CN118173121A

CN118173121A - Equipment running state evaluation method, device, computer equipment and storage medium

Info

Publication number: CN118173121A
Application number: CN202410360813.5A
Authority: CN
Inventors: 焦石; 谷裕; 张健成; 李舒维; 邓健俊; 赖桂森; 杨学广; 张朝辉; 龙建华
Original assignee: Guangzhou Bureau of Extra High Voltage Power Transmission Co
Current assignee: Guangzhou Bureau of Extra High Voltage Power Transmission Co
Priority date: 2024-03-27
Filing date: 2024-03-27
Publication date: 2024-06-11

Abstract

The application relates to a device running state evaluation method, a device, a computer device, a storage medium and a computer program product. The method comprises the following steps: and (3) acquiring the voiceprint signal generated by the equipment to be evaluated in the operation process, and carrying out convolution operation on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal, wherein the transformation operator is adopted to participate in the convolution operation, so that the data processing speed is improved. Under the condition that the convolution voiceprint signal meets noise reduction screening conditions, feature extraction is carried out on the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal.

Description

Equipment running state evaluation method, device, computer equipment and storage medium

Technical Field

The present application relates to the field of device evaluation technologies, and in particular, to a device operation state evaluation method, device, computer device, storage medium, and computer program product.

Background

Equipment under various working conditions is often in heavy load or full load operation, and in order to prolong the service life of the equipment and improve the stability and reliability of the system, the equipment needs to be overhauled.

In the existing mode, workers regularly disassemble and detect equipment, and the mode ensures the detection accuracy of the equipment to a certain extent, but simultaneously occupies more manpower resources, and irreparable service life loss exists for the equipment after the equipment is disassembled and reinstalled each time.

Disclosure of Invention

In view of the foregoing, it is desirable to provide an apparatus, a device, a computer-readable storage medium, and a computer program product that can reduce the lifetime loss of the device.

In a first aspect, the present application provides a method for evaluating an operating state of a device. The method comprises the following steps:

Acquiring a voiceprint signal generated by equipment to be evaluated in the running process; performing convolution operation on the voiceprint signal according to a transformation operator generated in time-frequency transformation of the voiceprint signal to obtain a convolution voiceprint signal; under the condition that the convolution voiceprint signal meets noise reduction screening conditions, extracting features of the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal; and evaluating the running state of the equipment based on the voiceprint characteristics to obtain a running state evaluation result of the equipment.

In one embodiment, the transformation operator includes a two-dimensional operator and a one-dimensional operator; the convolution operation is carried out on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal, and the method comprises the following steps: performing frequency domain conversion on the voiceprint signal to obtain a forward discrete transformation result of the voiceprint signal; performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal; the backward discrete transformation result comprises a two-dimensional operator; according to the disassembly dimension of the backward discrete transformation result, carrying out disassembly operation on the two-dimensional operator to obtain a one-dimensional operator of the backward discrete transformation result; and carrying out convolution operation on the voiceprint signal according to the one-dimensional operator to obtain a convolution voiceprint signal.

In one embodiment, the performing frequency domain conversion on the voiceprint signal to obtain a forward discrete transform result of the voiceprint signal includes: carrying out data point sampling on the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals; and generating a forward discrete transformation result of the voiceprint signal according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array.

In one embodiment, the feature extracting the convolved voiceprint signal to obtain a voiceprint feature of the convolved voiceprint signal includes: performing multiple feature extraction on the convolution voiceprint signal to obtain basic voiceprint features representing different feature extraction degrees in the convolution voiceprint signal, wherein the basic voiceprint features comprise basic channel numbers; weighting the basic voiceprint features corresponding to each basic channel number according to the channel weight matched with the basic channel number to obtain the feature weight of each basic voiceprint feature; and weighting the basic voiceprint features corresponding to each feature weight according to the feature weights to obtain the voiceprint features of the convolution voiceprint signals.

In one embodiment, the estimating the operation state of the device based on the voiceprint feature to obtain an operation state estimation result of the device includes: evaluating the running state of the equipment based on the voiceprint features, and determining the respective prediction probabilities of the equipment corresponding to a plurality of candidate faults; and determining an operation state evaluation result of the equipment according to each prediction probability.

In one embodiment, the estimating the operation state of the device based on the voiceprint feature to obtain an operation state estimation result of the device includes: acquiring an initial evaluation model for the evaluated equipment; based on the equipment association relation between the equipment to be evaluated and the evaluated equipment, performing transfer learning on the initial evaluation equipment to obtain an updated evaluation model matched with the equipment to be evaluated; and inputting the voiceprint features into the updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

In a second aspect, the application further provides a device for evaluating the running state of equipment. The device comprises:

The acquisition module is used for acquiring voiceprint signals generated in the running process of the equipment to be evaluated; the convolution module is used for carrying out convolution operation on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal; the extraction module is used for extracting the characteristics of the convolution voiceprint signal under the condition that the convolution voiceprint signal meets the noise reduction screening condition to obtain the voiceprint characteristics of the convolution voiceprint signal; and the evaluation module is used for evaluating the running state of the equipment based on the voiceprint characteristics to obtain a running state evaluation result of the equipment.

In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor which when executing the computer program performs the steps of:

In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of:

According to the equipment running state evaluation method, the equipment running state evaluation device, the computer equipment, the storage medium and the computer program product, the voiceprint signals generated by the equipment to be evaluated in the running process are obtained, convolution operation is carried out on the voiceprint signals according to the transformation operators generated by the voiceprint signals in time-frequency transformation, the convolution voiceprint signals are obtained, and the transformation operators are adopted to participate in convolution operation, so that the data processing speed is improved. Under the condition that the convolution voiceprint signal meets noise reduction screening conditions, feature extraction is carried out on the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal.

Drawings

FIG. 1 is an application environment diagram of a device operational state assessment method in one embodiment;

FIG. 2 is a flow diagram of a method for evaluating device operating state in one embodiment;

FIG. 3 is a flow diagram of a method for device operational state assessment based on attention mechanisms and parallel migration acceleration in one embodiment;

FIG. 4 is a block diagram of a device operation state evaluation apparatus in one embodiment;

fig. 5 is an internal structural diagram of a computer device in one embodiment.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

The method for evaluating the running state of the equipment, provided by the embodiment of the application, can be applied to an application environment shown in figure 1. Wherein the collection end 102 communicates with the server 104 through a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server.

The server 104 obtains a voiceprint signal generated by the device to be evaluated in the operation process from the acquisition end 102, the server 104 carries out convolution operation on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal, the server 104 carries out feature extraction on the convolution voiceprint signal under the condition that the convolution voiceprint signal meets noise reduction screening conditions to obtain voiceprint features of the convolution voiceprint signal, and the server 104 carries out evaluation on the operation state of the device based on the voiceprint features to obtain an operation state evaluation result of the device.

The acquisition end 102 may be various devices with voiceprint acquisition capabilities, and the devices carry various types of sensors, such as capacitive sensors, piezoelectric sensors, and MEMS sensors. The collection end 102 may collect the sound produced by the device, obtain a voiceprint signal, and send the voiceprint signal to the server 104. By way of example, the collection terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices, and portable wearable devices, which may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, and the like. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.

In one embodiment, as shown in fig. 2, there is provided a device operation state evaluation method, which is described by taking an example that the method is applied to the server in fig. 1, and includes:

s202, obtaining a voiceprint signal generated by equipment to be evaluated in the running process.

The device may be a device of a power station, for example a converter valve, a line reactor, a converter valve cooling device or other electrical device. The device can generate various sounds during operation, the sounds can be analyzed, and the operation state type of the device can be predicted in sequence.

The voiceprint signal, also called voiceprint, is a sound wave spectrum carrying speech information, and can be displayed by an electroacoustical instrument. Voiceprints have characteristics of specificity and relative stability.

Specifically, by acquiring the voiceprint signal of the device, the voiceprint signal can be analyzed to obtain an analysis result of whether the device is operating normally, and if the analysis result indicates that the device is not operating normally, the abnormal type of the device can be further judged.

By way of example, the cooling device of the converter valve of the power converter station is described by taking the device as an example, the cooling device of the converter valve comprises various parts such as bearings and gears, when the cooling device is abnormal, for example, the bearings are damaged and the gear box is abnormal, abnormal noise is usually generated, the voiceprint signal of the cooling device can be obtained, the voiceprint signal is analyzed to obtain an analysis result of whether the cooling device is abnormal, if the analysis result of the cooling device is abnormal, the occurrence of the abnormality of a part in the cooling device is represented, the voiceprint signal is required to be further analyzed, and the specific abnormal type in the cooling device is obtained, wherein the abnormal type can be the type such as bearing damage, bearing wear, looseness of a pipeline connection part and the like. Further, after obtaining the abnormal type of the cooling device, the related maintenance personnel can be informed to carry out targeted maintenance.

S204, performing convolution operation on the voiceprint signal according to a transformation operator generated in time-frequency transformation of the voiceprint signal to obtain a convolution voiceprint signal.

The transformation operator may be a laplace operator generated by the voiceprint signal in time-frequency transformation.

There are various forms of transformation operators, such as two-dimensional laplace operator, one-dimensional laplace operator, gradient operator. Specifically, taking into account the differential nature of the fourier transform, the difference between the difference and the difference is calculated in the fourier domain by applying a difference to the voiceprint signalTo obtain the gradient operator/>, of the voiceprint signal。

Wherein, the voiceprint signalThe fourier transform formula in the continuous domain is as follows:

Wherein, Representation/>And/>And/>Is a continuous frequency.

After the gradient operator is obtained, the method can also be used for controlling the voice print signalThe fourier transform formula in the continuous domain performs frequency domain transformation to obtain fourier transform in the discrete domain, i.e., forward discrete fourier transform.

In particular, the voiceprint signal needs to be first aligned in the discrete domainSampling is performed by having/>Finite square pair voiceprint signal of data points/>Sampling to obtain sampling result, and according to the sampling result, performing/>, on the voiceprint signalAnd performing frequency domain transformation in a Fourier transform formula of a continuous domain to obtain forward discrete Fourier transform.

After the forward discrete fourier transform is obtained, the forward discrete fourier transform can be subjected to data transposition to obtain the backward discrete fourier transform containing the two-dimensional laplace operator.

And further, calculating according to the two-dimensional Laplace operator, and performing convolution operation on the voiceprint signal according to the two-dimensional Laplace operator to obtain a convolution voiceprint signal.

It should be noted that, the two-dimensional laplace operator can be disassembled into a plurality of one-dimensional laplace operators, so that the calculation dimension is reduced, meanwhile, the dimension corresponding to each one-dimensional laplace operator is independent, no data dependency exists between independent one-dimensional calculations, and communication transmission of one calculation can be performed while the other calculation is performed, thereby realizing natural overlapping of calculation and communication and improving the calculation speed.

S206, under the condition that the convolution voiceprint signal meets the noise reduction screening condition, extracting features of the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal.

The noise reduction screening condition may be a discrimination condition for processing a single convolution voiceprint signal, or may be a discrimination condition for processing a plurality of convolution voiceprint signals.

Specifically, for a single convolution voiceprint signal, if the single convolution voiceprint signal meets the noise reduction screening condition, it represents that at least part of the voiceprint signals in the convolution voiceprint signal meet the noise reduction screening condition, and then part of the voiceprint signals which do not meet the noise reduction screening condition can be removed, at least part of the voiceprint signals are reserved, and then feature extraction is performed according to at least part of the voiceprint signals.

Specifically, for a plurality of convolution voiceprint signals, if the convolution voiceprint signals meet the noise reduction screening condition, signals representing the convolution voiceprint signals do not need noise reduction, and feature extraction can be directly performed on the convolution voiceprint signals.

For a plurality of convolution voiceprint signals, noise reduction screening can be performed on each convolution voiceprint signal to obtain at least one convolution voiceprint signal, and noise reduction screening is performed on the convolution voiceprint signal in the at least one convolution voiceprint signal to obtain at least part of the convolution voiceprint signals.

The feature extraction may be performed on a time sequence sensitive to the data sequence in the convolution voiceprint signal, specifically, feature extraction may be performed on the convolution voiceprint signal to obtain a plurality of basic features, and then re-extraction is performed on the convolution voiceprint information according to weights matched by the basic features to obtain voiceprint features reflecting importance of the features.

S208, the operation state of the equipment is evaluated based on the voiceprint characteristics, and an operation state evaluation result of the equipment is obtained.

The operating state of the device may be a state that the device presents when operating. The evaluation result of the operation state of the equipment comprises the following steps: normal evaluation results and abnormal evaluation results.

Wherein the exception evaluation result comprises a plurality of exception types. Illustratively, for a converter valve cooling apparatus, the anomaly types include: bearing damage, bearing wear, loosening of pipeline joints, etc.

It is understood that the evaluation result of the running state may be a probability value or a probability distribution.

Specifically, if the evaluation result of the operation state is a probability value, the higher the probability value is, the more the operation state of the representative device approaches to normal, the lower the probability value is, the more the operation state of the representative device approaches to abnormal, further, whether the device is normal or not can be determined according to the size of the probability value, and if the device is abnormal, the abnormal type of the device can be determined.

Specifically, if the evaluation result of the operation state is a probability distribution, the probability distribution represents a probability distribution condition of each operation state type of the apparatus, and the probability distribution includes, by way of example: the probability of bearing damage is 0.2, the probability of bearing abrasion is 0.6, the probability of looseness at the pipeline connection part is 0.1, and the probability of equipment normal is 0.1.

In the equipment operation state evaluation method, the voiceprint signals generated by the equipment to be evaluated in the operation process are obtained, convolution operation is carried out on the voiceprint signals according to the transformation operators generated by the voiceprint signals in the time-frequency transformation to obtain the convolution voiceprint signals, and the transformation operators are adopted to participate in the convolution operation, so that the data processing speed is improved. Under the condition that the convolution voiceprint signal meets noise reduction screening conditions, feature extraction is carried out on the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal.

In one embodiment, the transformation operator includes a two-dimensional operator and a one-dimensional operator, S204, including: performing frequency domain conversion on the voiceprint signal to obtain a forward discrete transformation result of the voiceprint signal, and performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal; the backward discrete transformation result comprises a two-dimensional operator, the two-dimensional operator is disassembled according to the disassembly dimension of the backward discrete transformation result to obtain a one-dimensional operator of the backward discrete transformation result, and the voiceprint signal is convolved according to the one-dimensional operator to obtain a convolved voiceprint signal.

The transformation operator can be a Laplace operator, wherein the Laplace operator comprises a two-dimensional Laplace operator and a one-dimensional Laplace operator, and the Laplace operator can be a gradient operator.

The frequency domain conversion is performed on the voiceprint signal, so as to obtain the voiceprint signal in the discrete domain, that is, the forward discrete conversion result of the voiceprint signal, in the discrete domain, which is the voiceprint signal in the continuous domain.

Specifically, for continuous domain voiceprint signalsExtracting to obtain voiceprint signal/>Two-dimensional array of (2)Two-dimensional array/>, based on voiceprint signalsPerforming forward discrete Fourier transform to obtain forward discrete Fourier transform/>Forward discrete Fourier transform/>The expression is as follows:

Wherein, And/>Representing voiceprint signals/>, to successive domainsParameters for extraction,/>Representing forward transform angular frequency,/>And/>Is a discrete frequency.

Wherein a discrete Fourier transform is obtainedAfter the expression, the forward discrete Fourier transform can be subjected to data transposition to obtain the backward discrete Fourier transform containing the two-dimensional Laplacian.

In particular, in the case of voiceprint signalsContinuous domain parameter part of voiceprint signal in time-frequency conversion processConversion to discrete domain parameter part/>Wherein/>Representing a continuous frequency, k representing a discrete frequency. Based on discrete domain parameter part/>For forward discrete Fourier transform/>Performing data transposition to obtain backward discrete Fourier transform/>, including two-dimensional LaplacianBackward discrete fourier transform/>The expression is as follows:

wherein due to the backward discrete Fourier transform Including the two-dimensional laplacian, i.e., the portions other than y and U. Also because the one-dimensional forward and backward transforms cancel each other out, therefore/>，/>AndIndependent of/>And/>。

Therefore, the backward discrete Fourier transform to which the two-dimensional operator belongs can be performed according to the disassembly dimension of the backward discrete transform resultDisassembling to obtain two partial/>, each of which contains one-dimensional operatorsAnd/>。

Specifically, the expression of the partial formula is:

；

Wherein, And/>May represent one-dimensional kernels along the columns and rows of the matrix, respectively, representing the computation/>, in the fourier domainWherein each dimension is calculated independently. At/>In the case of (a) one-dimensional kernel involves/>、/>Length is/>Forward and backward fourier transforms of (2), where and/>Is sandwiched between them, a similar definition relates to/>。

Specifically, after one-dimensional kernels along columns and rows of the matrix are obtained, convolution operation is performed according to the two one-dimensional kernel voiceprint signals, and convolution voiceprint signals are obtained.

In this embodiment, frequency domain conversion is performed on the voiceprint signal to obtain a forward discrete transformation result of the voiceprint signal;

The method comprises the steps of performing data transposition operation based on a forward discrete transformation result to obtain a backward discrete transformation result of a voiceprint signal, wherein the backward discrete transformation result comprises two-dimensional operators, performing disassembly operation on the two-dimensional operators according to disassembly dimensions of the backward discrete transformation result to obtain one-dimensional operators of the backward discrete transformation result, performing convolution operation on the voiceprint signal according to the one-dimensional operators to obtain a convolution voiceprint signal, and decomposing the two-dimensional laplace operators into a plurality of one-dimensional laplace operators.

In one embodiment, performing frequency domain conversion on the voiceprint signal to obtain a forward discrete transform result of the voiceprint signal, including: and carrying out data point sampling on the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals, and generating a forward discrete transformation result of the voiceprint signals according to the discrete frequency of the voiceprint signals in the frequency domain and the two-dimensional array.

Wherein, for the voiceprint signal belonging to the time domainSampling is performed in the frequency domain.

Specifically, the voiceprint signal is firstly processedSampling is performed by having/>Finite square pair voiceprint signal of data points/>Sampling to obtain a two-dimensional array/> -of the voiceprint signal。

Obtaining a two-dimensional array of voiceprint signalsThen, according to the discrete frequency/>, in the frequency domain, of the voiceprint signal、/>And a two-dimensional array for generating the forward discrete transformation result/>, of the voiceprint signal。

Wherein forward discrete transform resultsMay be the forward discrete fourier transform result, expressed as follows:

In this embodiment, data point sampling is performed on the voiceprint signal belonging to the time domain to obtain a two-dimensional array of the voiceprint signal, and according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array, a forward discrete transformation result of the voiceprint signal is generated to obtain an accurate forward discrete transformation result, so that a basis is provided for subsequent data processing according to the forward discrete transformation result, and the evaluation accuracy of the operation state evaluation result of the device is improved.

In one embodiment, S206 includes: performing multiple feature extraction on the convolution voice print signal to obtain each basic voice print feature representing different feature extraction degrees in the convolution voice print signal, wherein the basic voice print feature comprises a basic channel number, weighting the basic voice print feature corresponding to each basic channel number according to the channel weight matched by the basic channel number to obtain feature weights of each basic voice print feature, and performing weighting processing on the basic voice print feature corresponding to each feature weight according to the feature weights to obtain the voice print feature of the convolution voice print signal.

The convolution voiceprint signal is subjected to multiple feature extraction, and the multiple feature extraction is performed on the same convolution voiceprint feature to obtain basic voiceprint features with different feature extraction degrees, wherein the basic voiceprint features comprise basic channel data.

Illustratively, for convolved voiceprint signalsAnd extracting the characteristics to obtain 2 channels of the basic characteristics A, 4 channels of the basic characteristics B, 8 channels of the basic characteristics C and the like.

Specifically, the basic voiceprint feature Z _j can be compressed by the operation of f _sq, the formula of f _sq is as follows:

Wherein h and b are the length and width of the image, i and j are the voiceprint feature layer coordinates in the length and width directions of the voiceprint feature layer, and Z _j (i, j) is the voiceprint feature value of the voiceprint feature layer at the position corresponding to (i, j).

The scale of the compressed image is 1×1×c, and the importance of each channel information can be obtained through adaptive learning f _ex.

Where ω ₁,ω₂,ω₃,…,ω_c is the importance of each channel after adaptive learning.

Specifically, according to the feature weights ω ₁,ω₂,ω₃,…,ω_c, weighting the basic voiceprint features f ₁,f₂,…,f_c corresponding to each feature weight, so as to obtain the voiceprint features of the convolution voiceprint signal.

Illustratively, the importance of each channel is weighted onto the input voiceprint feature layer by the f _sc operation, resulting in the information weighted image Z _j as follows:

Where f ₁,f₂,…,f_c is the base voiceprint feature corresponding to each channel in input Z _j.

In this embodiment, feature extraction is performed on a convolution voiceprint signal for multiple times to obtain each basic voiceprint feature representing different feature extraction degrees in the convolution voiceprint signal, where the basic voiceprint feature includes a basic channel number, weights are performed on the basic voiceprint features corresponding to each basic channel number according to channel weights matched by the basic channel number, feature weights of each basic voiceprint feature are obtained, and weighting processing is performed on the basic voiceprint features corresponding to each feature weight according to the feature weights, so as to obtain voiceprint features reflecting feature importance.

In one embodiment, the operation state of the device is evaluated based on the voiceprint features to obtain an operation state evaluation result of the device, including: and evaluating the running state of the equipment based on the voiceprint characteristics, determining the respective prediction probabilities of the equipment corresponding to the multiple candidate faults, and determining the running state evaluation result of the equipment according to the respective prediction probabilities.

Wherein, the abnormal evaluation result comprises a plurality of candidate faults. For example, for a converter valve cooling apparatus, candidate faults include: bearing damage, bearing wear, loosening of pipeline joints, etc.

Specifically, if the prediction probability of a certain candidate fault is higher, the probability of the candidate fault occurring on the device is higher, and the device is more likely to occur.

The operation state evaluation result of the equipment can be the equipment fault type and also can be the distribution of the equipment fault type. Specifically, for the fault type, a threshold may be set for the prediction probability, and if the prediction probability exceeds the threshold, the candidate fault corresponding to the prediction probability is determined as the equipment fault type. Specifically, for fault distribution, the prediction probability of each candidate fault can be counted to obtain an equipment fault distribution table, and further analysis can be performed according to the fault distribution table, so that maintenance personnel can be helped to arrange a maintenance plan.

For example, the processing result of the voiceprint feature can be subjected to pattern recognition by using a Softmax function. The Softmax layer probability is calculated as follows:

In this embodiment, the operation state of the device is evaluated based on the voiceprint feature, so that the respective prediction probabilities of the device corresponding to the multiple candidate faults are determined, and the operation state evaluation result of the device is determined according to the respective prediction probabilities, so that the service life loss of the device due to the dismounting process is reduced on the premise that the device is not required to be dismounted.

In one embodiment, the operation state of the device is evaluated based on the voiceprint features to obtain an operation state evaluation result of the device, including:

acquiring an initial evaluation model for the evaluated equipment;

Based on the equipment association relation between the equipment to be evaluated and the evaluated equipment, performing transfer learning on the initial evaluation equipment to obtain an updated evaluation model matched with the equipment to be evaluated;

and inputting the voiceprint characteristics into an updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

Wherein the initial assessment model may be a model set for the assessed device. And evaluating the running state of the equipment through the initial evaluation model to obtain an evaluation result of the running state of the equipment.

The devices to be evaluated may be the same type of devices as the devices to be evaluated, for example, both may be converter valve cooling devices. The device to be evaluated may also be a different type of device than the one being evaluated. The device association relationship between the device to be evaluated and the evaluated device indicates that there is a position, type and time association between the device to be evaluated and the evaluated device.

Specifically, performing the migration learning on the initial evaluation apparatus includes: and acquiring an initial evaluation model, acquiring target domain data of the equipment to be evaluated, and performing transfer learning on the initial evaluation model according to the target domain data to obtain an updated evaluation model matched with the equipment to be evaluated.

And performing migration learning on the initial evaluation equipment to reduce the distribution difference between the source domain and the target domain, so as to realize feature migration and construct a small sample detection model. Essentially, a mapping function is found that minimizes the distance between the transformed source domain data and the target domain data.

Maximum mean difference: given two sets of model features xs= {1,2, … ns } and xt= {1,2, … nt } perform a nonlinear mapping of the hilbert space as an inner product with a kernel function, the difference between the two domains is measured:

The empirical estimate of the two distribution errors is taken as the distance between the two data distributions in the hilbert space. Here, x ^s _i and x ^t _i are nonlinear mapping functions in the reconstruction kernel hilbert space, and k is a kernel function for mapping source and target domain data to the hilbert space.

The source domain model (initial evaluation model) is adjusted by minimizing the maximum average difference between the source domain and the target domain. The goal of this step is to make the distribution between the two domains closer during migration to improve the generalization ability of the model over the target domains.

Since the same category has stronger correlation, we improve on this basis, match the global distribution and local distribution of the category by aligning the relevant subdomains of the samples with the same label, add the aligned maximum average difference to align the distribution of the relevant subdomains of the same category, the expression is as follows:

Where ω ^sc _i and ω ^tc _i are the weights of instances x _i ^s and x _i ^t belonging to class c; c is the number of categories in the dataset; y _ic is the c-th element of the tag vector y _i.

In this embodiment, by acquiring an initial evaluation model for an evaluated device, performing migration learning on the initial evaluation device based on a device association relationship between the device to be evaluated and the evaluated device, obtaining an updated evaluation model matched with the device to be evaluated, and inputting voiceprint features into the updated evaluation model, so as to obtain an accurate evaluation result of the running state of the device to be evaluated.

In one embodiment, the device may be a soft direct converter station valve cooling device, all referred to as a soft direct converter station valve cooling device, which is a device for cooling a soft direct converter valve. The flexible direct current converter valve is core equipment of a flexible direct current transmission system and is used for converting three-phase alternating current into direct current or reverse alternating current and controlling power. Since the converter valve generates a large amount of heat during operation, a device is required to efficiently remove the heat and maintain the normal operating temperature of the converter valve.

The soft direct converter station valve cooling equipment usually adopts a water cooling mode, takes away the heat of the converter valve through circulating water, and then discharges the heat to the outside through refrigeration equipment. The device can ensure that the converter valve can operate efficiently and maintain stable temperature, so that the service life of the device is prolonged, and the stability and reliability of the system are improved.

When internal faults occur in the valve cooling equipment of the flexible direct current converter station, such as bearing damage, abnormal noise of a gear box and the like, the equipment cannot normally operate or efficiency is reduced, and serious electric safety accidents are caused. Therefore, the machine disassembly detection is carried out once a month by staff, but the method is time-consuming and labor-consuming, and each reinstallation has irreparable service life loss for equipment. In view of the existing problems, there is a need for a method for predicting the status of a device that does not require disassembly.

The inventor finds that when internal faults occur or the service life of the soft direct converter station valve cooling equipment is near, abnormal vibration sound is generated by the equipment, the sound usually appears as continuous buzzing or crunching, the voiceprint characteristics are collected by means of an acoustic detection instrument based on the abnormal vibration sound, and the fault detection is carried out by utilizing the voiceprint characteristics, so that the equipment is not required to be disassembled, and the service life loss caused by the disassembly of the equipment is avoided.

In view of this, a flow chart of a device operation state evaluation method based on an attention mechanism and parallel migration acceleration as shown in fig. 3 includes:

s302, a voiceprint signal generated by equipment to be evaluated in the running process is obtained.

S304, sampling data points of the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals.

S306, generating a forward discrete transformation result of the voiceprint signal according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array.

And S308, performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal.

Wherein the backward discrete transformation result comprises a two-dimensional operator.

S310, according to the disassembly dimension of the backward discrete transformation result, carrying out disassembly operation on the two-dimensional operator to obtain a one-dimensional operator of the backward discrete transformation result.

S312, carrying out convolution operation on the voiceprint signal according to the one-dimensional operator to obtain a convolution voiceprint signal.

S314, under the condition that the convolution voiceprint signal meets the noise reduction screening condition, carrying out feature extraction on the convolution voiceprint signal for a plurality of times to obtain each basic voiceprint feature representing different feature extraction degrees in the convolution voiceprint signal, wherein the basic voiceprint features comprise basic channel numbers.

S316, weighting the basic voiceprint features corresponding to each basic channel number according to the channel weights matched with the basic channel number, and obtaining the feature weights of the basic voiceprint features.

And S318, weighting the basic voiceprint features corresponding to each feature weight according to the feature weights to obtain the voiceprint features of the convolution voiceprint signal.

S320, the running state of the equipment is evaluated based on the voiceprint characteristics, and the prediction probability of the equipment corresponding to each of a plurality of candidate faults is determined.

S322, determining an operation state evaluation result of the equipment according to each prediction probability.

S324, an initial assessment model for the assessed device is acquired.

And S326, performing migration learning on the initial evaluation equipment based on the equipment association relation between the equipment to be evaluated and the evaluated equipment to obtain an updated evaluation model matched with the equipment to be evaluated.

S328, inputting the voiceprint characteristics into an updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

In the embodiment, the voiceprint signal generated by the equipment to be evaluated in the operation process is obtained, the voiceprint signal is subjected to convolution operation according to the transformation operator generated by the voiceprint signal in the time-frequency transformation, the convolution voiceprint signal is obtained, and the transformation operator is adopted to participate in the convolution operation, so that the data processing speed is improved. Under the condition that the convolution voiceprint signal meets noise reduction screening conditions, feature extraction is carried out on the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal.

It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.

Based on the same inventive concept, the embodiment of the application also provides a device operation state evaluation device for realizing the above related device operation state evaluation method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation in the embodiment of the device operation state evaluation device or devices provided below may refer to the limitation of the device operation state evaluation method hereinabove, and will not be repeated herein.

In one embodiment, as shown in fig. 4, there is provided an apparatus for evaluating an operation state of a device, including: an acquisition module 402, a convolution module 404, an extraction module 406, and an evaluation module 408, wherein:

An acquisition module 402, configured to acquire a voiceprint signal generated by a device to be evaluated in a running process;

The convolution module 404 is configured to perform convolution operation on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation, so as to obtain a convolved voiceprint signal;

the extracting module 406 is configured to perform feature extraction on the convolved voiceprint signal to obtain voiceprint features of the convolved voiceprint signal when the convolved voiceprint signal meets a noise reduction screening condition;

And the evaluation module 408 is configured to evaluate the operation state of the device based on the voiceprint feature, so as to obtain an operation state evaluation result of the device.

In one embodiment, the convolution module 404 is further configured to perform frequency domain conversion on the voiceprint signal to obtain a forward discrete transform result of the voiceprint signal; performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal; the backward discrete transformation result comprises a two-dimensional operator; according to the disassembly dimension of the backward discrete transformation result, carrying out disassembly operation on the two-dimensional operator to obtain a one-dimensional operator of the backward discrete transformation result; and carrying out convolution operation on the voiceprint signal according to the one-dimensional operator to obtain a convolution voiceprint signal. .

In one embodiment, the convolution module 404 is further configured to sample data points of the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals; and generating a forward discrete transformation result of the voiceprint signal according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array.

In one embodiment, the extracting module 406 is further configured to perform feature extraction on the convolved voiceprint signal multiple times to obtain each basic voiceprint feature in the convolved voiceprint signal, where each basic voiceprint feature represents different feature extraction degrees, and the basic voiceprint feature includes a basic channel number; weighting the basic voiceprint features corresponding to each basic channel number according to the channel weight matched with the basic channel number to obtain the feature weight of each basic voiceprint feature; and weighting the basic voiceprint features corresponding to each feature weight according to the feature weights to obtain the voiceprint features of the convolution voiceprint signal.

In one embodiment, the evaluation module 408 is further configured to evaluate an operation state of the device based on the voiceprint feature, and determine a prediction probability of the device corresponding to each of the plurality of candidate faults; and determining an operation state evaluation result of the equipment according to each prediction probability.

In one embodiment, the evaluation module 408 is further configured to obtain an initial evaluation model for the evaluated device; based on the equipment association relation between the equipment to be evaluated and the evaluated equipment, performing transfer learning on the initial evaluation equipment to obtain an updated evaluation model matched with the equipment to be evaluated; and inputting the voiceprint characteristics into an updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

The respective modules in the above-described apparatus operation state evaluation device may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is for storing voiceprint signal data. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a device operating state evaluation method.

It will be appreciated by those skilled in the art that the structure shown in FIG. 5 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.

In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of:

Acquiring a voiceprint signal generated by equipment to be evaluated in the running process; performing convolution operation on the voiceprint signal according to a transformation operator generated in time-frequency transformation of the voiceprint signal to obtain a convolution voiceprint signal; under the condition that the convolution voiceprint signal meets the noise reduction screening condition, carrying out feature extraction on the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal; and evaluating the running state of the equipment based on the voiceprint characteristics to obtain an evaluation result of the running state of the equipment.

In one embodiment, the processor when executing the computer program further performs the steps of:

Performing frequency domain conversion on the voiceprint signal to obtain a forward discrete conversion result of the voiceprint signal; performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal; the backward discrete transformation result comprises a two-dimensional operator; according to the disassembly dimension of the backward discrete transformation result, carrying out disassembly operation on the two-dimensional operator to obtain a one-dimensional operator of the backward discrete transformation result; and carrying out convolution operation on the voiceprint signal according to the one-dimensional operator to obtain a convolution voiceprint signal.

carrying out data point sampling on the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals; and generating a forward discrete transformation result of the voiceprint signal according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array.

Performing multiple feature extraction on the convolution voiceprint signal to obtain basic voiceprint features representing different feature extraction degrees in the convolution voiceprint signal, wherein the basic voiceprint features comprise basic channel numbers; weighting the basic voiceprint features corresponding to each basic channel number according to the channel weight matched with the basic channel number to obtain the feature weight of each basic voiceprint feature; and weighting the basic voiceprint features corresponding to each feature weight according to the feature weights to obtain the voiceprint features of the convolution voiceprint signal.

Evaluating the running state of the equipment based on the voiceprint characteristics, and determining the respective prediction probabilities of the equipment corresponding to the multiple candidate faults; and determining an operation state evaluation result of the equipment according to each prediction probability.

Acquiring an initial evaluation model for the evaluated equipment; based on the equipment association relation between the equipment to be evaluated and the evaluated equipment, performing transfer learning on the initial evaluation equipment to obtain an updated evaluation model matched with the equipment to be evaluated; and inputting the voiceprint characteristics into an updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, the computer program when executed by the processor further performs the steps of:

In one embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, performs the steps of:

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are both information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data are required to meet the related regulations.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magneto-resistive random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (PHASE CHANGE Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in various forms such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims

1. A method of evaluating an operational state of a device, the method comprising:

Acquiring a voiceprint signal generated by equipment to be evaluated in the running process;

Performing convolution operation on the voiceprint signal according to a transformation operator generated in time-frequency transformation of the voiceprint signal to obtain a convolution voiceprint signal;

Under the condition that the convolution voiceprint signal meets noise reduction screening conditions, extracting features of the convolution voiceprint signal to obtain voiceprint features of the convolution voiceprint signal;

and evaluating the running state of the equipment based on the voiceprint characteristics to obtain a running state evaluation result of the equipment.

2. The method of claim 1, wherein the transformation operator comprises a two-dimensional operator and a one-dimensional operator;

The convolution operation is carried out on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal, and the method comprises the following steps:

performing frequency domain conversion on the voiceprint signal to obtain a forward discrete transformation result of the voiceprint signal;

Performing data transposition operation based on the forward discrete transformation result to obtain a backward discrete transformation result of the voiceprint signal; the backward discrete transformation result comprises a two-dimensional operator;

according to the disassembly dimension of the backward discrete transformation result, carrying out disassembly operation on the two-dimensional operator to obtain a one-dimensional operator of the backward discrete transformation result;

And carrying out convolution operation on the voiceprint signal according to the one-dimensional operator to obtain a convolution voiceprint signal.

3. The method according to claim 2, wherein said performing frequency domain conversion on said voiceprint signal to obtain forward discrete transform results of said voiceprint signal comprises:

Carrying out data point sampling on the voiceprint signals belonging to the time domain to obtain a two-dimensional array of the voiceprint signals;

and generating a forward discrete transformation result of the voiceprint signal according to the discrete frequency of the voiceprint signal in the frequency domain and the two-dimensional array.

4. The method of claim 1, wherein the performing feature extraction on the convolved voiceprint signal to obtain a voiceprint feature of the convolved voiceprint signal comprises:

Performing multiple feature extraction on the convolution voiceprint signal to obtain basic voiceprint features representing different feature extraction degrees in the convolution voiceprint signal, wherein the basic voiceprint features comprise basic channel numbers;

Weighting the basic voiceprint features corresponding to each basic channel number according to the channel weight matched with the basic channel number to obtain the feature weight of each basic voiceprint feature;

And weighting the basic voiceprint features corresponding to each feature weight according to the feature weights to obtain the voiceprint features of the convolution voiceprint signals.

5. The method according to claim 1, wherein the evaluating the operation state of the device based on the voiceprint features to obtain the operation state evaluation result of the device includes:

evaluating the running state of the equipment based on the voiceprint features, and determining the respective prediction probabilities of the equipment corresponding to a plurality of candidate faults;

And determining an operation state evaluation result of the equipment according to each prediction probability.

6. The method according to claim 1, wherein the evaluating the operation state of the device based on the voiceprint features to obtain the operation state evaluation result of the device includes:

acquiring an initial evaluation model for the evaluated equipment;

And inputting the voiceprint features into the updated evaluation model to obtain an operation state evaluation result of the equipment to be evaluated.

7. An apparatus for evaluating an operating state of a device, the apparatus comprising:

The acquisition module is used for acquiring voiceprint signals generated in the running process of the equipment to be evaluated;

The convolution module is used for carrying out convolution operation on the voiceprint signal according to a transformation operator generated by the voiceprint signal in time-frequency transformation to obtain a convolution voiceprint signal;

The extraction module is used for extracting the characteristics of the convolution voiceprint signal under the condition that the convolution voiceprint signal meets the noise reduction screening condition to obtain the voiceprint characteristics of the convolution voiceprint signal;

And the evaluation module is used for evaluating the running state of the equipment based on the voiceprint characteristics to obtain a running state evaluation result of the equipment.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.

9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.