Method and apparatus for determining a faulty component and a component remaining useful life

Abstract

The embodiment of the application provides a kind of method and device for determining fault component and component remaining service life.It includes: receiving the first real-time working signal and the second real-time working signal of multiple target components at preset time;Variational modal decomposition is carried out on the first real-time working signal to obtain multiple signal components;Determine the first feature data of each signal component;The second feature data is determined by pre-processing the second real-time working signal;According to all the first feature data and the second feature data, a feature matrix is constructed;The feature matrix is input into the trained network model to output the component that fails in multiple target components and the remaining service life of each target component.The above technical solution can monitor the system and components in all directions in real time, accurately determine the remaining service life of each target, and timely find the faulty components in the system, improve the management of the life of the components and improve the safety of the components.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection, and in particular to a method and device for determining a faulty component and a remaining service life of the component, a storage medium and a processor. BACKGROUND

[0002] Hydraulic systems are widely used in industrial machinery production, and have complex structures and strong air tightness. Therefore, once a fault occurs, it is difficult for an operator to observe the fault signs and to determine the fault cause and the fault damage degree. The state monitoring of a hydraulic system is an important means to ensure the safe operation of mechanical equipment. However, the hydraulic system fault presents complex physical properties such as nonlinearity, discontinuity and multi-coupling. The existing technology only analyzes and diagnoses the individual signals of key components of the system, ignoring the fact that the hydraulic signals under actual working conditions have strong fault multi-coupling and fault fuzziness. As a result, the fault determination and life prediction accuracy are not high, and the practicability is low. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a method and device for determining a faulty component and a remaining service life of the component, a storage medium and a processor.

[0004] To achieve the above-mentioned purpose, the first aspect of the present application provides a method for determining a faulty component and a remaining service life of the component, comprising:

[0005] In the case where a system in which a plurality of target components are located has a fault, first real-time working signals and second real-time working signals of the plurality of target components at a preset time are received;

[0006] The first real-time working signals are subjected to variational mode decomposition to obtain a plurality of signal components corresponding to the first real-time working signals;

[0007] First feature data of each signal component are determined;

[0008] The second real-time working signals are preprocessed, and second feature data of the preprocessed second real-time working signals are determined;

[0009] A feature matrix is constructed according to all the first feature data and the second feature data;

[0010] The feature matrix is input into a trained network model, so as to output, by the trained network model, a component having a fault in the plurality of target components and a remaining service life of each target component.

[0011] In the embodiment of the present application, the first feature data includes component energy, and determining the first feature data of each signal component includes: obtaining energy values of a plurality of signal points included in each signal component; for any one signal component, determining a sum of squares of the energy values of all signal points of the signal component as the component energy of the signal component.

[0012] In the embodiment of the present application, the first feature data further includes first energy entropy, and determining the first feature data of each signal component includes calculating the first energy entropy according to formula (1):

[0013]

[0014] wherein H En represents the first energy entropy of the nth signal component, n represents the number of signal components, E i represents the component energy of the ith component, E represents a sum of component energies of all signal components, and ln represents a logarithmic function.

[0015] In the embodiment of the present application, the second feature data includes signal energy, and determining the second feature data of the preprocessed second real-time working signal includes: obtaining energy values of a plurality of signal points included in the preprocessed second real-time working signal; and determining a sum of squares of the energy values of all signal points included in the preprocessed second real-time working signal as the signal energy.

[0016] In the embodiment of the present application, the second feature data further includes second energy entropy, and determining the second feature data of the preprocessed second real-time working signal includes calculating the second energy entropy according to formula (2):

[0017]

[0018] wherein H Ex represents the second energy entropy, x represents the number of all signal points included in the second real-time working signal, E j represents the energy value of the jth signal point, E represents the signal energy, and ln represents a logarithmic function.

[0019] In the embodiment of the present application, the first real-time working signal is subjected to variational mode decomposition to obtain a plurality of signal components corresponding to the first real-time working signal, including: obtaining a decomposition number of the variational mode decomposition for the first real-time working signal; processing the first real-time working signal to determine a center frequency of the first real-time working signal; determining a target bandwidth parameter of the first real-time working signal according to the decomposition number and the center frequency; and decomposing the first real-time working signal according to the target bandwidth parameter to obtain the plurality of signal components.

[0020] In the embodiment of the present application, the method further comprises: obtaining a fault type corresponding to a historical fault of each target component occurring at a preset historical time and a historical remaining service life of each target component at the preset historical time; generating a fault code for each fault type and a life code of the historical remaining service life of each target component; obtaining a first historical working signal and a second historical working signal of each target component at the preset historical time; performing variational mode decomposition on the first historical working signal to obtain a plurality of historical signal components corresponding to the first historical working signal; determining first historical feature data of each historical signal component; preprocessing the second historical working signal and determining second historical feature data of the preprocessed second historical working signal; constructing a historical feature matrix according to all the first historical feature data and the second historical feature data; generating a feature label of the historical feature matrix at the preset historical time according to all the fault codes and all the life codes; and inputting the historical feature matrix carrying the feature label into the network model to train the network model.

[0021] The second aspect of the present application provides a processor configured to execute the above-mentioned method for determining a faulty component and a component remaining service life.

[0022] The third aspect of the present application provides an apparatus for determining a faulty component and a component remaining service life, comprising the above-mentioned processor.

[0023] The fourth aspect of the present application provides a machine-readable storage medium having instructions stored thereon, which, when executed by a processor, cause the processor to be configured to execute the above-mentioned method for determining a faulty component and a component remaining service life.

[0024] The above technical solution, by receiving a first real-time working signal and a second real-time working signal of a plurality of target components at a preset time in the case of a failure of a system in which the plurality of target components are located; performing variational mode decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal; determining first feature data of each signal component; preprocessing the second real-time working signal and determining second feature data of the preprocessed second real-time working signal; constructing a feature matrix according to all the first feature data and the second feature data; and inputting the feature matrix into a trained network model to output a faulty component in the plurality of target components and a remaining service life of each target component through the trained network model. The above technical solution can comprehensively and real-timely monitor the system and the components, accurately determine the remaining service life of each target, and timely find the faulty component in the system, thereby improving the management of the life of the components and enhancing the use safety of the components.

[0025] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0026] The accompanying drawings are included to provide a further understanding of embodiments of the application, and are incorporated in and constitute a part of this specification, illustrate embodiments of the application, and together with the description serve to explain the principles of the application. In the drawings:

[0027] Figure 1 schematically shows a flow diagram of a method for determining a failed component and a remaining useful life of a component according to an embodiment of the application;

[0028] Figure 2 schematically shows another flow diagram of a method for determining a failed component and a remaining useful life of a component according to an embodiment of the application;

[0029] Figure 3 schematically shows an internal structure diagram of a computer device according to an embodiment of the application. DETAILED DESCRIPTION

[0030] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It should be understood that the specific embodiments described herein are merely used to explain and illustrate the embodiments of the present application, and should not be used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0031] Figure 1 schematically shows a flow diagram of a method for determining a failed component and a remaining useful life of a component according to an embodiment of the application. As shown in Figure 1 In an embodiment of the present application, a method for determining a failed component and a remaining useful life of a component is provided, comprising the following steps:

[0032] Step 101, in a case where a system in which a plurality of target components are located fails, receiving a first real-time working signal and a second real-time working signal of the plurality of target components at a preset time.

[0033] Step 102, performing variational mode decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal.

[0034] Step 103, determining first feature data of each signal component.

[0035] Step 104, preprocessing the second real-time working signal, and determining second feature data of the preprocessed second real-time working signal.

[0036] Step 105: construct a feature matrix based on all the first feature data and the second feature data.

[0037] In step 106 , the feature matrix is ​​input into the trained network model, so as to output the failed components among the multiple target components and the remaining service life of each target component through the trained network model.

[0038] Variational mode decomposition (VMD) refers to a method that iteratively searches for the optimal solution of a variational model to determine the frequency center and bandwidth of each component during the decomposition process, thereby adaptively achieving frequency domain decomposition of the signal and effective separation of the components. In the event of a failure in a system containing multiple target components, a processor may receive first and second real-time operating signals from the multiple target components at predetermined times. The first real-time operating signals may include at least pressure signals, vibration signals, and sound signals. The second real-time operating signals may include at least temperature signals, flow signals, and displacement signals. After receiving the first and second real-time operating signals, the processor may perform variational mode decomposition on the first real-time operating signal to obtain multiple signal components corresponding to the first real-time operating signal. After obtaining the multiple signal components, the processor may determine first feature data for each signal component. The processor may preprocess the second real-time operating signal and determine second feature data for the preprocessed second real-time operating signal. After determining the first and second feature data, the processor may construct a feature matrix based on all of the first and second feature data. After constructing the feature matrix, the processor may input the feature matrix into the trained network model to output the failed components among the plurality of target components and the remaining service life of each target component through the trained network model.

[0039] For example, in the case of a hydraulic system in which a hydraulic multi-way valve, a hydraulic pump and a hydraulic cylinder fail, the processor can receive pressure signals, vibration signals and sound signals of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder at the current time, and can also receive temperature signals, flow signals and displacement signals. The processor can perform variational modal decomposition on the pressure signals, the vibration signals and the sound signals to obtain eight signal components corresponding to the pressure signals, the vibration signals and the sound signals respectively. Specifically, for the pressure signals, the vibration signals and the sound signals, each signal is first decomposed into a plurality of eigenmode components with limited bandwidth, and then the sum of the bandwidths of all components is minimized. The minimized bandwidth is determined as the corresponding optimal bandwidth balance parameter, and finally eight signal components corresponding to the force signals, the vibration signals and the sound signals are determined according to the optimal bandwidth balance parameter. The processor can determine the first feature data of each signal component, such as energy, energy entropy, transient amplitude root mean square, transient frequency root mean square, dimensioned, dimensionless, etc. The processor can perform smoothing processing on the temperature signals, the flow signals and the displacement signals to remove abnormal signals in the temperature signals, the flow signals and the displacement signals. After smoothing processing on the temperature signals, the flow signals and the displacement signals, the processor can determine the second feature data of the smoothed temperature signals, the flow signals and the displacement signals. The processor can construct a feature matrix according to all the first feature data and the second feature data. And input the feature matrix into the trained network model to output the remaining service life of the stuck hydraulic multi-way valve and the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder through the trained network model.

[0040] The above technical solutions can monitor the system and components in all directions in real time, accurately determine the remaining service life of each target, and find the faulty components in the system in time, thereby improving the management of the life of the components and improving the safety of the use of the components.

[0041] In one embodiment, the first feature data includes component energy, and determining the first feature data of each signal component includes: obtaining energy values of a plurality of signal points included in each signal component; for any one signal component, the sum of the squares of the energy values of all signal points of the signal component is determined as the component energy of the signal component.

[0042] The first feature data includes component energies. The processor can determine the first feature data of each component energy. Specifically, the processor can obtain energy values of a plurality of signal points included in each signal component. After obtaining the energy values of each signal point, for any one signal component, the processor can determine a sum of squares of the energy values of all signal points of the signal component as the component energy of the signal component, so that the first feature data includes energy characteristics of each signal point, and the running state of each component can be comprehensively considered, and all components are comprehensively detected.

[0043] For example, the processor can obtain energy values of n signal points included in the signal component x. The processor can determine a sum of squares of the energy values of all signal points of the signal component x as the component energy of the signal component x, that is, the component energy E x = E1+E2+…+En, where En represents the square of the energy value of the nth signal point.

[0044] In one embodiment, the first feature data further includes a first energy entropy, and determining the first feature data of each signal component includes calculating the first energy entropy according to formula (1):

[0045]

[0046] where H En represents the first energy entropy of the nth signal component, n represents the number of signal components, E i represents the component energy of the ith component, E represents the sum of component energies of all signal components, and ln represents a logarithmic function.

[0047] In one embodiment, the second feature data includes signal energy, and determining the second feature data of the preprocessed second real-time working signal includes: obtaining energy values of a plurality of signal points included in the preprocessed second real-time working signal; and determining a sum of squares of the energy values of all signal points included in the preprocessed second real-time working signal as the signal energy.

[0048] The second feature data includes signal energy. The processor can determine the second feature data of the preprocessed second real-time working signal. Specifically, the processor can obtain energy values of a plurality of signal points included in the preprocessed second real-time working signal. And a sum of squares of the energy values of all signal points included in the preprocessed second real-time working signal is determined as the signal energy.

[0049] In one embodiment, the second feature data further includes a second energy entropy, and determining the second feature data of the preprocessed second real-time working signal includes calculating the second energy entropy according to formula (2):

[0050]

[0051] wherein, H Ex represents the second energy entropy, x represents the number of all signal points included in the second real-time working signal, E j represents the energy value of the jth signal point, E represents the signal energy, and ln represents the logarithmic function.

[0052] In one embodiment, the first real-time working signal is subjected to variational mode decomposition to obtain a plurality of signal components corresponding to the first real-time working signal, including: obtaining a decomposition number of the variational mode decomposition for the first real-time working signal; processing the first real-time working signal to determine a center frequency of the first real-time working signal; determining a target bandwidth parameter of the first real-time working signal according to the decomposition number and the center frequency; and decomposing the first real-time working signal according to the target bandwidth parameter to obtain the plurality of signal components.

[0053] The processor can perform variational mode decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal. Specifically, the processor can obtain a decomposition number of the variational mode decomposition for the first real-time working signal. The processor can also process the first real-time working signal to determine a center frequency of the first real-time working signal. After determining the decomposition number and the center frequency, the processor can determine a target bandwidth parameter of the first real-time working signal according to the decomposition number and the center frequency. After determining the target bandwidth parameter, the processor can decompose the first real-time working signal according to the target bandwidth parameter to obtain the plurality of signal components. By decomposing the first real-time working signal according to the target bandwidth parameter, the bandwidth of each signal component after decomposition can be the optimal bandwidth, thereby enabling the extracted feature data to accurately reflect the true condition of the component and improving the accuracy of the remaining useful life prediction of the component.

[0054] In one embodiment, the method further includes: obtaining a historical failure type corresponding to a historical failure of each target component occurring at a preset historical time and a historical remaining useful life of each target component at the preset historical time; generating a failure code for each failure type and a life code of the historical remaining useful life of each target component; obtaining a first historical working signal and a second historical working signal of each target component at the preset historical time; performing variational mode decomposition on the first historical working signal to obtain a plurality of historical signal components corresponding to the first historical working signal; determining first historical feature data of each historical signal component; preprocessing the second historical working signal and determining second historical feature data of the preprocessed second historical working signal; constructing a historical feature matrix according to all the first historical feature data and the second historical feature data; generating a feature label of the historical feature matrix at the preset historical time according to all the failure codes and all the life codes; and inputting the historical feature matrix carrying the feature label into the network model to train the network model.

[0055] The processor can obtain a failure type corresponding to a failure of each target component at a preset historical time and a historical remaining useful life of each target component at the preset historical time. After determining the failure type and the historical remaining useful life, the processor can generate a failure code for each failure type and a life code of the historical remaining useful life of each target component. The failure code and the life code can both use one-hot encoding. The processor can also obtain a first historical working signal and a second historical working signal of each target component at the preset historical time. The processor can perform variational mode decomposition on the first historical working signal to obtain a plurality of historical signal components corresponding to the first historical working signal. The processor can determine first historical feature data of each historical signal component. The processor can preprocess the second historical working signal and determine second historical feature data of the preprocessed second historical working signal. After determining the first historical feature data and the second historical feature data, the processor can construct a historical feature matrix according to all the first historical feature data and the second historical feature data. After constructing the historical feature matrix, the processor can generate a feature label of the historical feature matrix at the preset historical time according to all the failure codes and all the life codes. After generating the feature label, the processor can input the historical feature matrix carrying the feature label into a network model to train the network model, so that the diagnosis ability of the network model for the failed component is continuously improved, and the prediction accuracy of the network model for the remaining useful life is improved.

[0056] For example, the target components include a hydraulic multi-way valve, a hydraulic pump, and a hydraulic cylinder. The failure types corresponding to the historical failures of the hydraulic multi-way valve include valve core jamming, valve core and valve body wear, multi-way valve core main spring deformation or fracture, and seal damage. The processor can generate failure codes a1, b1, c1, and d1 for valve core jamming, valve core and valve body wear, multi-way valve core main spring deformation or fracture, and seal damage, respectively. The historical failures of the hydraulic pump include slipper wear failure, piston loose boot, and swash plate wear failure. The processor can generate failure codes a2, b2, and c2 for slipper wear failure, piston loose boot, and swash plate wear failure, respectively. The historical failures of the hydraulic cylinder include noise hydraulic cylinder pull, seal ring wear, and hydraulic rod wear. The processor can generate failure codes a3, b3, and c3 for noise hydraulic cylinder pull, seal ring wear, and hydraulic rod wear, respectively.

[0057] At a preset historical time t1, the hydraulic multi-way valve fails. The processor can obtain a failure type corresponding to a historical failure of the hydraulic multi-way valve at the preset historical time t1 and historical remaining useful lives Rul 11 , Rul 12 , and Rul 13 of the hydraulic multi-way valve, the hydraulic pump, and the hydraulic cylinder at the preset historical time t1. The processor can generate failure codes a1, b1, c1, and d1 for the historical remaining useful lives Rul 11life tag Rul ′ 11 , historical remaining useful life Rul 12 life tag Rul ′ 12 , historical remaining useful life Rul 13 life tag Rul ′ 13 .

[0058] At the preset historical moment t2, the hydraulic multi-way valve and the hydraulic pump fail. The processor can acquire the fault type corresponding to the historical failure of the hydraulic pump at the preset historical moment t2 and the historical remaining useful life Rul 21 , Rul 22 and Rul 23 of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder at the preset historical moment t2. The processor can generate the life tag Rul 21 for the historical remaining useful life Rul ′ 21 , historical remaining useful life Rul 22 life tag Rul ′ 22 , historical remaining useful life Rul 23 life tag Rul ′ 23 .

[0059] At the preset historical moment t2, the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder all fail. The processor can acquire the fault type corresponding to the historical failure of the hydraulic cylinder at the preset historical moment t3 and the historical remaining useful life Rul 31 , Rul 32 and Rul 33 of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder at the preset historical moment t3. The processor can generate the life tag Rul 31 for the historical remaining useful life Rul ′ 31 , historical remaining useful life Rul 32 life tag Rul ′ 32 , historical remaining useful life Rul 33 life tag Rul ′ 33 .

[0060] The processor can obtain first historical working signals and second historical working signals of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder at a preset historical time t1. The processor can perform variational mode decomposition on the first historical working signals to obtain a plurality of historical signal components corresponding to the first historical working signals. The processor can determine first historical feature data corresponding to each historical signal component. The processor can perform smoothing processing on the second historical working signals, and determine second historical feature data of the second historical working signals after the smoothing processing. The historical feature data at least includes energy, energy entropy, transient amplitude root mean square, transient frequency root mean square, dimensioned (mean value, peak value, root mean square, standard deviation, rectified average, variance), dimensionless (kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor) and the like. The processor can construct a historical feature matrix according to all the first historical feature data and the second historical feature data.

[0061] At the preset historical time t1, the processor can determine the feature label N1 of the historical feature matrix of the preset historical time t1 according to the fault labels a1, b1, c1 and d1 and the life label Rul ′ 11 , Rul ′ 12 , Rul ′ 13 At the preset historical time t1, the processor can determine the feature label N1 of the historical feature matrix of the preset historical time t1 according to the fault labels a1, b1, c1 and d1 and the life label Rul

[0062] At the preset historical time t2, the processor can determine the feature label N2 of the historical feature matrix of the preset historical time t2 according to the fault labels a1, b1, c1, d1, a2, b2, c2 and the life label Rul ′ 21 , Rul ′ 22 , Rul ′ 23 At the preset historical time t2, the processor can determine the feature label N2 of the historical feature matrix of the preset historical time t2 according to the fault labels a1, b1, c1, d1, a2, b2, c2 and the life label Rul

[0063] At the preset historical time t3, the processor can determine the feature label N3 of the historical feature matrix of the preset historical time t3 according to the fault labels a1, b1, c1, d1, a2, b2, c2, a3, b3, c3 and the life label Rul ′ 31 , Rul ′ 32 , Rul ′ 33 At the preset historical time t3, the processor can determine the feature label N3 of the historical feature matrix of the preset historical time t3 according to the fault labels a1, b1, c1, d1, a2, b2, c2, a3, b3, c3 and the life label Rul

[0064] In one embodiment, asFigure 2 As shown, in the case of a failure of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder, the processor can receive the pressure signal, the vibration signal and the sound signal of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder at the current time, and can also receive the temperature signal, the flow signal and the displacement signal. After receiving the pressure signal, the vibration signal, the sound signal, the temperature signal, the flow signal and the displacement signal, the processor can perform signal conditioning on the received signals. And collect data in the signal under the LabVIEW (program development environment) development environment.

[0065] Specifically, the processor can perform fault information mining and feature extraction from the collected data. For example, the processor can perform variational modal decomposition on the pressure signal, the vibration signal and the sound signal to obtain 8 signal components corresponding to the pressure signal, the vibration signal and the sound signal respectively. For the pressure signal, the vibration signal and the sound signal, each signal is first decomposed into a plurality of finite bandwidth eigenmode components, and then the sum of the bandwidths of all components is minimized. And the minimized bandwidth is determined as the corresponding optimal bandwidth balance parameter, and finally 8 signal components corresponding to the force signal, the vibration signal and the sound signal are determined according to the optimal bandwidth balance parameter. The processor can determine the first feature data of each signal component, such as energy, energy entropy, transient amplitude root mean square, transient frequency root mean square, dimensioned, dimensionless, etc. The processor can perform smoothing processing on the temperature signal, the flow signal and the displacement signal to remove abnormal signals in the temperature signal, the flow signal and the displacement signal. After smoothing the temperature signal, the flow signal and the displacement signal, the processor can determine the second feature data of the smoothed temperature signal, the flow signal and the displacement signal.

[0066] The processor can construct a feature matrix according to all the first feature data and the second feature data. After constructing the feature matrix, the processor can input the feature matrix to the hydraulic system intelligent monitoring based on deep neural network to obtain the fault components of the hydraulic system and the remaining service life of the hydraulic multi-way valve, the hydraulic pump and the hydraulic cylinder. The processor can visualize the fault data of the fault components, and can also alarm the component failure to enable predictive maintenance of the system.

[0067] The processor can also perform deep model iterative updating. Among them, the model can select E-DBN-LSTM (network model). As shown in Table 1, the E-DBN-LSTM model, in experiment 2, the training time Time of the model is 630, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 10.73, and the average relative error MAPE is 0.268. In experiment 3, the training time Time of the model is 694, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 12.69, and the average relative error MAPE is 0.377. In experiment 4, the training time Time of the model is 638, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 11.12, and the average relative error MAPE is 0.321. In experiment 5, the training time Time of the model is 674, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 10.82, and the average relative error MAPE is 0.352. The average of the training time Time of the model is 659, the average of the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 11.34, and the average of the average relative error MAPE is 0.329.

[0068] The DBN (Deep Belief Network) in experiment 2, the training time Time of the model is 2013, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 27.44, and the average relative error MAPE is 0.484. In experiment 3, the training time Time of the model is 1987, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 28.95, and the average relative error MAPE is 0.521. In experiment 4, the training time Time of the model is 1990, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 25.29, and the average relative error MAPE is 0.493. In experiment 5, the training time Time of the model is 1984, the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 31.45, and the average relative error MAPE is 0.594. The average of the training time Time of the model is 1993, the average of the average absolute error MAE between the predicted remaining useful life and the actual remaining useful life is 28.28, and the average of the average relative error MAPE is 0.523.

[0069] The LSTM (Long Short Term Memory neural network) has a training time Time of 2034, an average absolute error MAE of 31.12 between the predicted residual service life and the actual residual service life, and an average relative error MAPE of 0.596 in experiment 2. The model has a training time Time of 2019, an average absolute error MAE of 30.64 between the predicted residual service life and the actual residual service life, and an average relative error MAPE of 0.683 in experiment 3. The model has a training time Time of 2001, an average absolute error MAE of 34.69 between the predicted residual service life and the actual residual service life, and an average relative error MAPE of 0.629 in experiment 4. The model has a training time Time of 1997, an average absolute error MAE of 32.36 between the predicted residual service life and the actual residual service life, and an average relative error MAPE of 0.583 in experiment 5. The average of the training time Time of the model is 2012, the average of the average absolute error MAE between the predicted residual service life and the actual residual service life is 32.21, and the average of the average relative error MAPE is 0.623.

[0070]

[0071] Table I Experimental data of E-DBN-LSTM, DBN and LSTM

[0072] The accuracy of the E-DBN-LSTM is greater than that of the DBN and the LSTM. The E-DBN-LSTM can be used to more accurately predict the residual service life of each component, thereby more effectively managing the components.

[0073] The above technical solution receives the first real-time working signal and the second real-time working signal of the plurality of target components at the preset time in the case of a system failure of the plurality of target components; performs variational mode decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal; determines the first feature data of each signal component; pre-processes the second real-time working signal and determines the second feature data of the pre-processed second real-time working signal; constructs a feature matrix according to all the first feature data and the second feature data; inputs the feature matrix into the trained network model to output the component that fails in the plurality of target components and the residual service life of each target component through the trained network model. The above technical solution can comprehensively and real-time monitor the system and the components, accurately determine the residual service life of each target, and timely find the component that fails in the system, thereby improving the management of the life of the components and improving the use safety of the components.

[0074] Figure 1 ,2 FIG. 1 is a flow chart of a method for determining a faulty component and a component's remaining useful life in one embodiment. Figure 1 、 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 、 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0075] An embodiment of the present application provides a processor, which is used to run a program, wherein the program executes the above-mentioned method for determining a faulty component and a remaining service life of the component when running.

[0076] An embodiment of the present application provides a device for determining a faulty component and the remaining useful life of the component, including the above-mentioned processor.

[0077] An embodiment of the present application provides a storage medium having a program stored thereon, which, when executed by a processor, implements the above-mentioned method for determining a faulty component and the remaining useful life of the component.

[0078] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor A01, a network interface A02, a memory (not shown in the figure) and a database (not shown in the figure) connected via a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02 and a database (not shown in the figure). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A04. The database of the computer device is used to store a first real-time working signal, a second real-time working signal, a first characteristic data and a second characteristic data. The network interface A02 of the computer device is used to communicate with an external terminal via a network connection. When the computer program B02 is executed by the processor A01, a method for determining a faulty component and the remaining service life of a component is implemented.

[0079] Those skilled in the art can understand that Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0080] The embodiment of the present application provides a device, the device comprising a processor, a memory and a program stored on the memory and executable on the processor, when the processor executes the program, the following steps are implemented: in the case that a system where a plurality of target components are located fails, a first real-time working signal and a second real-time working signal of the plurality of target components at a preset time are received; the first real-time working signal is subjected to variational mode decomposition to obtain a plurality of signal components corresponding to the first real-time working signal; first feature data of each signal component is determined; the second real-time working signal is preprocessed, and second feature data of the preprocessed second real-time working signal is determined; a feature matrix is constructed according to all the first feature data and the second feature data; and the feature matrix is input into a network model trained to output a component that fails among the plurality of target components and a remaining service life of each target component through the network model trained.

[0081] In one embodiment, the first feature data comprises component energy, and the determination of the first feature data of each signal component comprises: obtaining energy values of a plurality of signal points included in each signal component; and for any one signal component, a sum of squares of the energy values of all the signal points of the signal component is determined as component energy of the signal component.

[0082] In one embodiment, the first feature data further comprises first energy entropy, and the determination of the first feature data of each signal component comprises calculating the first energy entropy according to formula (1):

[0083]

[0084] Wherein, H En represents the first energy entropy of the nth signal component, n represents the number of signal components, E i represents component energy of the ith component, E represents a sum of component energies of all the signal components, and ln represents a logarithmic function.

[0085] In one embodiment, the second feature data comprises signal energy, and the determination of the second feature data of the preprocessed second real-time working signal comprises: obtaining energy values of a plurality of signal points included in the preprocessed second real-time working signal; and determining a sum of squares of the energy values of all the signal points included in the preprocessed second real-time working signal as signal energy.

[0086] In one embodiment, the second feature data further comprises a second energy entropy, and determining the second feature data of the preprocessed second real-time working signal comprises calculating the second energy entropy according to formula (2):

[0087]

[0088] wherein H Ex represents the second energy entropy, x represents the number of all signal points included in the second real-time working signal, E j represents the energy value of the jth signal point, E represents the signal energy, and ln represents a logarithmic function.

[0089] In one embodiment, the variational mode decomposition is performed on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal, comprising: obtaining a decomposition number of the variational mode decomposition for the first real-time working signal; processing the first real-time working signal to determine a center frequency of the first real-time working signal; determining a target bandwidth parameter of the first real-time working signal according to the decomposition number and the center frequency; and decomposing the first real-time working signal according to the target bandwidth parameter to obtain the plurality of signal components.

[0090] In one embodiment, the method further comprises: obtaining a fault type corresponding to a historical fault of each target component occurring at a preset historical time and a historical remaining useful life of each target component at the preset historical time; generating a fault code for each fault type and a life code of the historical remaining useful life of each target component; obtaining a first historical working signal and a second historical working signal of each target component at the preset historical time; performing variational mode decomposition on the first historical working signal to obtain a plurality of historical signal components corresponding to the first historical working signal; determining first historical feature data of each historical signal component; pre-processing the second historical working signal and determining second historical feature data of the preprocessed second historical working signal; constructing a historical feature matrix according to all the first historical feature data and the second historical feature data; generating a feature label of the historical feature matrix at the preset historical time according to all the fault codes and all the life codes; and inputting the historical feature matrix carrying the feature label into the network model to train the network model.

[0091] The present application also provides a computer program product adapted to perform the steps of the method for determining a faulty component and a remaining useful life of the component when executed on a data processing device.

[0092] Those skilled in the art will appreciate that embodiments of the application can be readily used as software, hardware, or a combination of software and hardware. In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0093] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0094] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0095] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0096] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0097] The memory can include non-persistent memory and / or persistent memory, such as flash memory, read-only memory (ROM), and / or volatile or non-volatile random access memory (RAM), among others. The memory is an example of computer-readable media.

[0098] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0099] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0100] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for determining a faulty component and the remaining useful life of the component, characterized in that The method comprises: receiving, in the event of a failure in a system where the plurality of target components are located, first real-time operating signals and second real-time operating signals of the plurality of target components at a preset time, wherein the first real-time operating signal includes at least a pressure signal, a vibration signal, and a sound signal, and the second real-time operating signal includes at least a temperature signal, a flow signal, and a displacement signal; performing variational mode decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal; determining first characteristic data for each signal component; Preprocessing the second real-time working signal and determining second characteristic data of the preprocessed second real-time working signal, wherein the preprocessing refers to smoothing processing; Constructing a feature matrix according to all the first feature data and the second feature data; Inputting the feature matrix into a trained network model to output the failed components among the plurality of target components and the remaining service life of each target component through the trained network model; The performing variational modal decomposition on the first real-time working signal to obtain a plurality of signal components corresponding to the first real-time working signal includes: Obtaining the number of decompositions of the variational mode decomposition for the first real-time working signal; processing the first real-time operating signal to determine a center frequency of the first real-time operating signal; determining a target bandwidth parameter of the first real-time working signal according to the decomposition number and the center frequency; Decomposing the first real-time working signal according to the target bandwidth parameter to obtain a plurality of signal components; The method further comprises: Obtaining the fault type corresponding to the historical fault that occurred at a preset historical moment and the historical remaining service life of each target component at the preset historical moment; Generate a fault code for each fault type and a life code for the historical remaining service life of each target component; Acquire a first historical operating signal and a second historical operating signal of each target component at the preset historical moment; Performing variational mode decomposition on the first historical working signal to obtain a plurality of historical signal components corresponding to the first historical working signal; Determining first historical feature data of each historical signal component; preprocessing the second historical working signal, and determining second historical characteristic data of the preprocessed second historical working signal; Constructing a historical feature matrix based on all the first historical feature data and the second historical feature data; Generating a feature label of the historical feature matrix at a preset historical moment according to all fault codes and all life codes; The historical feature matrix carrying the feature labels is input into the network model to train the network model.

2. The method for determining a faulty component and a component's remaining useful life according to claim 1, wherein: The first characteristic data includes component energy, and determining the first characteristic data of each signal component includes: Obtaining energy values ​​of a plurality of signal points included in each signal component; For any signal component, the sum of squares of energy values ​​of all signal points of the signal component is determined as the component energy of the signal component.

3. The method for determining a faulty component and a component's remaining useful life according to claim 2, wherein: The first characteristic data further includes a first energy entropy, and determining the first characteristic data of each signal component includes calculating the first energy entropy according to formula (1): （1） in, represents the first energy entropy of the nth signal component, n represents the number of signal components, represents the component energy of the i-th component, E represents the component energy sum of all signal components, Represents a logarithmic function.

4. The method for determining a faulty component and a component's remaining useful life according to claim 1, wherein: The second characteristic data includes signal energy, and the second characteristic data of the second real-time working signal after preprocessing is determined to include: Acquiring the preprocessed second real-time working signal including energy values ​​of multiple signal points; The signal energy is determined by taking the square sum of energy values ​​of all signal points included in the preprocessed second real-time working signal.

5. The method for determining a faulty component and a component's remaining useful life according to claim 4, wherein: The second characteristic data further includes a second energy entropy, and determining the second characteristic data of the preprocessed second real-time working signal includes calculating the second energy entropy according to formula (2): （2） in, represents the second energy entropy, x represents the number of all signal points included in the second real-time working signal, represents the energy value of the j-th signal point, E represents the signal energy, and ln represents a logarithmic function.

6. A processor, characterized in that: The device is configured to perform the method for determining a faulty component and a remaining useful life of a component according to any one of claims 1 to 5.

7. A device for determining a faulty component and the remaining useful life of the component, characterized in that: The apparatus comprises a processor according to claim 6.

8. A machine-readable storage medium having instructions stored thereon, characterized in that: When executed by a processor, the instructions cause the processor to be configured to perform the method for determining a faulty component and a remaining useful life of a component according to any one of claims 1 to 5.

Images