Particle accelerator equipment fault diagnosis and prediction method and system

Abstract

The invention belongs to the field of equipment fault diagnosis, and provides a particle accelerator equipment fault diagnosis prediction method and system, and the method comprises the steps: obtaining a plurality of operation signals of a particle accelerator; extracting energy distribution characteristics of the particle accelerator for the plurality of operation signals based on wavelet packet decomposition to obtain a plurality of operation signal characteristics; on the basis of the multiple operation signal features, a trained reverse rebroadcasting neural network model is used for identifying and predicting the states of the operation signals of the particle accelerator, and multiple operation signal state features are obtained; and carrying out feature fusion on the identified state features of the operation signals, comparing with normal features of the operation signals, and outputting a fault judgment result. According to the invention, the operation fault of the particle accelerator can be effectively identified; and the method does not depend on experience and priori knowledge of operators too much.

Description

Technical Field

[0001] The invention belongs to the technical field of equipment fault diagnosis, and in particular relates to a particle accelerator equipment fault diagnosis prediction method and system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Vacuum, water cooling, power supply system, etc. are key components of particle accelerators. They play an important role in the operation of particle accelerators. The integrity of system equipment directly affects the safety and economy of the entire accelerator operation. Therefore, it is necessary to accurately identify the operating status of related equipment through fault prediction and diagnosis technology.

[0004] The operation control system of the particle accelerator is the core to ensure the stable, efficient and safe operation of the accelerator. The principles of the particle accelerator operation control system can be summarized as follows: Data acquisition and monitoring, sensors and detectors: used to collect status data of various parts of the accelerator, such as temperature, pressure, magnetic field strength, particle flow intensity, etc. Data acquisition module: collect sensor data in real time and transmit it to the central control system through the network. Data processing and analysis, real-time processing: use high-performance computing units to process the collected data in real time to generate control instructions. Algorithms and models: based on physical models and control algorithms, analyze and predict data and adjust accelerator parameters. Control and feedback, actuators: according to the instructions of the central control system, adjust various parameters of the accelerator, such as magnetic field strength, voltage, vacuum degree, etc. Feedback loop: real-time monitoring of adjustment effects, continuous optimization of control parameters through feedback loops, to ensure stable operation of the accelerator. Safety and protection, emergency shutdown system: when abnormalities or faults are detected, the emergency shutdown procedure is automatically triggered to protect the safety of equipment and personnel. Safety monitoring: through a multi-level safety monitoring system, potential safety hazards are discovered and handled in a timely manner.

[0005] At present, particle accelerator operation mainly adopts patrol inspection to identify fault reports, and the operators in the main control room make processing instructions and detect fault signals through real-time monitoring and feedback loops of the control system. The operators in the main control room judge the type of fault based on relevant feedback information and then conduct manual intervention. This is all done by manually mining the shallow characteristics of the signal and then identifying the type for fault diagnosis, which greatly relies on the experience and prior knowledge of the operators. However, according to statistics, human error is one of the factors that have the greatest impact on the safe operation of the system. (Human error: people fail to accurately, appropriately, fully, and acceptably complete tasks within the prescribed performance standards) At the same time, frequent shutdowns and restarts may cause cumulative losses to equipment, reduce the reliability and service life of the overall system, and reduce economic efficiency. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a particle accelerator equipment fault diagnosis prediction method and system, which is used to predict and diagnose accelerator equipment faults and can effectively identify particle accelerator operation faults; it can greatly improve the reliance on the experience and prior knowledge of the operating personnel.

[0007] According to some embodiments, a first solution of the present invention provides a particle accelerator equipment fault diagnosis and prediction method, which adopts the following technical solution:

[0008] A particle accelerator equipment fault diagnosis and prediction method, comprising:

[0009] Acquire multiple operation signals of particle accelerator;

[0010] For multiple operation signals, the energy distribution characteristics of the particle accelerator are extracted based on wavelet packet decomposition to obtain multiple operation signal characteristics;

[0011] Based on multiple operation signal features, the trained reverse propagation neural network model is used to identify and predict the state of the particle accelerator's operation signal, and multiple operation signal state features are obtained;

[0012] The identified multiple operating signal status features are fused and compared with the normal features of the operating signal to output the fault judgment result.

[0013] Furthermore, the energy distribution feature extraction of the particle accelerator is performed based on wavelet packet decomposition for the multiple operation signals to obtain multiple operation signal features, specifically:

[0014] Perform multi-layer wavelet packet decomposition on a single running signal;

[0015] Reconstruct the wavelet packet decomposition coefficients to extract the signals in each frequency band, i.e., the signals in each frequency band;

[0016] Calculate the power of each frequency band signal and solve the energy of each frequency band signal;

[0017] Setting a signal threshold, eliminating interference components of the frequency band signals based on the signal threshold, and obtaining processed signals of each frequency band;

[0018] Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to reconstruct the processed signals of each frequency band.

[0019] The reconstructed signal is decomposed from low to high frequency, and the mode based on the maximum characteristic energy ratio coefficient value is compared with the mode of the second largest characteristic energy ratio coefficient. The maximum value of the comparison result represents the best decomposition result, and the corresponding operation signal characteristics are obtained.

[0020] Furthermore, the multi-layer wavelet packet decomposition of the single running signal is specifically performed as follows:

[0021] Perform three-layer wavelet packet decomposition on vibration signals, hydraulic signals, sound signals, temperature signals, and vacuum gauge signals;

[0022] Perform four-layer wavelet packet decomposition on motion signals and current signals.

[0023] Furthermore, based on the norm inequality and the Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to reconstruct the processed signals of each frequency band, specifically:

[0024] Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and empirical wavelet function are defined;

[0025] Determine the detail coefficient according to the inner product of the empirical wavelet function and the operating signal, and determine the approximation function according to the inner product of the empirical scaling function and the operating signal;

[0026] The processed signals of each frequency band are reconstructed based on the approximation function and the detail coefficients to obtain a reconstructed signal.

[0027] Furthermore, the reconstructed signal is specifically:

[0028]

[0029] in, is the approximation function, Experience is a scaling function. is the detail function, ψ k (t) is the kth term of the empirical wavelet function, ω represents the frequency domain variable; ω n Represents the nth component of the upper boundary of the frequency domain; and is the Fourier transform of the empirical wavelet function and the empirical scaling function.

[0030] Furthermore, the training process of the reverse broadcast neural network model is specifically as follows:

[0031] Based on the signal extracted from the particle accelerator features as input, normal operation and abnormal operation as output, the initial back propagation neural network model is iteratively trained after preprocessing the input and output;

[0032] Select BP neural networks with different numbers of layers for training respectively, use hidden nodes to select transfer functions, and determine the best BP neural network structure;

[0033] In the process of iterative training, the mean square error is used to represent the error of the vector, and the LM algorithm is used to correct the connection weights in the training process until the training accuracy reaches the set value, thereby obtaining a trained back-propagation neural network model.

[0034] Furthermore, the identified multiple operating signal status features are fused, compared with the normal operating signal features, and the fault determination result is output, specifically:

[0035] The running signal state features corresponding to different signals are spliced ​​to form a comprehensive feature vector, and the spliced ​​comprehensive feature vector is subjected to dimensionality reduction processing to obtain a comprehensive fusion feature vector;

[0036] Based on the comprehensive fusion feature vector and compared with the normal characteristics of the operating signal, the final fault judgment result is obtained.

[0037] According to some embodiments, a second solution of the present invention provides a particle accelerator equipment fault diagnosis and prediction system, which adopts the following technical solution:

[0038] A particle accelerator equipment fault diagnosis and prediction system, comprising:

[0039] A signal acquisition module, configured to acquire a plurality of operation signals of a particle accelerator;

[0040] A signal feature extraction module is configured to extract energy distribution features of a particle accelerator based on wavelet packet decomposition for a plurality of operation signals to obtain a plurality of operation signal features;

[0041] The signal state recognition module is configured to identify and predict the state of the particle accelerator's operating signal based on multiple operating signal features using a trained reverse propagation neural network model to obtain multiple operating signal state features;

[0042] The state feature fusion diagnosis module is configured to fuse the identified multiple operation signal state features to obtain the operation signal fusion feature, compare the operation signal state feature with the operation signal normal feature, and output the fault judgment result.

[0043] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium.

[0044] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in a particle accelerator equipment fault diagnosis and prediction method as described in the first aspect above.

[0045] According to some embodiments, a fourth aspect of the present invention provides a computer device.

[0046] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the particle accelerator device fault diagnosis and prediction method as described in the first aspect above are implemented.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention can effectively identify particle accelerator operation failures; it can greatly improve the reliance on the experience and prior knowledge of operators, which is conducive to improving the operation of particle accelerators affected by human errors. To a certain extent, it can reduce the cumulative loss and economy of the accelerator caused by the start and stop of the particle accelerator due to insufficient inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0050] Figure 1 It is a flow chart of a particle accelerator equipment fault diagnosis and prediction method in an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of signal wavelet packet decomposition in an embodiment of the present invention;

[0052] Figure 3 Schematic diagram of multi-signal data fusion in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0054] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0055] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0056] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0057] Embodiment 1

[0058] like Figure 1 As shown, this embodiment provides a particle accelerator equipment fault diagnosis and prediction method. This embodiment uses the method applied to a server as an example for illustration. It can be understood that the method can also be applied to a terminal, and can also be applied to a terminal, a server, and a system, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, and this application is not limited here. In this embodiment, the method includes the following steps:

[0059] Acquire multiple operation signals of particle accelerator;

[0060] For multiple operation signals, the energy distribution characteristics of the particle accelerator are extracted based on wavelet packet decomposition to obtain multiple operation signal characteristics;

[0061] Based on multiple operation signal features, the trained reverse propagation neural network model is used to identify and predict the state of the particle accelerator's operation signal, and multiple operation signal state features are obtained;

[0062] The identified multiple operating signal status features are fused and compared with the normal features of the operating signal to output the fault judgment result.

[0063] The plurality of operation signals include a vibration signal, a hydraulic signal, a sound signal, a temperature signal, a vacuum gauge signal, a current signal, and a motion signal.

[0064] The energy distribution feature extraction of the particle accelerator is performed based on wavelet packet decomposition for multiple operation signals to obtain multiple operation signal features, specifically:

[0065] Perform multi-layer wavelet packet decomposition on a single running signal;

[0066] Reconstruct the wavelet packet decomposition coefficients to extract the signals in each frequency band, i.e., the signals in each frequency band;

[0067] Calculate the power of each frequency band signal and solve the energy of each frequency band signal;

[0068] Setting a signal threshold, eliminating interference components of the frequency band signals based on the signal threshold, and obtaining processed signals of each frequency band;

[0069] Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to reconstruct the processed signals of each frequency band.

[0070] The reconstructed signal is decomposed from low to high frequency, and the mode based on the maximum characteristic energy ratio coefficient value is compared with the mode of the second largest characteristic energy ratio coefficient. The maximum value of the comparison result represents the best decomposition result, and the corresponding operation signal characteristics are obtained.

[0071] Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to perform signal reconstruction, specifically:

[0072] Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and empirical wavelet function are defined;

[0073] Determine the detail coefficient according to the inner product of the empirical wavelet function and the operating signal, and determine the approximation function according to the inner product of the empirical scaling function and the operating signal;

[0074] The processed signals of each frequency band are reconstructed based on the approximation function and the detail coefficients to obtain a reconstructed signal.

[0075] The reconstructed signal is specifically:

[0076]

[0077] in, is the approximation function, Experience is a scaling function. is the detail function, ψ k (t) is the kth term of the empirical wavelet function, ω represents the frequency domain variable; ω n Represents the nth component of the upper boundary of the frequency domain; and is the Fourier transform of the empirical wavelet function and the empirical scaling function.

[0078] The multi-layer wavelet packet decomposition of a single running signal is specifically as follows:

[0079] Perform three-layer wavelet packet decomposition on vibration signals, hydraulic signals, sound signals, temperature signals, and vacuum gauge signals;

[0080] Perform four-layer wavelet packet decomposition on motion signals and current signals.

[0081] The training process of the reverse-transmission neural network model is specifically as follows:

[0082] Based on the signal extracted from the particle accelerator features as input, normal operation and abnormal operation as output, the initial back propagation neural network model is iteratively trained after preprocessing the input and output;

[0083] Select BP neural networks with different numbers of layers for training respectively, use hidden nodes to select transfer functions, and determine the best BP neural network structure;

[0084] In the process of iterative training, the mean square error is used to represent the error of the vector, and the LM algorithm is used to correct the connection weights in the training process until the training accuracy reaches the set value, thereby obtaining a trained back-propagation neural network model.

[0085] The method described in this embodiment specifically includes the following three aspects:

[0086] 1. Energy distribution feature extraction based on wavelet packet decomposition

[0087] The specific model is: x(t) = s(t) + n(t), where x(t) is the received signal, s(t) is the signal to be detected, and n(t) is a zero-mean Gaussian white noise with a variance of σ. 2 The specific steps are as follows:

[0088] 1) First, perform multi-layer wavelet packet decomposition on the original A / D sampling signal. The determination of the number of decomposition layers should be analyzed according to the specific situation. Taking three layers as an example, the signal characteristics of the eight frequency components from low frequency to high frequency in the third layer need to be extracted respectively. It should be noted that the method described in this embodiment performs three-layer wavelet packet decomposition on vibration, hydraulic pressure, sound, temperature, and vacuum gauge sensor signals; and performs four-layer wavelet packet decomposition on motion sensor signals and current signals. Figure 2 shown.

[0089] 2) Reconstruct the wavelet packet decomposition coefficients and extract the signals in each frequency band. It is understandable that the wavelet packet decomposition coefficients can be reconstructed by conventional means in the prior art, which will not be described in detail or limited in detail.

[0090] 3) Find the energy of the signal in each frequency band.

[0091] Calculate the power P of each frequency band signal. The signal power P is defined as

[0092]

[0093] Among them, signal power P, ω is a frequency domain variable, X T (ω) is the Fourier amplitude spectrum in the frequency domain.

[0094] Then, based on the product of signal power and time, the energy of the signal in each frequency band is obtained.

[0095] 4) Introduce signal threshold processing.

[0096] The signal threshold is defined as: η = coefficient × P, where the coefficient can be 1, 2, 3, ..., N.

[0097] Based on the above definition of signal threshold, the corresponding coefficients are selected according to the time sequence of the frequency band signals to set the sizes of different signal thresholds, and then different signal threshold sizes are set for different frequency band signals to eliminate the interference components of the frequency band signals and obtain the processed frequency band signals.

[0098] 5) Signal reconstruction.

[0099] For each frequency band signal after processing, signal reconstruction is performed. The details are as follows:

[0100] Firstly, based on Littlewood-Paley (norm inequality) and Meyer wavelet (Meyer wavelet) construction, the empirical scaling function and empirical wavelet are defined as expressed by equation (2) and equation (3) respectively.

[0101]

[0102]

[0103] Among them, ω represents the frequency domain variable; ω n Represents the nth component of the upper boundary of the frequency domain; and is the Fourier transform of the empirical wavelet function and the empirical scaling function; τ n is the bandwidth parameter; λ n is the frequency division point, defining the boundary of the frequency interval, indicating the center frequency of the nth frequency band; β(x) is the transition function, used to smooth the transition between frequency intervals, and the specific transition function is β(x) = 35x 4 -8x 5 +70x 6 -20x 7 ;

[0104] γ<min[(ω n+1 -ω n ) / (ω n+1 +ω n )].

[0105] Specifically, the detail coefficient Wfe(n,t) is given by the empirical wavelet function ψ n The inner product with the signal x(t) is generated as follows:

[0106]

[0107] Approximation function By empirical scaling function The inner product with the signal x(t) is generated as follows:

[0108]

[0109] in, and is the Fourier transform of the empirical wavelet function and the empirical scaling function, and represents the complex conjugate of the empirical wavelet function and the empirical scaling function. and represents the inverse Fourier transform.

[0110] The frequency band signal after basis pair decomposition is reconstructed to obtain the reconstructed signal x'(t) as follows:

[0111]

[0112] The reconstructed signal x'(t) is decomposed into single-component components of amplitude modulation and frequency modulation from low to high frequencies, x' k (t), k = 0, 1, 2, 3, ..., k, that is, k is an integer greater than or equal to zero, specifically:

[0113]

[0114]

[0115] Where x'0(t) is the single-component component of AM-FM when k is 0, and x' k (t) is the single-component component of AM-FM when k is taken, is the approximation function, Experience is a scaling function. is the kth term of the detail function, ψ k (t) is the kth term of the empirical wavelet function.

[0116] 6) Select the best decomposition based on the characteristic energy ratio ER.

[0117] The characteristic energy ratio ER is as follows:

[0118] ER=(E1+E2+E3+...+E n ) / E (9);

[0119] Among them, E1, E2, E3, ..., E N It is the signal energy of each frequency band, which is represented by the fault characteristic frequency and its harmonics. Here, N corresponds to the coefficient value in the definition of the signal threshold; E is the total energy.

[0120] According to P1, P2, P3, ..., P N The decomposition results of each result are composed of several modes. The mode with the largest ER coefficient value is ER N,max The modal ER with the second largest ER coefficient N,secondmax For comparison, it is expressed as follows:

[0121]

[0122] Among them, A N The maximum value is recorded as A Nmax , indicating the best decomposition result. At this time, ER N,max The mode contains the best fault information.

[0123] The reconstructed signal is decomposed from low to high frequency, and the mode based on the maximum characteristic energy ratio coefficient value is compared with the mode of the second largest characteristic energy ratio coefficient. The maximum value of the comparison result represents the best decomposition result, and the corresponding operation signal characteristics are obtained.

[0124] According to the above steps, the corresponding wavelet packet analysis software was compiled to analyze the relevant sensor signals under the running state of the particle accelerator.

[0125] 2. Identify and predict the status of particle accelerator operation signals based on BP neural network

[0126] First, the reverse propagation neural network model (BP neural network model) is trained as follows:

[0127] 1) Determine the BP neural network structure.

[0128] The original sample signal is collected and feature extracted based on wavelet packet decomposition to obtain the sample signal feature data and the state corresponding to the sample signal as the data set, which is divided into a training set, a validation set and a test set;

[0129] The characteristic data of the sample signal in the training set is used as input, and the state corresponding to the sample signal is used as output to train the BP neural network. The output is either 1 or 0, that is, whether the sample signal is normal or abnormal.

[0130] Specifically, the neuron output is normal operation or abnormal operation, such as the output vibration signal is an abnormal vibration signal. Specifically, the signals include vibration signals, hydraulic signals, sound signals, temperature signals, vacuum gauge signals, motion signals and current signals.

[0131] 2) Determine the number of hidden units and layers to determine the optimal BP structure.

[0132] Based on the training set, BP neural networks with different numbers of layers are selected for training respectively, and the transfer function is selected by using hidden nodes to determine the optimal BP neural network structure.

[0133] In this embodiment, the hidden unit of the structure of the optimal BP neural network model is determined to be 17; the transfer function selected for the hidden node is SIGMOID, and the transfer function for the output layer node is a linear function.

[0134] 3) Iterative training.

[0135] Define mean square error: Among them, E n is the error of the nth characterization vector, t ni is the expected value of the ith output neuron, o ni is the actual value of the i-th output neuron;

[0136] In this method, the initial value is selected as the default value and the training accuracy ε is set to 0.05.

[0137] The BP neural network model is iteratively trained based on the training set until the training accuracy is less than the set threshold and the error of the characterization vector corresponding to the output result is less than the set value, thereby obtaining a preliminarily trained BP neural network model.

[0138] 4) Model training and verification.

[0139] Preprocess the input and output data, including normalization and compression of input data;

[0140] The preliminarily trained BP neural network model is verified based on the validation set. In order to improve the generalization ability of the network, the Early Stopping method is selected to prevent overfitting during the training of the BP neural network model, monitor the validation set error, and terminate the training in advance when the validation error begins to rise; at the same time, the Sigmoid function is used to adjust the parameters to obtain the trained BP neural network model.

[0141] 5) Connection weight correction.

[0142] After obtaining the trained BP neural network model, the LM (Levenberg-Marquardt) method is selected to correct the connection weights with each use of the network model. Experiments have found that the algorithm in this article has the characteristics of fast and accurate.

[0143] 6) BP neural network model test

[0144] The trained BP neural network model is tested using the test set to obtain the state corresponding to the sample signal.

[0145] Based on this, the trained BP neural network model is used to perform state recognition and prediction on the signal features extracted in step 2 to obtain the operating state corresponding to each signal.

[0146] 3. Multi-signal data fusion

[0147] At present, the multi-sensor data fusion model can be divided into three levels: data layer fusion, feature layer fusion, and decision layer fusion. Figure 3 As shown, this paper will use wavelet packet feature energy extraction and BP neural network to fuse the current signal, vibration signal, motion signal, temperature signal, hydraulic signal, vacuum gauge signal and sound signal, and also use the parameter data of the normal operation of the particle accelerator as input data for decision-level fusion.

[0148] The operating signal status features corresponding to each signal identification in the particle accelerator are spliced ​​and fused to obtain a comprehensive fusion feature vector; through decision fusion, the normal operating status corresponding to the normal operating data of each signal in the particle accelerator operating status is comprehensively compared with the comprehensive fusion feature vector to determine the fault type of the particle accelerator.

[0149] The running signal state features corresponding to different signals are spliced ​​to form a comprehensive feature vector, and the spliced ​​comprehensive feature vector is subjected to dimensionality reduction processing to obtain a comprehensive fusion feature vector;

[0150] Based on the comprehensive fusion feature vector and compared with the normal characteristics of the operating signal, the final fault judgment result is obtained.

[0151] 1) Feature splicing and fusion

[0152] The running signal state features corresponding to the signals from different sensors are spliced ​​to form a comprehensive feature vector. This step integrates the features of each sensor into a unified feature space, laying the foundation for subsequent feature fusion and decision fusion.

[0153] According to the application requirements, it is selected whether to perform dimensionality reduction processing on the combined feature vector after splicing to obtain a comprehensive fusion feature vector. For example, dimensionality reduction methods such as principal component analysis (PCA) can be used to reduce redundant features and improve the efficiency and effect of feature fusion.

[0154] 2) Decision Fusion

[0155] Based on the comprehensive fusion feature vector, the comprehensive judgment result is obtained by comparing it with the normal feature of the running signal, which includes but is not limited to the fusion by rule-based fusion or classifier fusion method.

[0156] Rule-based fusion

[0157] Formulate rules to integrate the judgment results of each sensor. For example, you can set the following rules:

[0158] -When both the temperature sensor and the current sensor detect a fault, it is determined to be a system fault.

[0159] -When the humidity sensor detects an abnormality but other sensors are normal, it is determined to be a possible environmental abnormality.

[0160] Classifier fusion

[0161] In order to improve the accuracy of decision-making, the judgment results of each sensor can be integrated by using voting method or weighted average method:

[0162] Voting method: The judgment results of each sensor are combined and the final judgment result is determined by majority voting.

[0163] Weighted average: Different weights are assigned to each sensor according to its importance, and the judgment results are weighted averaged.

[0164] 3) Fault determination and output

[0165] Fault determination

[0166] The final fault judgment is made based on the comprehensive fusion feature vector and comprehensive judgment results.

[0167] A threshold for fault judgment can be set. When the fused features or decision results exceed the threshold, it is judged as a fault.

[0168] Output

[0169] Output the results of fault determination, including detailed information such as fault type, fault severity, and diagnostic suggestions. The results can be output to system maintenance personnel through the display screen, alarm system or network interface.

[0170] Alarm and feedback

[0171] If it is finally determined to be a fault, the alarm mechanism is triggered:

[0172] An alarm is sounded through an audible and visual alarm to alert system maintenance personnel.

[0173] Feedback fault information to the control center for remote monitoring and fault handling.

[0174] Automatically adjust system parameters based on the fault type and try to restore normal operation.

[0175] Embodiment 2

[0176] This embodiment provides a particle accelerator equipment fault diagnosis and prediction system, including:

[0177] A signal acquisition module, configured to acquire a plurality of operation signals of a particle accelerator;

[0178] A signal feature extraction module is configured to extract energy distribution features of a particle accelerator based on wavelet packet decomposition for a plurality of operation signals to obtain a plurality of operation signal features;

[0179] The signal state recognition module is configured to identify and predict the state of the particle accelerator's operating signal based on multiple operating signal features using a trained reverse propagation neural network model to obtain multiple operating signal state features;

[0180] The state feature fusion diagnosis module is configured to fuse the identified multiple operating signal state features, compare them with the normal features of the operating signals, and output a fault judgment result.

[0181] The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment 1. It should be noted that the above modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0182] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0183] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0184] Embodiment 3

[0185] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the particle accelerator device fault diagnosis and prediction method described in the first embodiment are implemented.

[0186] Embodiment 4

[0187] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the particle accelerator device fault diagnosis method described in the first embodiment are implemented.

[0188] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0189] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0190] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0192] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0193] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A particle accelerator equipment fault diagnosis and prediction method, characterized in that: include: Acquire multiple operating signals of particle accelerator; For multiple operation signals, the energy distribution characteristics of the particle accelerator are extracted based on wavelet packet decomposition to obtain multiple operation signal characteristics; Based on multiple operation signal features, the trained reverse propagation neural network model is used to identify and predict the state of the particle accelerator's operation signal, and multiple operation signal state features are obtained; The identified multiple operating signal status features are fused and compared with the normal features of the operating signal to output the fault judgment result.

2. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 1, characterized in that: The energy distribution feature extraction of the particle accelerator is performed based on wavelet packet decomposition for multiple operation signals to obtain multiple operation signal features, specifically: Perform multi-layer wavelet packet decomposition on a single running signal; Reconstruct the wavelet packet decomposition coefficients to extract the signals in each frequency band, i.e., the signals in each frequency band; Calculate the power of each frequency band signal and solve the energy of each frequency band signal; Setting a signal threshold, eliminating interference components of the frequency band signals based on the signal threshold, and obtaining processed signals of each frequency band; Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to reconstruct the processed signals of each frequency band. The reconstructed signal is decomposed from low to high frequency, and the mode based on the maximum characteristic energy ratio coefficient value is compared with the mode of the second largest characteristic energy ratio coefficient. The maximum value of the comparison result represents the best decomposition result, and the corresponding operation signal characteristics are obtained.

3. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 2, characterized in that: The multi-layer wavelet packet decomposition of a single running signal is specifically as follows: Perform three-layer wavelet packet decomposition on vibration signals, hydraulic signals, sound signals, temperature signals, and vacuum gauge signals; Perform four-layer wavelet packet decomposition on motion signals and current signals.

4. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 2, characterized in that: Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and the empirical wavelet function are defined to reconstruct the processed signals of each frequency band, specifically: Based on the norm inequality and Mayer wavelet construction, the empirical scaling function and empirical wavelet function are defined; Determine the detail coefficient according to the inner product of the empirical wavelet function and the operating signal, and determine the approximation function according to the inner product of the empirical scaling function and the operating signal; The processed signals of each frequency band are reconstructed based on the approximation function and the detail coefficients to obtain a reconstructed signal.

5. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 4, characterized in that: The reconstructed signal is specifically: in, is the approximation function, Experience is a scaling function. is the detail function, ψ k (t) is the kth term of the empirical wavelet function, ω represents the frequency domain variable; ω n Represents the nth component of the upper boundary of the frequency domain; and is the Fourier transform of the empirical wavelet function and the empirical scaling function.

6. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 1, characterized in that: The training process of the reverse-transmission neural network model is specifically as follows: Based on the signal extracted from the particle accelerator features as input, normal operation and abnormal operation as output, the initial back propagation neural network model is iteratively trained after preprocessing the input and output; Select BP neural networks with different numbers of layers for training respectively, use hidden nodes to select transfer functions, and determine the best BP neural network structure; In the process of iterative training, the mean square error is used to represent the error of the vector, and the LM algorithm is used to correct the connection weights in the training process until the training accuracy reaches the set value, thereby obtaining a trained back-propagation neural network model.

7. A particle accelerator equipment fault diagnosis and prediction method as claimed in claim 1, characterized in that: The feature fusion of the identified multiple operating signal status features is compared with the normal features of the operating signal, and the fault determination result is output, specifically: The running signal state features corresponding to different signals are spliced ​​to form a comprehensive feature vector, and the spliced ​​comprehensive feature vector is subjected to dimensionality reduction processing to obtain a comprehensive fusion feature vector; Based on the comprehensive fusion feature vector and compared with the normal characteristics of the operating signal, the final fault judgment result is obtained.

8. A particle accelerator equipment fault diagnosis and prediction system, characterized in that: include: A signal acquisition module, configured to acquire a plurality of operation signals of a particle accelerator; A signal feature extraction module is configured to extract energy distribution features of a particle accelerator based on wavelet packet decomposition for a plurality of operation signals to obtain a plurality of operation signal features; The signal state recognition module is configured to identify and predict the state of the particle accelerator's operating signal based on multiple operating signal features using a trained reverse propagation neural network model to obtain multiple operating signal state features; The state feature fusion diagnosis module is configured to fuse the identified multiple operating signal state features, compare them with the normal features of the operating signals, and output a fault judgment result.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in a particle accelerator equipment fault diagnosis and prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the particle accelerator equipment fault diagnosis and prediction method according to any one of claims 1 to 7 are implemented.

Images