Water turbine anomaly detection and fault diagnosis method based on multi-sensor data fusion

Abstract

The invention provides a water turbine anomaly detection and fault diagnosis method based on multi-sensor data fusion, and the method comprises the steps: collecting stress data, audio signals, vibration signals and working condition data during the operation of a water turbine, and carrying out the preprocessing of the collected data, including data cleaning, alignment and normalization; extracting time-domain, frequency-domain and time-frequency-domain features of the preprocessed stress data, audio signals and vibration signals, forming a high-dimensional feature vector in combination with the preprocessed working condition data, and performing dimensionality reduction on the high-dimensional feature vector by using principal component analysis; performing anomaly detection on the dimensionality-reduced high-dimensional feature vector according to a clustering algorithm, a local outlier factor algorithm and an anomaly detection algorithm constructed by an auto-encoder neural network, and generating a comprehensive anomaly score in combination with a preset weight to obtain an anomaly detection result; and when the water turbine is in an abnormal state, the fault type is diagnosed through the similarity between the dimensionality-reduced high-dimensional feature vector and the historical fault feature template.

Description

Technical Field

[0001] The present application relates to the technical field of industrial equipment monitoring, and in particular to a method and device for detecting anomalies and diagnosing faults of a hydraulic turbine based on multi-sensor data fusion. Background Art

[0002] As the core equipment of a hydropower station, the operating status of a turbine directly affects the power generation efficiency and equipment life. In the prior art, a single type of sensor (such as a stress sensor or a vibration sensor) is usually used to monitor the operating status of a turbine. These technologies mainly focus on the collection of a specific parameter and cannot fully reflect the multi-dimensional information during the operation of the turbine. Especially under complex working conditions, such as mechanical wear, abnormal vibration or hydraulic system failure, it is difficult to detect abnormalities and diagnose faults in a timely and accurate manner by relying only on a single signal.

[0003] In addition, the existing technology often cannot obtain stress, audio, vibration and working condition data at the same time, resulting in a lack of comprehensive monitoring of equipment status. In addition, the monitoring coverage of key parts such as friction cranks and relay cylinders is insufficient, making it difficult to capture subtle anomalies in equipment operation. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first objective of the present application is to propose a method for turbine anomaly detection and fault diagnosis based on multi-sensor data fusion.

[0006] The second objective of the present application is to propose a turbine anomaly detection and fault diagnosis device based on multi-sensor data fusion.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] A fifth object of the present application is to provide a computer program product.

[0010] To achieve the above objectives, the first embodiment of the present application proposes a method for detecting anomalies and diagnosing faults in a hydraulic turbine based on multi-sensor data fusion, comprising:

[0011] Collect stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and pre-process the collected data, including data cleaning, alignment and normalization;

[0012] Extracting the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal, combining the preprocessed working condition data to form a high-dimensional feature vector, and reducing the dimension of the high-dimensional feature vector by principal component analysis;

[0013] The anomaly detection algorithm constructed by the clustering algorithm, the local outlier factor algorithm and the autoencoder neural network is used to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction, and a comprehensive anomaly score is generated by combining the preset weights to obtain the anomaly detection result;

[0014] When the turbine is in an abnormal state, the fault type is diagnosed by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template.

[0015] Optionally, the collection of stress data, audio signals, vibration signals and operating condition data during the operation of the turbine includes:

[0016] At the side end of the turbine friction arm, each friction arm is equipped with a set of stress sensors to monitor the stress data generated during the operation of the friction arm in real time, wherein the stress sensor is fixed by magnetic attraction;

[0017] At the upper end of the turbine friction arm, a set of acoustic vibration temperature sensors is installed every two friction arms to collect audio signals during the operation of the turbine in real time, wherein the acoustic vibration temperature sensors are fixed by magnetic attraction;

[0018] A set of acoustic vibration temperature sensors are installed at the upper end of the turbine friction crank arm and at both ends of the relay cylinder to monitor the vibration signals of the friction crank arm and the relay in real time during operation, wherein the acoustic vibration temperature sensors are fixed by magnetic attraction;

[0019] Install the guide vane opening sensor on the guide vane drive mechanism to collect the guide vane opening data in real time;

[0020] The speed sensor is installed on the turbine shaft to monitor the speed of the turbine shaft in real time;

[0021] The cavity pressure sensor is installed in the cavity position of the hydraulic system of the turbine to monitor the pressure of the hydraulic cavity;

[0022] The power sensor is installed at the output end of the turbine generator to monitor the power output of the turbine.

[0023] Optionally, extracting the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal includes:

[0024] Perform time domain analysis on the preprocessed stress data, audio signal, and vibration signal, and extract the maximum value, minimum value, mean value, variance, peak factor, and kurtosis of the preprocessed stress data, audio signal, and vibration signal;

[0025] Perform frequency domain analysis on the preprocessed stress data, audio signal, and vibration signal, perform fast Fourier transform on the preprocessed stress data, audio signal, and vibration signal, extract the main frequency and bandwidth, and draw a spectrum diagram;

[0026] The preprocessed stress data, audio signal, and vibration signal are analyzed in the time and frequency domains. The dynamic frequency characteristics of the preprocessed stress data, audio signal, and vibration signal are extracted using short-time Fourier transform, and the multi-scale characteristics of the preprocessed stress data, audio signal, and vibration signal are collected using wavelet transform.

[0027] Optionally, the anomaly detection algorithm constructed according to the clustering algorithm, the local outlier factor algorithm and the autoencoder neural network respectively performs anomaly detection on the high-dimensional feature vector after dimensionality reduction, including:

[0028] Construct a historical normal data set through high-dimensional feature vector samples of historical normal operation data;

[0029] The historical normal data is clustered into multiple normal classes using a K-means clustering algorithm to obtain feature class centers of multiple normal states, and the maximum distance between the high-dimensional feature vector after dimensionality reduction and all feature class centers is used as the first abnormal score score1;

[0030] Based on the historical normal data set, a normal data density model is constructed using a local outlier factor algorithm, the outlier degree of the high-dimensional feature vector after dimensionality reduction in the normal data density model is calculated, and the outlier degree is used as a second abnormal score score2;

[0031] The autoencoder neural network is trained according to the historical normal data set to obtain an autoencoder model of normal data, the high-dimensional feature vector after dimensionality reduction is input into the autoencoder model, and a third abnormal score score3 is calculated.

[0032] Optionally, the generating a comprehensive anomaly score by combining preset weights to obtain an anomaly detection result includes:

[0033] The comprehensive anomaly score score is generated according to the following formula:

[0034] score=ω1·score1+ω2·score2+ω3·score3

[0035] Among them, ω1, ω2, and ω3 are algorithm weight parameters;

[0036] The comprehensive abnormality score is compared with a set threshold value, and when the comprehensive abnormality score exceeds the set threshold value, it is determined that the turbine is in an abnormal state.

[0037] Optionally, when the turbine is in an abnormal state, diagnosing the fault type by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template includes:

[0038] Collect historical fault data, and use time domain, frequency domain and time-frequency domain analysis methods to extract fault feature vectors, wherein if the historical fault data is a continuous time process, what is extracted is a set of feature vector sequences with a time series relationship;

[0039] If the fault feature vector is a single feature vector, each feature vector represents a fault type, and the high-dimensional feature vector after dimensionality reduction is calculated with all fault feature vectors for similarity; if the fault feature vector is a set of feature vector sequences, the dynamic time warping algorithm is used to calculate the similarity between the high-dimensional feature vector after dimensionality reduction and the fault feature vector sequence;

[0040] The fault type corresponding to the highest similarity is selected as the diagnosis result. If the highest similarity exceeds the similarity threshold of the corresponding fault type, it is determined that the turbine has a corresponding fault, wherein the similarity thresholds of different fault types are different.

[0041] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a turbine anomaly detection and fault diagnosis device based on multi-sensor data fusion, comprising:

[0042] The data acquisition and preprocessing module is used to collect stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and preprocess the collected data, including data cleaning, alignment and normalization;

[0043] A feature extraction module is used to extract the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal, combine the preprocessed working condition data to form a high-dimensional feature vector, and use principal component analysis to reduce the dimension of the high-dimensional feature vector;

[0044] The anomaly detection module is used to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction according to the clustering algorithm, the local outlier factor algorithm and the anomaly detection algorithm constructed by the autoencoder neural network, and generate a comprehensive anomaly score in combination with the preset weights to obtain the anomaly detection result;

[0045] A fault diagnosis module is used to diagnose the fault type when the turbine is in an abnormal state by comparing the high-dimensional feature vector after dimensionality reduction with the historical fault feature template.

[0046] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0047] The memory stores computer-executable instructions;

[0048] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0049] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer-readable storage medium is stored computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0050] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, which implements any method in the first aspect when executed by a processor.

[0051] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0052] By rationally arranging multiple types of sensors at key parts of the turbine such as the friction arm and the relay cylinder, the comprehensive collection of stress, audio, vibration and working condition data is achieved, avoiding the diagnostic limitations caused by the single monitoring data of a single sensor in the existing technology, improving the comprehensiveness and accuracy of equipment status monitoring, and improving the comprehensiveness and accuracy of equipment operation status monitoring. Through data cleaning, alignment and normalization preprocessing, unified processing and fusion of different types of data are achieved, avoiding the time offset and dimensional inconsistency between multi-source data, and improving the accuracy of data analysis. By using clustering algorithm, local outlier factor algorithm and autoencoder neural network to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction, multi-algorithm fusion analysis of anomaly detection is achieved, avoiding the problem of insufficient recognition ability of a single algorithm for complex fault patterns, and improving the robustness and sensitivity of anomaly detection. By calculating the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template, accurate diagnosis of specific fault types is achieved, avoiding the inaccurate or misjudgment of fault identification in the existing technology, and improving the reliability and practicality of turbine fault diagnosis.

[0053] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0055] Figure 1 A schematic flow chart of a method for detecting and diagnosing abnormalities of a hydraulic turbine based on multi-sensor data fusion provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of a hydraulic turbine anomaly detection and fault diagnosis device based on multi-sensor data fusion provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0058] In view of the technical problems existing in the prior art, the present application provides a method for detecting anomalies and diagnosing faults of a hydraulic turbine based on multi-sensor data fusion. Figure 1 The present invention provides a flow chart of a method for detecting and diagnosing a turbine anomaly based on multi-sensor data fusion. Figure 1 As shown, the method comprises the following steps:

[0059] Step 101, collecting stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and preprocessing the collected data, including data cleaning, alignment and normalization.

[0060] In the embodiment of the present application, the operating condition data includes guide vane opening, open (closed) cavity pressure, power, speed, etc.

[0061] In the embodiment of the present application, stress data, audio signals, vibration signals and operating condition data of the turbine during operation can be collected by installing sensors at different locations of the turbine.

[0062] In one embodiment of the present application, a set of stress sensors is installed on the side end of each friction arm of the turbine to monitor the stress data generated during the operation of the friction arm in real time, wherein the stress sensors are fixed by magnetic attraction.

[0063] In one embodiment of the present application, at the upper end of the turbine friction arm, a set of acoustic vibration and temperature sensors is installed every two friction arms to collect audio signals during the operation of the turbine in real time, wherein the acoustic vibration and temperature sensors are fixed by magnetic attraction.

[0064] In one embodiment of the present application, a set of acoustic vibration and temperature sensors are installed at the upper end of the turbine friction crank arm and at both ends of the relay cylinder to monitor the vibration signals of the friction crank arm and the relay during operation in real time, wherein the acoustic vibration and temperature sensors are fixed by magnetic attraction.

[0065] In one embodiment of the present application, a guide vane opening sensor is installed on the guide vane driving mechanism to collect guide vane opening data in real time.

[0066] In one embodiment of the present application, a rotation speed sensor is installed on the turbine shaft to monitor the rotation speed of the turbine shaft in real time.

[0067] In one embodiment of the present application, a cavity pressure sensor is installed at a cavity position of a hydraulic system of a water turbine to monitor the pressure of the hydraulic cavity.

[0068] In one embodiment of the present application, a power sensor is installed at the output end of the turbine generator to monitor the power output of the turbine.

[0069] It should be noted that the number of different sensors needs to be set according to the actual scenario, and this application does not make any specific limitation on this.

[0070] It is understandable that in order to achieve data consistency and reliability, the collected stress data, audio signals, vibration signals and operating condition data need to be preprocessed, including data cleaning, alignment and normalization.

[0071] In an embodiment of the present application, the step of data cleaning includes: deleting physically impossible abnormal values, such as values ​​in stress data that exceed the sensor range or noise spikes in audio signals; processing missing values ​​in sensor data, replacing missing points by linear interpolation, moving average or historical mean; filtering random noise in the signal, and denoising the signal using low-pass filters, Gaussian smoothing and other techniques.

[0072] The steps of data alignment include: synchronizing the data collected by multiple sensors based on timestamps to ensure that all data are aligned under the same time reference; using interpolation methods to time-align sensor data with different sampling frequencies, for example, interpolating low-frequency sampling data to a time point consistent with high-frequency sampling.

[0073] The step of data normalization includes normalizing different types of data (such as stress values, vibration amplitudes, and audio signal powers) and mapping them to the same dimension range (such as [0, 1] or a range with a mean of 0 and a standard deviation of 1).

[0074] Step 102, extract the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal, combine them with the preprocessed working condition data to form a high-dimensional feature vector, and use principal component analysis to reduce the dimension of the high-dimensional feature vector.

[0075] In the embodiment of the present application, time domain analysis is performed on the preprocessed stress data, audio signal, and vibration signal.

[0076] Specifically, the preprocessed stress data, audio signal and vibration signal are analyzed in the time domain to extract the following features:

[0077] (1) Maximum and minimum values: reflect the amplitude range of the signal and are used to evaluate the extreme conditions of equipment operation;

[0078] (2) Mean: describes the overall level of the signal and reflects the long-term operating trend of the equipment;

[0079] (3) Variance: evaluates the degree of signal fluctuation and is used to detect operational stability;

[0080] (4) Crest factor: measures the fluctuation amplitude of the signal, which is the ratio of the maximum value of the signal to the mean value;

[0081] (5) Kurtosis: reflects the non-Gaussian nature of the signal and is used to identify the impact characteristics of abnormal signals.

[0082] Specifically, the preprocessed stress data, audio signal and vibration signal are subjected to frequency domain analysis. Fast Fourier transform is used to convert the signal from the time domain to the frequency domain, and the spectral characteristics of the signal are extracted, including the following characteristics:

[0083] (1) Main frequency: extract the frequency component with the largest energy in the signal to determine the main vibration during operation;

[0084] (2) Bandwidth: The spectral range of the analyzed signal, used to evaluate the complexity of the frequency distribution;

[0085] (3) Spectrum diagram: It intuitively displays the relationship between frequency and signal amplitude and is used to detect anomalies in specific frequency bands.

[0086] In addition, the preprocessed stress data, audio signals and vibration signals are subjected to time-frequency domain analysis, including the use of short-time Fourier transform to extract the dynamic frequency characteristics of the signal and analyze the frequency change trend over time; and the use of wavelet transform to extract the multi-scale characteristics of the signal and capture the changing characteristics of the signal at different time scales for identifying abnormal local characteristics.

[0087] Finally, the above-extracted time domain, frequency domain and time-frequency domain features are combined with the operating condition data (such as guide vane opening, speed, cavity pressure, power, etc.) to form a high-dimensional feature vector to comprehensively describe the operating status of the equipment.

[0088] In addition, the principal component analysis method is used to reduce the dimension of high-dimensional feature vectors, extract the main components, reduce redundant features, and retain key information. The feature vector after dimension reduction is convenient for subsequent anomaly detection and fault diagnosis, and improves computational efficiency and accuracy.

[0089] Step 103, performing anomaly detection on the high-dimensional feature vector after dimensionality reduction according to the clustering algorithm, the local outlier factor algorithm and the anomaly detection algorithm constructed by the autoencoder neural network, and generating a comprehensive anomaly score in combination with the preset weights to obtain an anomaly detection result.

[0090] In the embodiments of the present application, an anomaly detection model is constructed by analyzing historical normal operation data, and an anomaly score is calculated based on three algorithms to achieve accurate detection of the operating status of the turbine.

[0091] First, high-dimensional feature vector samples are extracted from historical normal operation data through feature extraction and dimensionality reduction methods (such as step 102) to construct a historical normal data set. This data set comprehensively describes the feature distribution of the turbine under various normal states and provides basic training data for the anomaly detection algorithm.

[0092] Then, the following three algorithms are used to calculate the anomaly score for the real-time high-dimensional feature vector after dimensionality reduction, providing a basis for anomaly judgment from multiple angles:

[0093] (1) Use the K-means clustering algorithm to cluster the historical normal data set and divide the normal data into multiple normal state categories; then, calculate the feature class center of each normal class, and these center points represent the feature distribution of different normal states; finally, calculate the maximum distance between the reduced-dimensional feature vector of the real-time monitoring data (i.e., the reduced-dimensional high-dimensional feature vector obtained in step 102) and the center of all feature classes, and use this distance as the first abnormal score score1.

[0094] score1 reflects the degree to which real-time data deviates from the center of all normal states and is suitable for identifying anomalies in discrete distributions.

[0095] (2) Using the local outlier factor algorithm, a density distribution model of normal data is constructed based on the historical normal data set; then the local density of each data point is calculated to analyze its closeness to the data points in the neighborhood; finally, the reduced-dimensional feature vector of the real-time monitoring data is input into the model, its local outlier degree is calculated, and the outlier degree is used as the second anomaly score score2.

[0096] score2 reflects whether the real-time data is a rare point in the normal distribution and is suitable for detecting local anomalies.

[0097] (3) The autoencoder neural network is trained based on the historical normal data set to enable it to learn the feature reconstruction ability of the historical normal data; then the dimension-reduced feature vector of the real-time monitoring data is input into the trained autoencoder model to calculate the reconstruction error; finally, the reconstruction error is used as the third anomaly score score3.

[0098] Score3 reflects the overall difference between real-time data and normal distribution and is suitable for identifying anomalies in complex patterns.

[0099] Finally, the anomaly scores (score1, score2, score3) obtained by the three algorithms are weighted and fused according to the preset weights to generate a comprehensive anomaly score (score), which is expressed as:

[0100] score=ω1·score1+ω2·score2+ω3·score3

[0101] Among them, ω1, ω2, and ω3 are algorithm weight parameters.

[0102] It can be understood that the calculation of the comprehensive anomaly score takes into account the combined effects of global outliers, local density features, and feature reconstruction errors, thereby improving the robustness of anomaly detection.

[0103] In addition, the comprehensive abnormality score is compared with the set threshold, and when the comprehensive abnormality score exceeds the set threshold, it is determined that the turbine is in an abnormal state.

[0104] It is understandable that the threshold needs to be set according to the actual scenario, and this application does not make any specific limitations on this.

[0105] Step 104, when the turbine is in an abnormal state, diagnose the fault type by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template.

[0106] In the embodiment of the present application, fault diagnosis is performed on the characteristic vector of the abnormal state to further analyze the specific type of the abnormality.

[0107] First, historical fault data is collected and fault feature vectors are extracted using time domain, frequency domain and time-frequency domain analysis methods. Depending on whether it is a continuous time process, it is divided into the following two cases:

[0108] (1) Single feature vector: If the fault data is a static feature at a single time point, each fault feature vector corresponds to a fault type;

[0109] (2) Feature vector sequence: If the fault data is a continuous time process, a set of feature vector sequences with a time series relationship is extracted, and each set of sequences corresponds to a fault type.

[0110] Then, according to the historical fault feature type (single feature vector or feature vector sequence), different methods are used to calculate the similarity between the real-time data and the fault template.

[0111] If the fault feature vector is a single feature vector, the real-time high-dimensional feature vector after dimensionality reduction is respectively similar to all historical fault feature vectors, and the cosine similarity or Euclidean distance is used to calculate the similarity between the real-time data and each fault template; in the calculation results, the fault type corresponding to the highest similarity is taken as the diagnosis result.

[0112] If the fault feature vector is a set of feature vector sequences, the real-time high-dimensional feature vector sequence after dimensionality reduction is compared with the historical fault feature vector sequence, and the dynamic time warping (DTW) algorithm is used to calculate the similarity of the two sets of feature vector sequences, and the fault type corresponding to the highest similarity is selected as the diagnosis result.

[0113] It should be noted that the dynamic time warping algorithm can find the best matching path when the sequence lengths are different or the time alignment is inconsistent, ensuring the accuracy of the similarity calculation.

[0114] In the embodiment of the present application, a different similarity threshold is also set for each fault type, which is determined based on historical fault data and actual operating experience.

[0115] It can be understood that when the highest similarity of the real-time data exceeds the similarity threshold of the corresponding fault type, it is determined that the corresponding fault has occurred in the turbine, and a diagnosis result is output, including the fault type, cause of occurrence and recommended maintenance measures.

[0116] In order to implement the above-mentioned embodiment, the present application also proposes a turbine anomaly detection and fault diagnosis device based on multi-sensor data fusion. Figure 2 The structure diagram of a hydraulic turbine abnormality detection and fault diagnosis device 10 based on multi-sensor data fusion provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the device comprises:

[0117] The data acquisition and preprocessing module 100 is used to collect stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and to preprocess the collected data, including data cleaning, alignment and normalization;

[0118] The feature extraction module 200 is used to extract the time domain, frequency domain and time-frequency domain features of the pre-processed stress data, audio signal and vibration signal, combine the pre-processed working condition data to form a high-dimensional feature vector, and use principal component analysis to reduce the dimension of the high-dimensional feature vector;

[0119] Anomaly detection module 300, used to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction according to the clustering algorithm, the local outlier factor algorithm and the anomaly detection algorithm constructed by the autoencoder neural network, and generate a comprehensive anomaly score in combination with preset weights to obtain anomaly detection results;

[0120] Fault diagnosis module 400 is used to diagnose the fault type when the turbine is in an abnormal state by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template.

[0121] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0122] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0123] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0124] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0125] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0126] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0127] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0128] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0129] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0130] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0131] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0132] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0133] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0134] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

[0135] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0136] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-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 in the protection scope of this application.

Claims

1. A method for abnormality detection and fault diagnosis of a hydraulic turbine based on multi-sensor data fusion, characterized in that: The following steps are involved: Collect stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and pre-process the collected data, including data cleaning, alignment and normalization; Extracting the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal, combining the preprocessed working condition data to form a high-dimensional feature vector, and reducing the dimension of the high-dimensional feature vector by principal component analysis; The anomaly detection algorithm constructed by the clustering algorithm, the local outlier factor algorithm and the autoencoder neural network is used to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction, and a comprehensive anomaly score is generated by combining the preset weights to obtain the anomaly detection result; When the turbine is in an abnormal state, the fault type is diagnosed by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template.

2. The method according to claim 1, characterized in that: The collection of stress data, audio signals, vibration signals and operating condition data during the operation of the turbine includes: At the side end of the turbine friction arm, each friction arm is equipped with a set of stress sensors to monitor the stress data generated during the operation of the friction arm in real time, wherein the stress sensor is fixed by magnetic attraction; At the upper end of the turbine friction arm, a set of acoustic vibration temperature sensors is installed every two friction arms to collect audio signals during the operation of the turbine in real time, wherein the acoustic vibration temperature sensors are fixed by magnetic attraction; A set of acoustic vibration temperature sensors are installed at the upper end of the turbine friction crank arm and at both ends of the relay cylinder to monitor the vibration signals of the friction crank arm and the relay in real time during operation, wherein the acoustic vibration temperature sensors are fixed by magnetic attraction; Install the guide vane opening sensor on the guide vane drive mechanism to collect the guide vane opening data in real time; The speed sensor is installed on the turbine shaft to monitor the speed of the turbine shaft in real time; The cavity pressure sensor is installed in the cavity position of the hydraulic system of the turbine to monitor the pressure of the hydraulic cavity; The power sensor is installed at the output end of the turbine generator to monitor the power output of the turbine.

3. The method according to claim 2, characterized in that The extracting of the time domain, frequency domain and time-frequency domain features of the pre-processed stress data, audio signal and vibration signal includes: Perform time domain analysis on the preprocessed stress data, audio signal, and vibration signal, and extract the maximum value, minimum value, mean value, variance, peak factor, and kurtosis of the preprocessed stress data, audio signal, and vibration signal; Perform frequency domain analysis on the preprocessed stress data, audio signal, and vibration signal, perform fast Fourier transform on the preprocessed stress data, audio signal, and vibration signal, extract the main frequency and bandwidth, and draw a spectrum diagram; The preprocessed stress data, audio signal, and vibration signal are analyzed in the time and frequency domains. The dynamic frequency characteristics of the preprocessed stress data, audio signal, and vibration signal are extracted using short-time Fourier transform, and the multi-scale characteristics of the preprocessed stress data, audio signal, and vibration signal are collected using wavelet transform.

4. The method according to claim 3, characterized in that The anomaly detection algorithm constructed according to the clustering algorithm, the local outlier factor algorithm and the autoencoder neural network respectively performs anomaly detection on the high-dimensional feature vector after dimensionality reduction, including: Construct a historical normal data set through high-dimensional feature vector samples of historical normal operation data; The historical normal data is clustered into multiple normal classes using a K-means clustering algorithm to obtain feature class centers of multiple normal states, and the maximum distance between the high-dimensional feature vector after dimensionality reduction and all feature class centers is used as the first abnormal score score1; Based on the historical normal data set, a normal data density model is constructed using a local outlier factor algorithm, the outlier degree of the high-dimensional feature vector after dimensionality reduction in the normal data density model is calculated, and the outlier degree is used as a second abnormal score score2; The autoencoder neural network is trained according to the historical normal data set to obtain an autoencoder model of normal data, the high-dimensional feature vector after dimensionality reduction is input into the autoencoder model, and a third abnormal score score3 is calculated.

5. The method according to claim 4, characterized in that The method of generating a comprehensive anomaly score by combining the preset weights to obtain an anomaly detection result includes: The comprehensive anomaly score score is generated according to the following formula: score=ω1·score1+ω2·score2+ω3·score3 Among them, ω1, ω2, and ω3 are algorithm weight parameters; The comprehensive abnormality score is compared with a set threshold value, and when the comprehensive abnormality score exceeds the set threshold value, it is determined that the turbine is in an abnormal state.

6. The method according to claim 5, characterized in that When the turbine is in an abnormal state, diagnosing the fault type by the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template includes: Collect historical fault data, and use time domain, frequency domain and time-frequency domain analysis methods to extract fault feature vectors, wherein if the historical fault data is a continuous time process, what is extracted is a set of feature vector sequences with a time series relationship; If the fault feature vector is a single feature vector, each feature vector represents a fault type, and the high-dimensional feature vector after dimensionality reduction is calculated with all fault feature vectors for similarity; if the fault feature vector is a set of feature vector sequences, the dynamic time warping algorithm is used to calculate the similarity between the high-dimensional feature vector after dimensionality reduction and the fault feature vector sequence; The fault type corresponding to the highest similarity is selected as the diagnosis result. If the highest similarity exceeds the similarity threshold of the corresponding fault type, it is determined that the turbine has a corresponding fault, wherein the similarity thresholds of different fault types are different.

7. A hydraulic turbine anomaly detection and fault diagnosis device based on multi-sensor data fusion, characterized in that: include: The data acquisition and preprocessing module is used to collect stress data, audio signals, vibration signals and operating condition data of the turbine during operation, and preprocess the collected data, including data cleaning, alignment and normalization; A feature extraction module is used to extract the time domain, frequency domain and time-frequency domain features of the preprocessed stress data, audio signal and vibration signal, combine the preprocessed working condition data to form a high-dimensional feature vector, and use principal component analysis to reduce the dimension of the high-dimensional feature vector; The anomaly detection module is used to perform anomaly detection on the high-dimensional feature vector after dimensionality reduction according to the clustering algorithm, the local outlier factor algorithm and the anomaly detection algorithm constructed by the autoencoder neural network, and generate a comprehensive anomaly score in combination with the preset weights to obtain the anomaly detection result; The fault diagnosis module is used to diagnose the fault type when the turbine is in an abnormal state through the similarity between the high-dimensional feature vector after dimensionality reduction and the historical fault feature template.

8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images