Fault diagnosis method, device, medium and computer program product

Abstract

The invention relates to the technical field of mechanical fault detection, and particularly provides a fault diagnosis method and device, a medium and a computer program product, and the method comprises the steps: obtaining an operation state signal of a mechanical device; preprocessing the operation state signal to obtain operation characteristic data of the mechanical equipment, and performing frequency coding on the operation characteristic data to obtain a peak sequence to be processed; inputting the to-be-processed peak sequence into a liquid machine LSM model, processing the to-be-processed peak sequence based on a pulse time sequence dependent plasticity STDP algorithm, updating the weight between neurons in the LSM model, and outputting a to-be-identified peak sequence; and inputting the to-be-identified peak sequence into the convolutional neural network model for feature identification to obtain a fault diagnosis result of the mechanical equipment. According to the invention, the accuracy and efficiency of fault diagnosis can be improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of mechanical fault detection, and particularly to a fault diagnosis method, device, medium and computer program product. Background Art

[0002] The reliability of mechanical equipment is the foundation of modern industrial systems, directly affecting operational efficiency, safety and cost-effectiveness; as systems become increasingly complex, it is particularly important to perform real-time and reliable detection of mechanical equipment in a dynamic environment.

[0003] In related technologies, based on a neural network model, the operating state signals collected during the operation of mechanical equipment can be processed to determine whether the mechanical equipment has a fault. For example, a spiking neural network (SNN) or a convolutional neural network (CNN) can be used for mechanical fault detection.

[0004] However, in the prior art, the method for detecting mechanical equipment faults based on a spiking neural network still has problems of poor accuracy and low efficiency of detection results. Summary of the Invention

[0005] In view of the above problems, the present disclosure is proposed. The present disclosure provides a fault diagnosis method, device, medium and computer program product; it can improve the accuracy and efficiency of fault diagnosis.

[0006] According to one aspect of the present disclosure, a fault diagnosis method is provided, including:

[0007] Obtaining the operating state signals of mechanical equipment;

[0008] Preprocessing the operating state signals to obtain the operating characteristic data of the mechanical equipment, and performing frequency encoding on the operating characteristic data to obtain a to-be-processed spike sequence;

[0009] Inputting the to-be-processed spike sequence into a liquid state machine LSM model to process the to-be-processed spike sequence based on the spike timing-dependent plasticity STDP algorithm, updating the weights between neurons in the LSM model, and outputting a to-be-identified spike sequence;

[0010] Inputting the to-be-identified spike sequence into a convolutional neural network model for feature recognition to obtain the fault diagnosis result of the mechanical equipment.

[0011] According to another aspect of the present disclosure, an electronic device is provided, including a memory, a processor and a computer program stored on the memory, and the processor executes the computer program to implement the above-mentioned fault diagnosis method.

[0012] According to another aspect of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned fault diagnosis method is implemented.

[0013] According to still another aspect of the present disclosure, there is provided a computer program product including a computer program, and when the computer program is executed by a processor, the above-mentioned fault diagnosis method is implemented.

[0014] For the fault diagnosis method, device, medium, and computer program product according to the embodiments of the present disclosure, on the one hand, an LSM model capable of capturing the temporal dependence of mechanical fault features is used to accurately identify the fault features of mechanical equipment, and a spike sequence to be identified containing richer mechanical equipment fault features is obtained; and during the fault feature extraction process, the weights between neurons in the LSM model are updated based on the STDP algorithm, further preventing the problem of insufficient ability to identify the temporal dependence of fault features during the process of using the LSM model with fixed weights for fault feature identification, so as to further improve the accuracy of the LSM model in identifying fault features; on the other hand, a convolutional neural network is used to identify the fault features to obtain a fault identification result, which can solve the problem of low identification efficiency caused by the double-layer spiking neural network and improve the fault identification efficiency.

[0015] It should be understood that both the foregoing general description and the following detailed description are exemplary and are intended to provide further explanation of the claimed technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present disclosure will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure, and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation to the present disclosure. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 is a flowchart illustrating the fault diagnosis method according to the embodiments of the present disclosure.

[0018] Figure 2 is a schematic diagram illustrating the change of the neuron membrane potential according to the embodiments of the present disclosure.

[0019] Figure 3 is a schematic diagram illustrating the structure of the LSM model according to the embodiments of the present disclosure.

[0020] Figure 4 is a schematic diagram illustrating the adjustment of the synaptic weights of neurons according to the embodiments of the present disclosure.

[0021] Figure 5 It is a schematic circuit diagram showing the LIF neuron model of the embodiments of the present disclosure.

[0022] Figure 6 It is a schematic diagram showing the fault diagnosis process of the embodiments of the present disclosure.

[0023] Figure 7 It is a schematic diagram showing the confusion matrix of a target data set of the embodiments of the present disclosure.

[0024] Figure 8 It is a schematic diagram showing the mapping result of the operation characteristic data of the embodiments of the present disclosure.

[0025] Figure 9 It is a schematic diagram showing the mapping result of the spike train to be recognized of the embodiments of the present disclosure.

[0026] Figure 10 It is a schematic diagram showing the ROC curve and AUC result of the embodiments of the present disclosure.

[0027] Figure 11 It is a block diagram showing the fault diagnosis device of the embodiments of the present disclosure.

[0028] Figure 12 It is a schematic diagram showing the computer program product according to the embodiments of the present disclosure.

[0029] Figure 13 It is a hardware block diagram showing the electronic device of the embodiments of the present disclosure. Detailed implementation manners

[0030] In order to make the objectives, technical solutions and advantages of the present disclosure more apparent, exemplary embodiments according to the present disclosure will be described in detail below with reference to the accompanying drawings. Apparently, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments of the present disclosure. It should be understood that the present disclosure is not limited by the exemplary embodiments described herein.

[0031] In the related art, when applying a spiking neural network for mechanical fault detection, a two-layer spiking neural network can be selected for fault detection of mechanical equipment. Among them, one layer of the spiking neural network is used for fault feature extraction, and the other layer of the spiking neural network is used for fault type diagnosis.

[0032] However, in the solution of mechanical fault diagnosis based on a two-layer spiking neural network, during the feature extraction process, it is difficult for a traditional spiking neural network to capture fault features with time dependence, resulting in poor accuracy of the determined fault diagnosis result. At the same time, the fault diagnosis framework based on a two-layer spiking neural network requires a large amount of computing resources, resulting in low diagnosis efficiency.

[0033] To solve the above problems, an embodiment of the present disclosure provides a fault diagnosis method, which can be applied to a terminal device. The terminal device can be an electronic device such as a computer, a notebook, or a server. For example, Figure 1 As shown, the fault diagnosis method includes:

[0034] Step S101, obtaining the operating state signal of the mechanical equipment;

[0035] Step S102, preprocessing the operating state signal to obtain the operating characteristic data of the mechanical equipment, and frequency encoding the operating characteristic data to obtain a spike sequence to be processed;

[0036] Step S103, inputting the spike sequence to be processed into a liquid state machine LSM model, processing the spike sequence to be processed based on the spike-timing-dependent plasticity STDP algorithm, updating the weights between neurons in the LSM model, and outputting a spike sequence to be recognized;

[0037] Step S104, inputting the spike sequence to be recognized into a convolutional neural network model for feature recognition to obtain the fault diagnosis result of the mechanical equipment.

[0038] In summary, for the fault diagnosis method provided by the embodiment of the present disclosure, on the one hand, an LSM model capable of capturing the temporal dependence of mechanical fault characteristics is used to accurately identify the mechanical equipment fault characteristics, and a spike sequence to be recognized containing richer mechanical equipment fault characteristics is obtained; and during the fault feature extraction process, the weights between neurons in the LSM model are updated based on the STDP algorithm, further preventing the problem of insufficient ability to recognize the temporal dependence of fault characteristics during the process of using the LSM model with fixed weights for fault feature recognition, so as to further improve the recognition accuracy of the LSM model for fault characteristics; on the other hand, through the convolutional neural network, the fault features are recognized to obtain the fault recognition result, which can solve the problem of low recognition efficiency caused by the double-layer spiking neural network and improve the fault recognition efficiency.

[0039] The following elaborates in detail on the specific implementation manners of each step in the Figure 1 shown embodiment:

[0040] In step S101, the terminal device obtains the operating state signal of the mechanical equipment.

[0041] In the embodiment of the present disclosure, the operating state signal of the mechanical equipment can be the operating state signal of the entire mechanical equipment or a certain component in the mechanical equipment; the operating state signal can be a key index signal that can characterize the operating condition of the mechanical equipment. For example, vibration signals, temperature values, pressure values, and noise signals, etc. Specifically, it can be determined based on actual needs, and the embodiment of the present disclosure does not limit this.

[0042] It can be understood that in the process of monitoring the operating state of mechanical equipment, sensors are usually used to collect the operating state signals of the mechanical equipment and send them to the terminal device, where the operating state signals are time series signals.

[0043] In an alternative embodiment, the process by which the terminal device obtains the operating state signals of the mechanical equipment includes: the terminal device receives the operating state signals of the mechanical equipment collected by the sensor.

[0044] In step S102, the terminal device preprocesses the operating state signals to obtain the operating characteristic data of the mechanical equipment, and frequency-encodes the operating characteristic data to obtain the to-be-processed spike sequence.

[0045] In the embodiments of the present disclosure, the operating characteristic data is data extracted from the original operating state signals and can more accurately characterize the operating state characteristics of the mechanical equipment. It can not only reduce the amount of data processing in the fault diagnosis process, but also further improve the accuracy of the determined fault diagnosis results.

[0046] It should be noted that in the embodiments of the present disclosure, the process by which the terminal device preprocesses the operating state signals may include: determining at least one of the time-domain characteristic data, frequency-domain characteristic data, and decomposition characteristic data of the operating state signals. Specifically, it can be determined based on actual needs, and the embodiments of the present disclosure do not limit this. Among them, the decomposition characteristic data may be local mean decomposition (LMD) data, empirical mode decomposition (EMD) data, variational mode decomposition (VMD) data, etc.

[0047] In an alternative embodiment, the process by which the terminal device preprocesses the operating state signals to obtain the operating characteristic data of the mechanical equipment includes: extracting the time-domain characteristic data of the mechanical equipment by performing time-domain characteristic extraction on the operating state signals; then, performing Fourier transform on the time-domain characteristic data to obtain the frequency-domain characteristic data of the mechanical equipment; and performing signal decomposition processing on the time-domain characteristic data to obtain the decomposition characteristic data of the mechanical equipment; further, determining the time-domain characteristic data, the frequency-domain characteristic data, and the decomposition characteristic data as the operating characteristic data of the mechanical equipment. The time-domain characteristic data, frequency-domain characteristic data, and decomposition characteristic data extracted from the operating state signals can be determined as the operating characteristic data of the mechanical equipment, so as to characterize the operating state characteristics of the mechanical equipment through multi-dimensional characteristic data and improve the richness and accuracy of the characterization of the operating state characteristics of the mechanical equipment.

[0048] Among them, the type of time-domain feature data extracted by the terminal device can be determined based on actual needs, and the embodiments of the present disclosure do not limit this. For example, as shown in Table 1, the terminal device can extract 11 time-domain feature data such as the average value, standard deviation, and root mean square from the operation state signal of the mechanical equipment; in Table 1, x(n) is a time series where n = 1, 2, …, N, and N is the number of sampling points.

[0049] Table 1 Time-Domain Feature Data

[0050]

[0051] Table 2 Frequency-Domain Feature Data

[0052]

[0053] Similarly, the type of frequency-domain feature data extracted by the terminal device can be determined based on actual needs, and the embodiments of the present disclosure do not limit this. For example, as shown in Table 2, the terminal device can extract 13 frequency-domain feature data such as the average frequency, frequency variance, and frequency skewness from the Fourier transform result of the time-domain feature data of the mechanical equipment; in Table 2, s(k) is the spectrum where k = 1, 2, …, K, and K is the number of spectral lines; f k is the frequency value of the Xth spectral line.

[0054] In an alternative embodiment, when the decomposition strategy used by the terminal device in the signal decomposition of the time-domain feature data is VMD, the decomposition feature data obtained by the terminal device includes variational mode decomposition (VMD) data of multiple modes. Among them, the terminal device can also: determine the energy value of the VMD data of each mode to obtain energy data associated with the VMD data of each mode; then the process by which the terminal device determines the time-domain feature data, the frequency-domain feature data, and the decomposition feature data as the operation feature data of the mechanical equipment includes: determining the time-domain feature data, the frequency-domain feature data, the decomposition feature data, and the energy data associated with the VMD data of each mode as the operation feature data of the mechanical equipment. When determining the VMD data as the operation feature data of the mechanical equipment, the energy value of the VMD data of each mode can be determined, and the energy value of the VMD data of each mode can also be determined as the operation feature data of the mechanical equipment, so as to more accurately characterize the operation state characteristics of the mechanical equipment through the operation feature data including the energy value of the VMD data.

[0055] It can be understood that when the terminal device decomposes the time-domain feature data using the VMD algorithm, and the obtained decomposed feature data includes variational mode decomposition (VMD) data of multiple modes. Among them, the process of the terminal device decomposing the time-domain feature data based on the VMD algorithm can be implemented based on the first formula, and the first formula is:

[0056]

[0057] In Formula 1, K is the k-th mode, u k is the component of the k-th mode, w k is the central frequency of the k-th mode, is the partial derivative with respect to time t, ‖·‖ 2 is the norm for measuring the energy of the signal, is to find the analytic signal of the source signal, represents shifting the spectrum to the baseband, and j is the imaginary unit.

[0058] Optionally, when the terminal device decomposes the time-domain feature data using the VMD algorithm, the terminal device can also determine the energy value of the VMD data of each mode, and obtain energy data associated with the VMD data of each mode. Among them, the process of the terminal device determining the energy value of the VMD data of each mode can be implemented based on the second formula, and the second formula is:

[0059]

[0060] In Formula 2, IEF k is the energy value of the VMD data of the k-th mode.

[0061] It should be noted that in the embodiments of the present disclosure, after obtaining the operation feature data of the mechanical equipment, the terminal device can encode the operation feature data based on different frequency encoding methods to obtain a to-be-processed spike sequence. Specifically, it can be determined based on actual needs, and the embodiments of the present disclosure do not limit this. For example, Poisson pulse coding, random pulse sequence coding, Gaussian pulse coding, etc.

[0062] In step S103, the terminal device inputs the to-be-processed spike sequence into the liquid state machine (LSM) model to process the to-be-processed spike sequence based on the spike-timing-dependent plasticity (STDP) algorithm, update the weights between neurons in the LSM model, and output a to-be-recognized spike sequence.

[0063] In the embodiments of the present disclosure, the to-be-recognized spike sequence contains the fault feature information of the mechanical equipment recognized by the LSM.

[0064] It should be noted that in the embodiments of the present disclosure, the spiking neural network is similar to the communication method of biological neurons. The spiking neural network transmits precise time information through spikes. Generally, when the membrane potential of a neuron reaches a specific threshold, the neuron emits a spike and the potential is reset, as Figure 2 shown Figure 2 Figure Figure 2 shows a schematic diagram of the change in the membrane potential of neuron A in a spiking neural network under the influence of neurons i and j connected to it. Among them, V m (t) represents the change in membrane potential over time, V thre is the threshold potential for neuron firing, and V rev is the resting membrane potential. When the potential is lower than the resting membrane potential, the neuron remains inactive.

[0065] The LSM model is a computational model that uses a spiking neuron library to process time-varying inputs. The structure of the LSM model is as Figure 3 shown, including an input layer 301, a liquid layer 302, and an output layer 303. Among them, the liquid layer is used to convert the input data provided by the input layer into liquid output data. This process can be implemented based on Formula 3, and Formula 3 is:

[0066] x M (t) = L M (S in (t)); (Formula 3)

[0067] In Formula 3, S in (t) is the input data, x M (t) is the liquid output data, and L M is the liquid conversion function.

[0068] The output layer is used to convert the liquid output data into model output data. This process can be implemented based on Formula 4, and Formula 4 is:

[0069] y(t) = f M (x M (t)); (Formula 4)

[0070] In Formula 4, y(t) is the model output data, and f M is the output mapping function.

[0071] In an alternative embodiment, the process by which the terminal device processes the to-be-processed spike train based on the spike-timing-dependent plasticity (STDP) algorithm, updates the weights between neurons in the LSM model, and outputs the to-be-identified spike train includes: processing the to-be-processed spike train based on the STDP algorithm to update the weights between neurons in the input layer and the liquid layer of the LSM model, and / or the weights between neurons in the liquid layer, and determining and outputting the to-be-identified spike train based on the updated weights between neurons. The weights between neurons in the input layer and the liquid layer of the LSM model, and / or the weights between neurons in the liquid layer can be updated based on the STDP algorithm, so as to select the spiking neural network layer in the LSM model for weight update according to the actual computing power or the need for high recognition accuracy, and meet the special requirements of the actual situation while ensuring the recognition accuracy of fault features.

[0072] In an alternative embodiment, the process by which the terminal device processes the to-be-processed spike train based on the STDP algorithm, updates the weights between neurons in the input layer and the liquid layer of the LSM model, and / or the weights between neurons in the liquid layer, and determines the to-be-identified spike train based on the updated weights between neurons includes: in the input layer of the LSM model, processing the to-be-processed spike train based on the STDP algorithm to update the weights between neurons in the input layer and the liquid layer, and, based on the updated weights between neurons in the input layer and the liquid layer, determining an initial spike train; then, inputting the initial spike train into the liquid layer of the LSM model and processing the initial spike train based on the STDP algorithm to update the weights between neurons in the liquid layer, and, based on the updated weights between neurons in the liquid layer, determining the to-be-identified spike train. The weights between neurons in the input layer and the liquid layer of the LSM model, and, the weights between neurons in the liquid layer can be updated based on the STDP algorithm to improve the recognition accuracy of fault features of mechanical equipment based on the LSM model by adjusting the weights of neurons between the input layer and the liquid layer of the LSM model, and within the liquid layer itself.

[0073] Among them, the STDP algorithm can be implemented by the third formula, where the third formula is:

[0074]

[0075] In formula 5, t post is the spike emission time of the presynaptic neuron, t pre is the spike emission time of the postsynaptic neuron, Δt is the spike emission time difference, and the weight update function W is:

[0076]

[0077] In Equation 6, A + is the learning rate of the potential, A- is the learning rate of inhibition, and τ + and τ - are time constants. Here, Δt > 0 indicates that the presynaptic spike precedes the postsynaptic spike, and the synapse is strengthened by increasing the synaptic weight. On the contrary, Δt < 0 indicates that the postsynaptic spike occurs before the presynaptic spike, and the synapse is weakened by reducing the synaptic weight, as Figure 4 shown Figure 4 shows a schematic diagram of synaptic weight adjustment according to the present disclosure. W(Δt) represents the functional relationship between the weight change and the time difference, increasing the weight when it is positive and decreasing the weight when it is negative.

[0078] In an alternative embodiment, in the embodiments of the present disclosure, during the process of determining the spike sequence to be recognized based on the LSM model, the terminal device can also process the to-be-processed spike sequence based on the STDP algorithm, update the weights between the input layer and the liquid layer of the LSM model or between the neurons in the liquid layer, and determine the spike sequence to be recognized based on the updated weights between the neurons; wherein, the process of determining the spike sequence to be recognized is similar to the process in the above embodiments, where the terminal device processes the to-be-processed spike sequence based on the STDP algorithm, updates the weights of the neurons between the input layer and the liquid layer of the LSM model, and the weights between the neurons in the liquid layer, and determines the spike sequence to be recognized based on the updated weights between the neurons. The embodiments of the present disclosure will not elaborate on this.

[0079] It should be noted that, in the embodiments of the present disclosure, in order to further improve the recognition ability of neurons for time-dependent features, the neurons in the LSM model are integrate-and-fire (LIF) neurons based on conductance synapses, where, as Figure 5 shown Figure 5 shows a circuit schematic diagram of a LIF neuron model. Among them, the membrane voltage of the LIF neuron is:

[0080]

[0081] In Equation 7, C m is the membrane capacitance, R m is the membrane resistance, I(t) is the input current of the neuron at time t. Among them, after the to-be-processed spike sequence is input into the LSM model, it can trigger a successive connection; where, when the membrane voltage V m (t) reaches the threshold voltage V there , the neuron emits a spike represented by the Dirac delta function δ(t - t i ), and at the same time, the membrane voltage is reset.

[0082] Optionally, the neurons between the input layer and the liquid layer are LIF neurons based on conductance synapses. The terminal device determines an initial spike sequence based on the updated weights of the neurons between the input layer and the liquid layer, including: inputting the updated weights of the LIF neurons between the input layer and the liquid layer into a conductance dynamics model to determine the updated conductance values of the LIF neurons in the input layer; and inputting the updated conductance values into a neuron membrane voltage control model based on conductance to obtain the updated membrane voltages of the LIF neurons between the input layer and the liquid layer. Then, based on a spike sequence generation strategy and the updated membrane voltages of the LIF neurons between the input layer and the liquid layer, the initial spike sequence is determined. The synaptic connections between the output layer and the liquid layer in the LSM model can be adjusted based on the LIF neurons with conductance synapses to ensure that the output layer can transmit an initial spike sequence containing richer and more accurate time-dependent features to the liquid layer for fault feature recognition, further improving the accuracy of the fault feature recognition result determined based on the LSM model.

[0083] Among them, the conductance dynamics model is:

[0084]

[0085] In Equation 8, τ g is the synaptic time constant, g syn is the conductance, w 1 is the synaptic weight, t spike represents the time of the presynaptic spike; this conductance dynamics model simulates the relationship between the instantaneous increase in conductance when a spike arrives and its subsequent exponential decay. Among them, the synaptic time constant can be determined based on actual needs, and the embodiments of the present disclosure do not limit this.

[0086] The neuron membrane voltage control model based on conductance is:

[0087]

[0088] In Equation 9, τ m is the membrane time constant, V rev is the reversal membrane potential. Among them, the membrane time constant can be determined based on actual needs, and the embodiments of the present disclosure do not limit this.

[0089] It can be understood that in the embodiments of the present disclosure, the spike sequence generation strategy is: when the membrane potential of a neuron reaches a specific threshold, the neuron will emit a spike.

[0090] In an alternative embodiment, the neurons in the liquid layer are conductance-based synaptic LIF neurons. The process of determining the spike sequence to be recognized based on the updated weights between the neurons in the liquid layer includes: inputting the updated weights between the neurons in the liquid layer into a conductance dynamics model to determine the updated conductance values of the LIF neurons in the liquid layer; and inputting the updated conductance values into a neuron membrane voltage control model based on conductance to obtain the updated membrane voltages of the LIF neurons in the liquid layer. Then, based on a spike sequence generation strategy and the updated membrane voltages of the LIF neurons in the liquid layer, the spike sequence to be recognized is determined. The LIF neurons based on conductance synapses can adjust the synaptic connections in the liquid layer of the LSM model to ensure that the liquid layer can transmit the spike sequence to be recognized, which contains richer and more accurate time-dependent features, to the convolutional neural network model for fault feature recognition, further improving the accuracy of the fault feature recognition result determined based on the LSM model.

[0091] In step S104, the terminal device inputs the spike sequence to be recognized into a convolutional neural network model for feature recognition to obtain the fault diagnosis result of the mechanical equipment.

[0092] In the embodiments of the present disclosure, the convolutional neural network is used to perform fault classification according to the spike count results of each neuron in the liquid layer of the LSM model within a preset time window; where the spike count result within the preset time window can be expressed as:

[0093]

[0094] In formula 10, δ is the Dirac delta function, t i is the spike time of the neuron, and the start time of the preset time window is t 0 .

[0095] In an alternative embodiment, the process by which the terminal device inputs the spike sequence to be recognized into a convolutional neural network model for feature recognition to obtain the fault diagnosis result of the mechanical equipment is implemented based on a decision function, where the decision function is:

[0096] f(s) = w 2 T φ(s) + b; (Formula 11)

[0097] where f(s) is the fault diagnosis result, w 2 is the weight vector of the convolutional neural network, b is the bias term, φ is the kernel function, and s is the set of spike count results of each neuron in the liquid layer of the LSM model determined based on the spike sequence to be recognized within a preset time window, where s = (s 1 , s2 ,..., s N ), where N is the number of neurons in the liquid layer of the LSM model.

[0098] It should be noted that in the embodiments of the present disclosure, the type of the convolutional neural network model can be determined according to actual needs, and the embodiments of the present disclosure do not limit this. For example, the convolutional neural network model can be a Support Vector Machine (SVM) model, and among them, the convolutional neural network model can be pre-trained.

[0099] Among them, the SVM model finds the optimal hyperplane that maximizes the margin between different types of faults by solving an optimization problem to determine the fault diagnosis result. Among them, the process of the SVM model finding the optimal hyperplane that maximizes the margin between different classes by solving an optimization problem is as follows:

[0100]

[0101] In Equation 12, ξ k is the Slack variable, and C is the regularization parameter.

[0102] Among them, the fault diagnosis result output by the SVM model is determined by the sign of the output result of the decision function y(s) of the SVM model. For example, if y(s) is a positive value, it means there is a certain type of fault, and if y(s) is a negative value, it means there is no certain type of fault. Among them, the type of fault is determined by the number of optimal hyperplanes, that is, the SVM model can find several hyperplanes, and then it can diagnose several types of faults. The decision function y(s) of the SVM model is:

[0103] y(s) = sign(f(s)) = sign(w 2 T φ(s) + b); (Equation 13)

[0104] For example, the terminal device evaluates the experimental results obtained by the fault diagnosis solution provided in the embodiments of the present disclosure on the Case Western Reserve University Bearing Fault Diagnosis Dataset (CWRU). This dataset includes fault data at three positions on the inner race, outer race, and rolling elements. Among them, the terminal device uses the target dataset collected under a motor load of 1 horsepower (HP) and a motor speed of approximately 1772 rpm.

[0105] Among them, the target dataset includes the healthy y data of three fault locations and the fault data of the inner ring, ball, and outer ring (6 o'clock direction) of bearings with diameters of 0.007 inches, 0.014 inches, and 0.021 inches, respectively, as well as the corresponding 10 fault types: normal, 3 ball faults (B007, B014, B021), 3 inner ring faults (IR007, IR014, IR021), and 3 outer ring faults (OR007@6, OR014@6, OR021@6); moreover, the basic configuration of the terminal device can be determined based on actual needs, and the embodiments of the present disclosure do not limit this.

[0106] Exemplarily, as Figure 6 shown, Figure 6 shows a schematic diagram of the fault diagnosis process for this experiment, where the terminal device uses the SVM model for fault diagnosis; as Figure 7 shown, Figure 7 shows a schematic diagram of the confusion matrix of the target dataset used in the experiment provided by the embodiments of the present disclosure.

[0107] It can be understood that during this experiment, the terminal device can be based on the Figure 7 data in, use the fault diagnosis scheme provided in the embodiments of the present disclosure, and determine the fault diagnosis result of the bearing according to the fault diagnosis process shown in Figure 6 ; among them, as Figure 8 shown, Figure 8 shows a schematic diagram of the operating characteristic data of the bearing during this experiment being mapped to a two-dimensional space, and, as Figure 9 shown, Figure 9 shows a schematic diagram of the spike sequence to be identified determined by the LSM model during this experiment being mapped to a two-dimensional space.

[0108] Optionally, the terminal device can also evaluate the reliability of the fault diagnosis scheme provided in the embodiments of the present disclosure based on the Receiver Operating Characteristic curve (ROC) and the Area Under the Curve (AUC); among them, as Figure 10 shown, Figure 10 shows a schematic diagram of the ROC curve and AUC results determined during this experiment, where only two misclassifications occurred in the target dataset: B014 was misclassified as B021, and OR014@6 was misclassified as B014. Despite these misclassifications, the ROC curve and the Area Under the Curve (AUC) value are close to 1, confirming that the fault diagnosis scheme provided in the embodiments of the present disclosure has strong fault identification capabilities and always maintains high resolution capabilities among fault types.

[0109] An exemplary embodiment of the present disclosure provides a fault diagnosis device, which can be a server or a chip applied to a server. Figure 11 FIG. shows a schematic block diagram of functional modules of a fault diagnosis device according to an exemplary embodiment of the present disclosure. As Figure 11 shown, the fault diagnosis device 1100 includes:

[0110] An acquisition module 1101, configured to acquire an operation state signal of a mechanical device;

[0111] A preprocessing module 1102, configured to preprocess the operation state signal to obtain operation characteristic data of the mechanical device, and frequency-encode the operation characteristic data to obtain a to-be-processed spike sequence;

[0112] An identification module 1103, configured to input the to-be-processed spike sequence into a liquid state machine LSM model, process the to-be-processed spike sequence based on a spike timing-dependent plasticity STDP algorithm, update weights between neurons in the LSM model, and output a to-be-identified spike sequence;

[0113] A diagnosis module 1104, configured to input the to-be-identified spike sequence into a convolutional neural network model for feature recognition to obtain a fault diagnosis result of the mechanical device.

[0114] Optionally, the identification module 1103 is configured to:

[0115] Process the to-be-processed spike sequence based on the STDP algorithm, update weights between neurons between the input layer and the liquid layer in the LSM model, and / or weights between neurons in the liquid layer, and determine and output a to-be-identified spike sequence based on the updated weights between neurons.

[0116] Optionally, the identification module 1103 is configured to:

[0117] In the input layer of the LSM model, process the to-be-processed spike sequence based on the STDP algorithm to update weights between neurons between the input layer and the liquid layer, and, based on the updated weights between neurons between the input layer and the liquid layer, determine an initial spike sequence;

[0118] Input the initial spike sequence into the liquid layer of the LSM model, and process the initial spike sequence based on the STDP algorithm to update weights between neurons in the liquid layer, and, based on the updated weights between neurons in the liquid layer, determine the to-be-identified spike sequence.

[0119] Optionally, the identification module 1103 is configured to:

[0120] Input the updated weight of the LIF neuron between the input layer and the liquid layer into the conductance dynamics model to determine the updated conductance value of the LIF neuron in the input layer;

[0121] Input the updated conductance value into the conductance-based neuron membrane voltage control model to obtain the updated membrane voltage of the LIF neuron between the input layer and the liquid layer;

[0122] Determine the initial spike sequence based on the spike sequence generation strategy and the updated membrane voltage of the LIF neuron between the input layer and the liquid layer.

[0123] Optionally, the recognition module 1103 is configured to:

[0124] Input the updated weight between the neurons in the liquid layer into the conductance dynamics model to determine the updated conductance value of the LIF neuron in the liquid layer;

[0125] Input the updated conductance value into the conductance-based neuron membrane voltage control model to obtain the updated membrane voltage of the LIF neuron in the liquid layer;

[0126] Determine the spike sequence to be recognized based on the spike sequence generation strategy and the updated membrane voltage of the LIF neuron in the liquid layer.

[0127] Optionally, the preprocessing module 1102 is configured to:

[0128] Extract the time-domain features of the operation state signal to obtain the time-domain feature data of the mechanical equipment;

[0129] Perform Fourier transform on the time-domain feature data to obtain the frequency-domain feature data of the mechanical equipment;

[0130] Perform signal decomposition processing on the time-domain feature data to obtain the decomposed feature data of the mechanical equipment;

[0131] Determine the time-domain feature data, the frequency-domain feature data, and the decomposed feature data as the operation feature data of the mechanical equipment.

[0132] Optionally, the device further includes a determination module 1105, which is configured to:

[0133] Determine the energy value of the VMD data of each mode to obtain the energy data associated with the VMD data of each mode;

[0134] The step of determining the time-domain feature data, the frequency-domain feature data, and the decomposed feature data as the operation feature data of the mechanical equipment includes:

[0135] The time-domain feature data, the frequency-domain feature data, the decomposed feature data, and the energy data associated with the VMD data of each mode are determined as the operation feature data of the mechanical equipment.

[0136] An exemplary embodiment of the present disclosure also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, and when executed by the at least one processor, the computer program is used to cause the electronic device to execute the method according to the embodiment of the present disclosure.

[0137] An exemplary embodiment of the present disclosure also provides a non-transitory computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor of a computer, the computer is caused to execute the method according to the embodiment of the present disclosure.

[0138] As Figure 12 shown, an exemplary embodiment of the present disclosure also provides a computer program product 1200, including a computer program 1201, wherein when the computer program is executed by a processor of a computer, the computer is caused to execute the method according to the embodiment of the present disclosure.

[0139] Referring Figure 13 to, a block diagram of an electronic device 1300 that can be used as a terminal device of the present disclosure will now be described, which is an example of a hardware device applicable to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0140] As Figure 13 shown, the electronic device 1300 includes a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage unit 1308 into a random access memory (RAM) 1303. In the RAM 1303, various programs and data required for the operation of the electronic device 1300 can also be stored. The computing unit 1301, the ROM 1302, and the RAM 1303 are connected to each other through a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0141] Multiple components in the electronic device 1300 are connected to the I / O interface 1305, including: an input unit 1306, an output unit 1307, a storage unit 1308, and a communication unit 1309. The input unit 1306 can be any type of device capable of inputting information into the electronic device 1300. The input unit 1306 can receive input digital or character information and generate key signal inputs related to the user settings and / or function controls of the electronic device. The output unit 1307 can be any type of device capable of presenting information and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 1308 can include, but is not limited to, magnetic disks and optical disks. The communication unit 1309 allows the electronic device 1300 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0142] The computing unit 1301 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1301 executes the various methods and processes described above. For example, in some embodiments, the methods of the exemplary embodiments of the present disclosure can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 1308. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 1300 via the ROM 1302 and / or the communication unit 1309. In some embodiments, the computing unit 1301 can be configured to execute the methods of the exemplary embodiments of the present disclosure by any other suitable means (e.g., by means of firmware).

[0143] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, executed as an independent software package partially on the machine and partially on a remote machine, or executed entirely on a remote machine or server.

[0144] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0145] As used in this disclosure, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (e.g., a disk, an optical disk, a memory, a programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0146] In order to provide an interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide an interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0147] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), and the Internet.

[0148] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present disclosure are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a terminal, a user device, or other programmable devices. The computer program or instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrating one or more available media. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it can also be an optical medium, such as a digital video disc (DVD); or it can be a semiconductor medium, such as a solid state drive (SSD).

[0149] Although the present disclosure has been described in connection with specific features and their embodiments, it is obvious that various modifications and combinations can be made without departing from the spirit and scope of the present disclosure. Accordingly, this specification and the drawings are merely exemplary illustrations of the present disclosure defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present disclosure. Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalent technologies, the present disclosure is also intended to include these changes and modifications.

Claims

1. A fault diagnosis method, characterized in that: include: Obtain the operating status signal of mechanical equipment; Preprocessing the operation status signal to obtain operation characteristic data of the mechanical equipment, and frequency encoding the operation characteristic data to obtain a spike sequence to be processed; Inputting the spike sequence to be processed into the liquid state machine LSM model, processing the spike sequence to be processed based on the pulse timing dependent plasticity STDP algorithm, updating the weights between neurons in the LSM model, and outputting the spike sequence to be identified; The spike sequence to be identified is input into a convolutional neural network model for feature recognition to obtain a fault diagnosis result of the mechanical equipment.

2. The fault diagnosis method according to claim 1, characterized in that: The step of processing the spike sequence to be processed based on the pulse timing dependent plasticity STDP algorithm, updating the weights between neurons in the LSM model, and outputting the spike sequence to be identified includes: The spike sequence to be processed is processed based on the STDP algorithm, the weights of the neurons between the input layer and the liquid layer of the LSM model and / or the weights between the neurons in the liquid layer are updated, and based on the updated weights between the neurons, the spike sequence to be identified is determined and output.

3. The fault diagnosis method according to claim 2, characterized in that: Processing the spike sequence to be processed based on the STDP algorithm, updating the weights of neurons between the input layer and the liquid layer of the LSM model, and / or the weights between neurons in the liquid layer, and determining the spike sequence to be identified based on the updated weights between neurons, including: In the input layer of the LSM model, the to-be-processed spike sequence is processed based on the STDP algorithm to update the weights of neurons between the input layer and the liquid layer, and an initial spike sequence is determined based on the updated weights of neurons between the input layer and the liquid layer; The initial spike sequence is input into the liquid layer of the LSM model, and the initial spike sequence is processed based on the STDP algorithm to update the weights between the neurons in the liquid layer, and the spike sequence to be identified is determined based on the updated weights between the neurons in the liquid layer.

4. The fault diagnosis method according to claim 3, characterized in that: The neurons between the input layer and the liquid layer are integrated-release (LIF) neurons based on conductance synapses, and the initial spike sequence is determined based on the updated weights of the neurons between the input layer and the liquid layer, including: Inputting the updated weights of the LIF neurons between the input layer and the liquid layer into the conductance dynamics model to determine the updated conductance values ​​of the LIF neurons in the input layer; Inputting the updated conductance value into a conductance-based neuron membrane voltage control model to obtain an updated membrane voltage of the LIF neuron between the input layer and the liquid layer; The initial spike train is determined based on a spike train generation strategy and an updated membrane voltage of a LIF neuron between the input layer and the fluid layer.

5. The fault diagnosis method according to claim 3, characterized in that: The neurons in the liquid layer are LIF neurons based on conductance synapses, and the spike sequence to be identified is determined based on the updated weights between the neurons in the liquid layer, including: Inputting the updated weights between neurons in the liquid layer into the conductance dynamics model to determine the updated conductance values ​​of the LIF neurons in the liquid layer; Inputting the updated conductance value into a conductance-based neuron membrane voltage control model to obtain an updated membrane voltage of the LIF neuron in the liquid layer; The spike sequence to be identified is determined based on the spike sequence generation strategy and the updated membrane voltage of the LIF neurons in the liquid layer.

6. The fault diagnosis method according to any one of claims 1 to 5, characterized in that: The preprocessing of the operation status signal to obtain the operation characteristic data of the mechanical equipment includes: Extracting time domain features of the operating status signal to obtain time domain feature data of the mechanical equipment; Performing Fourier transformation on the time domain characteristic data to obtain frequency domain characteristic data of the mechanical equipment; Performing signal decomposition processing on the time domain characteristic data to obtain decomposed characteristic data of the mechanical equipment; The time domain characteristic data, the frequency domain characteristic data and the decomposed characteristic data are determined as the operation characteristic data of the mechanical equipment.

7. The fault diagnosis method according to claim 6, characterized in that: The decomposition feature data includes variational mode decomposition (VMD) data of multiple modes, and the method further includes: Determine the energy value of each mode of VMD data to obtain energy data associated with each mode of VMD data; The step of determining the time domain feature data, the frequency domain feature data, and the decomposed feature data as the operation feature data of the mechanical equipment includes: The time domain characteristic data, the frequency domain characteristic data, the decomposed characteristic data, and the energy data associated with the VMD data of each mode are determined as the operation characteristic data of the mechanical equipment.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the fault diagnosis method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fault diagnosis method according to any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the fault diagnosis method according to any one of claims 1 to 7 is implemented.

Images