

[0048] In addition, the present invention also introduces transfer learning and domain adaptation technology to apply the established general fault knowledge base to new equipment fields to improve the applicability and deployment efficiency of the fault analysis model in new fields.

[0049] This invention significantly improves the intelligence and accuracy of fault diagnosis through the close integration of innovative multi-source data fusion, pattern recognition combining unsupervised and active learning, time series mining, interactive reasoning and other technical modules. It can effectively overcome the defects of existing technologies such as the use of a single data source, limited pattern recognition capabilities, and poor human-computer interaction, and provide efficient and reliable fault diagnosis solutions for industrial equipment, transportation vehicles, medical equipment and other fields.

[0050] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for analyzing the root cause of a fault based on multi-source equipment data fusion, characterized in that: include: Collecting multi-source operation data from the equipment, the multi-source operation data including equipment operation data, maintenance record data and environmental data; Preprocessing the multi-source operation data to eliminate noise, outliers and missing values, and standardize data format, dimension and sampling frequency; Performing multi-source data fusion on the pre-processed multi-source operation data to fuse key features of different data sources and generate a fused feature vector; Based on the fused feature vector, identifying potential failure modes through a pattern recognition technique combining unsupervised learning and active learning; Performing time series pattern mining on the identified fault mode to discover time series evolution characteristics related to the fault; Establishing a knowledge base between fault modes and fault causes, and matching the identified fault modes and the discovered timing characteristics with the knowledge base to obtain possible fault causes; An interactive Bayesian network reasoning framework is introduced to infer the root cause of the fault by taking the possible causes of the fault obtained as a priori and combining it with manual feedback; The root cause of the fault obtained by the reasoning is presented in a visual form for human-computer interaction and correction.

2. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The multi-source data fusion includes: extracting features from time series data using time-frequency analysis technology based on wavelet transform; extracting features from discrete state data using embedded coding technology based on Word2vec; extracting features from structured data using unsupervised learning technology based on autoencoder network; The extracted features of each source data are concatenated to form a fused feature vector.

3. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The active learning-based pattern recognition includes: selecting unlabeled samples that show the greatest distinguishing ability; presenting the selected unlabeled samples to manual labeling; and using the labeled samples as supervision signals to iteratively train a fault mode supervision classifier.

4. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The time series pattern mining includes: using the time-frequency analysis technology of wavelet transform to discover local time series patterns in the data; using a sequence model based on the attention mechanism to discover long-range time series evolution characteristics in the data.

5. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The establishment of a knowledge base between fault modes and fault causes includes: marking the corresponding relationship between fault modes and causes of historical fault cases and expert knowledge as knowledge base items; Through machine learning technology, the correspondence between potential failure modes and causes is automatically discovered and added to the knowledge base.

6. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The method further includes: introducing transfer learning and domain adaptation technology to apply the established general fault knowledge base to a new equipment field to improve the applicability of the fault analysis model in the new field.

7. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The visual presentation of the root cause of the fault includes: intuitively displaying the fault analysis process and reasoning results; and visually displaying key evidence and basis supporting the inferred root cause of the fault.

8. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The human-computer interaction correction includes: receiving manual feedback on the inferred root cause of the fault; using the feedback as new evidence, inputting it into the interactive Bayesian network, and re-inferring the root cause of the fault.

9. The method for analyzing the root cause of a fault according to claim 1, characterized in that: The multi-source data fusion is based on the framework of evidence theory to perform decision-level fusion, taking the fault possibility generated by each source data as evidence and fusing them to obtain the overall fault possibility distribution.

10. The method for analyzing the root cause of a fault according to any one of claims 1 to 9, characterized in that: The method is used for diagnosing faults of industrial equipment, transportation vehicles or medical equipment.

Images