CN110415835B - Method and device for predicting residual life of mechanical equipment - Google Patents

Method and device for predicting residual life of mechanical equipment Download PDF

Info

Publication number
CN110415835B
CN110415835B CN201811125938.0A CN201811125938A CN110415835B CN 110415835 B CN110415835 B CN 110415835B CN 201811125938 A CN201811125938 A CN 201811125938A CN 110415835 B CN110415835 B CN 110415835B
Authority
CN
China
Prior art keywords
state data
state
layer
cluster
clusters
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201811125938.0A
Other languages
Chinese (zh)
Other versions
CN110415835A (en
Inventor
孙梅玉
孙帮成
李明高
刘天赋
齐洪峰
王坚
戴毅茹
周一青
张强
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
CRRC Industry Institute Co Ltd
Original Assignee
Tongji University
CRRC Industry Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tongji University, CRRC Industry Institute Co Ltd filed Critical Tongji University
Priority to CN201811125938.0A priority Critical patent/CN110415835B/en
Publication of CN110415835A publication Critical patent/CN110415835A/en
Application granted granted Critical
Publication of CN110415835B publication Critical patent/CN110415835B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention discloses a method and a device for predicting the residual life of mechanical equipment. Wherein the method comprises the following steps: acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset; obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset; predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established. The device is used for executing the method. According to the method and the device for predicting the residual life of the mechanical equipment, the residual life of the mechanical equipment is predicted by combining the single-layer sensor model and the back propagation neural network model, so that the accuracy of predicting the residual life of the mechanical equipment is improved.

Description

Method and device for predicting residual life of mechanical equipment
Technical Field
The invention relates to the technical field of mechanical equipment, in particular to a method and a device for predicting the residual life of the mechanical equipment.
Background
The mechanical equipment consists of various parts, the parts of the mechanical equipment are gradually worn and aged in the long-term operation process, the residual service life is gradually reduced, the mechanical equipment is finally stopped or even has safety accidents, and unreasonable replacement of the parts causes waste, so that the residual service life of the mechanical equipment is correctly predicted, and the method has great significance for guaranteeing the safe operation of the mechanical equipment and improving the economic benefit.
In the prior art, methods for estimating the remaining life of mechanical equipment can be divided into three categories: a physical model-based remaining life estimate, a knowledge model-based remaining life estimate, and a data-driven model-based remaining life estimate. The remaining life estimation based on the physical model is usually based on differential equations to describe various working conditions of the mechanical equipment, however, the physical model of the mechanical equipment is usually difficult to obtain, and the method for estimating the remaining life of the mechanical equipment based on the physical model has no universality. In the estimation of the remaining life based on the knowledge model, an expert system is often adopted, however, the knowledge model adopting the expert system has very high requirements on the qualification of experts and the mastery degree of the experts on the knowledge in the field, and it is difficult to find a proper expert to ensure the accuracy of the estimation of the remaining life of the mechanical equipment. The method for estimating the residual life based on the data driving model is based on a statistical principle, information and knowledge hidden in massive data are captured by collecting massive data, the current commonly used data driving model is established through an artificial neural network or a support vector machine, a neural network model or a support vector machine model is established, and then a connection weight between layers of the neural network, the number of nodes of a hidden layer and a penalty coefficient of the support vector machine are optimized respectively through a corresponding algorithm, such as a genetic algorithm, a particle swarm algorithm or a gradient descent algorithm, so that the data driving model is obtained. However, since the acquired data often has highly nonlinear characteristics, the data-driven model is often difficult to converge or easily falls into a locally optimal condition, and the obtained prediction result is not accurate enough.
Therefore, how to provide a method for predicting the remaining life of a mechanical device, which can improve the accuracy of the prediction of the remaining life of the mechanical device, is an important issue to be solved in the industry.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a method and a device for predicting the residual life of mechanical equipment.
In one aspect, the present invention provides a method for predicting a remaining life of a mechanical device, including:
acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset;
obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset;
predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
In another aspect, the present invention provides a remaining life prediction apparatus for a mechanical device, including:
the acquisition unit is used for acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset;
the obtaining unit is used for obtaining an intermediate prediction result according to the state data and the single-layer sensor model corresponding to each state parameter; wherein the single-layer perceptron model is preset;
the prediction unit is used for predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
In yet another aspect, the present invention provides an electromechanical device comprising: a processor, a memory, and a communication bus, wherein:
the processor and the memory are communicated with each other through the communication bus;
the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the method for predicting the residual life of the mechanical equipment provided by the embodiments.
In yet another aspect, the present invention provides a non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform a method for predicting remaining life of a mechanical device as provided in the above embodiments.
According to the method and the device for predicting the residual life of the mechanical equipment, the state data corresponding to each state parameter of the mechanical equipment in the preset time period can be obtained, the intermediate prediction result is obtained according to the state data corresponding to each state parameter and the single-layer sensor model, then the back propagation neural network model is predicted according to the intermediate prediction result and the residual life, and the residual life of the mechanical equipment is predicted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a method for predicting remaining life of a mechanical device according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating a method for predicting remaining life of a mechanical device according to another embodiment of the present invention;
FIG. 3 is a flowchart illustrating a method for predicting remaining life of a mechanical device according to another embodiment of the present invention;
FIG. 4 is a flowchart illustrating a method for predicting remaining life of a mechanical device according to another embodiment of the present invention;
fig. 5 is a schematic structural diagram of a remaining life prediction apparatus for a mechanical device according to an embodiment of the present invention;
fig. 6 is a schematic physical structure diagram of an electromechanical device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flowchart of a method for predicting remaining life of a mechanical device according to an embodiment of the present invention, and as shown in fig. 1, the method for predicting remaining life of a mechanical device according to the present invention includes:
s101, acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset;
specifically, when the mechanical device is running, the sensor arranged on the mechanical device may be used to obtain the state data corresponding to each state parameter of the mechanical device in the preset time period, where the state parameter may be a state parameter of a main component that affects the running of the mechanical device, and may affect the remaining life of the mechanical device. A remaining life prediction device (hereinafter referred to as a remaining life prediction device) of the mechanical device may obtain state data corresponding to each of the state parameters of the mechanical device within the preset time period. The state parameters are preset and are set according to actual experience, and the embodiment of the invention is not limited; the preset time period is set according to actual needs, and the embodiment of the invention is not limited.
For example, the mechanical device is a wind power generator, and each of the state parameters is a temperature of a lubricating oil of the gearbox, a vibration frequency of the gearbox, a temperature of a rotor winding, a converter current, a converter voltage, a temperature of the engine, a rotational speed of the engine, and a vibration frequency of the engine.
S102, obtaining an intermediate prediction result according to the state data and the single-layer sensor model corresponding to each state parameter; the single-layer perceptron model is preset;
specifically, after obtaining the state data corresponding to each state parameter, the remaining life prediction apparatus may obtain an output result of the single-layer sensor by using the state data corresponding to each state parameter as an input of the single-layer sensor, where the output result is an intermediate prediction result. Wherein the single-layer perceptron model is preset.
S103, predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
Specifically, after obtaining the intermediate prediction result, the remaining life prediction apparatus may obtain an output result of the remaining life prediction back propagation neural network model by using the intermediate prediction result as an input of the remaining life prediction back propagation neural network model, where the output result of the remaining life prediction back propagation neural network model is the remaining life of the mechanical device, so as to predict the remaining life of the mechanical device. Wherein the remaining life prediction back propagation neural network model is pre-established.
For example, the remaining life predicting apparatus obtains historical state data corresponding to each of the state parameters, and obtains a second preset number of sets of initial training data according to the historical state data corresponding to each of the state parameters, where each set of initial training data includes the historical state data corresponding to each of the state parameters in the preset time period; then obtaining a second preset number of groups of residual life prediction back propagation neural network model training data according to each group of the initial training data and the single-layer perceptron model; then training initial back propagation neural network model training data based on the second preset number of groups of the residual life prediction back propagation neural network model training data to obtain a residual life prediction back propagation neural network model; wherein the initial back propagation neural network model includes a hidden layer.
According to the method for predicting the residual life of the mechanical equipment, the state data corresponding to each state parameter of the mechanical equipment in the preset time period can be obtained, the intermediate prediction result is obtained according to the state data corresponding to each state parameter and the single-layer sensor model, then the back propagation neural network model is predicted according to the intermediate prediction result and the residual life, and the residual life of the mechanical equipment is predicted.
On the basis of the foregoing embodiments, further, the transfer function of the output layer of the single-layer sensor model is:
Figure BDA0001812386470000061
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure BDA0001812386470000062
wherein, CiThe connection weight of the ith input layer node of the input layer of the single-layer perceptron model, n is the number of the input layer nodes of the input layer of the single-layer perceptron model, HiIs obtained according to the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron modeljThe connection weight and the first threshold are preset as a first threshold of a jth output layer node of an output layer of the single-layer perceptron model.
Specifically, the single-layer perceptron is composed of two layers of neurons, an input layer and an output layer, wherein the input layer comprises a plurality of input layer nodes, and the output layer comprises at least one output layer node. In this embodiment of the present invention, the state parameters correspond to the input layer nodes one to one, the state data corresponding to the state parameters are input into the single-layer perceptron from the corresponding input layer nodes, the state data input into the input layer nodes of the single-layer perceptron is the state data corresponding to the input layer nodes of the single-layer perceptron model, and the number of the output layer nodes is set according to actual experience. The transfer function of the output layer node is a Sigmoid function:
Figure BDA0001812386470000063
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure BDA0001812386470000071
wherein, CiIs the ith input layer of the input layers of the single-layer perceptron modelThe connection weight of the node is preset and is set according to actual experience, and the embodiment of the invention is not limited; hiObtained according to state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model, HiThe specific obtaining process is described in the following; n is the number of input layer nodes of the input layer of the single-layer perceptron model, thetajThe first threshold is a first threshold of a j-th output layer node of an output layer of the single-layer perceptron model, and the first threshold is preset and set according to practical experience, and the embodiment of the invention is not limited. Wherein i and j are positive integers.
Fig. 2 is a schematic flow chart of a method for predicting remaining life of a mechanical device according to another embodiment of the present invention, as shown in fig. 2, based on the foregoing embodiments, further, the step H is performediThe obtaining of the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model comprises the following steps:
s201, performing cluster analysis on state data corresponding to an ith input layer node of an input layer of the single-layer sensor model to obtain a first preset number of clusters, wherein the first preset number of clusters comprises m-1 normal state clusters and one abnormal state cluster; wherein m is the first preset number;
specifically, when the state data corresponding to each state parameter is input into the single-layer perceptron, the state data corresponding to one corresponding state parameter is input into the ith input layer node, and the state data input into the ith input layer node becomes the state data corresponding to the ith input layer node. The residual life prediction device performs cluster analysis on state data corresponding to an ith input layer node of an input layer of the single-layer sensor model to obtain a first preset number of clusters, wherein the first preset number of clusters comprises m-1 normal state clusters and one abnormal state cluster; the first preset number is set according to an actual situation, and the embodiment of the invention is not limited. It is understood that m is the first predetermined number.
For example, the remaining life predicting apparatus randomly selects m-1 state data from state data corresponding to an ith input layer node of an input layer of the single-layer sensor model, and respectively uses the m-1 state data as a cluster center of m-1 normal state clusters, and calculates an average value of the m-1 state data as an initial average value. Then, the distance between each state data in the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model and the initial average value is calculated. Secondly, after judging that the distance is larger than or equal to a second threshold value, classifying the state data corresponding to the distance into the abnormal state cluster; after judging that the distance is smaller than the second threshold value, classifying the state data corresponding to the distance into a normal state cluster corresponding to the state data with the shortest distance in the m-1 state data. And thirdly, after the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model are clustered once, calculating to obtain the current average value according to the state data of the m-1 normal state clusters. Then, after judging and knowing that the current average value is equal to the initial average value, outputting m-1 normal state clusters and one abnormal state cluster as the first preset number of clusters; otherwise, updating the initial average value to be the current average value, updating the clustering centers of the m-1 normal state clusters, and clustering again until the current average value obtained after clustering again is equal to the initial average value of the last clustering.
S202, obtaining an average value of the number of the state data of the normal clusters according to the number of the state data of each normal cluster in m-1 normal clusters, and obtaining a result of a product of the average value of the number of the state data of the normal clusters and a preset value;
specifically, after obtaining m-1 normal state clusters, the remaining life predicting apparatus may statistically obtain the number of state data included in each of the m-1 normal state clusters, where the number of state data included in the normal state cluster is the number of state data of the normal state cluster, then calculate and obtain an average value of the number of state data of the normal clusters according to the number of state data of each normal state cluster, and multiply the average value of the number of state data of the normal clusters by a preset value to obtain a result of a product of the average value of the number of state data of the normal clusters and the preset value. The preset value is set according to actual experience, and the embodiment of the invention is not limited.
For example, the remaining life predicting means obtains 7 normal state clusters, and the number of state data included in the 7 normal state clusters is q, respectively1、q2、q3、q4、q5、q6And q is7Average value of the number of state data of the normal cluster
Figure BDA0001812386470000091
Assuming that the preset value is 30%, the result of the product of the average value of the number of state data of the normal cluster and the preset value
Figure BDA0001812386470000092
S203, comparing the state data quantity of each normal state cluster of m-1 normal state clusters with the result of the product to obtain a normal state cluster of which the state data quantity of the normal state cluster is less than the result of the product;
specifically, the remaining life predicting means may obtain the normal state cluster whose number of state data is smaller than the result of the product by comparing the number of state data of each of the m-1 normal state clusters with the result of the product after obtaining the result of the product of the average value of the number of state data of the normal cluster and a preset value.
For example, the remaining life predicting means obtains the result of the multiplication as Q, and the numbers of the state data included in 7 of the normal state clusters are Q, respectively1、q2、q3、q4、q5、q6And q is7. The remaining life predicting means predicts the remaining life of q1、q2、q3、q4、q5、q6And q is7Are compared with Q respectively, if Q is1Greater than Q, Q2Less than Q, Q3Greater than Q, Q4Greater than Q, Q5Greater than Q, Q6Greater than Q, Q7Less than Q, then the normal state cluster whose state data quantity is less than the result of the product is: q. q.s2Corresponding to the normal cluster and q7The corresponding normal cluster.
S204, obtaining the quantity of abnormal state data according to the quantity of the state data of the abnormal state cluster and the quantity of the state data of the normal state cluster which is smaller than the quantity of the state data of the normal state cluster of the product result;
specifically, the remaining life predicting means adds the number of state data of the normal state cluster, the number of state data of which is smaller than the result of the multiplication, of each of the normal state clusters to the number of state data of the abnormal state cluster to obtain the number of abnormal state data, after obtaining the normal state clusters, the number of state data of which is smaller than the result of the multiplication.
For example, the remaining life predicting means obtains two normal state clusters having state data numbers smaller than the result of the multiplication, the corresponding state data numbers being q2And q is7The number of state data of the abnormal state cluster is Q', and the number of abnormal state data QDifferent from each other＝q2+q7+q′。
S205, obtaining H according to the state data quantity and the abnormal state data quantity of the first cluster with the preset quantityi。
Specifically, the remaining lifetime prediction apparatus may statistically obtain the number of state data of m-1 normal state clusters and the number of state data of abnormal state clusters in m clusters, use the sum of the number of state data of each normal state cluster and the number of state data of abnormal state clusters as the number of state data of the first preset number of clusters, and then divide the number of state data of the first preset number of clusters by the number of abnormal state data to obtain Hi。
For example, the remaining life predicting means obtains the number q of pieces of state data included in 7 of the normal state clusters, respectively1、q2、q3、q4、q5、q6And q is7And the number of state data of the abnormal state cluster is Q', and the number of state data of the first preset number of clusters is QGeneral assembly＝q1+q2+q3+q4+q5+q6+q7+ Q', the number of abnormal state data QDifferent from each otherThen, then
Figure BDA0001812386470000101
Fig. 3 is a schematic flow chart of a method for predicting a remaining life of a mechanical device according to another embodiment of the present invention, as shown in fig. 3, based on the foregoing embodiments, further performing cluster analysis on state data corresponding to an i-th input layer node of an input layer of the single-layer sensor model to obtain a first preset number of clusters, where the first preset number of clusters includes m-1 normal state clusters and an abnormal state cluster, and the method includes:
s2011, randomly selecting m-1 state data from state data corresponding to an ith input layer node of an input layer of the single-layer sensor model, respectively taking each state data of the m-1 state data as a clustering center of the normal state cluster, and calculating an average value of the clustering centers of the m-1 normal state clusters as an initial average value;
specifically, when the state data corresponding to each state parameter is input into the single-layer perceptron, the state data corresponding to one corresponding state parameter is input into the ith input layer node, and the state data input into the ith input layer node becomes the state data corresponding to the ith input layer node. The residual life prediction device randomly selects m-1 state data from state data corresponding to the ith input layer node of the input layer of the single-layer sensor model, and takes each state data in the m-1 state data as a clustering center of the normal state cluster to obtain the clustering centers of the m-1 normal state clusters. The residual life prediction device obtains the clustering centers of m-1 normal state clusters, and takes the average value of the clustering centers of m-1 normal state clusters as an initial average value.
S2012, calculating the distance between each state data in the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model and the initial average value;
specifically, the remaining life predicting means may calculate, after obtaining the initial average value, a distance between each of the state data corresponding to the ith input layer node of the input layer from which the single-layer perceptron model is obtained and the initial average value.
S2013, if the distance is judged to be larger than a second threshold value, classifying the state data corresponding to the distance into the abnormal state cluster;
specifically, after calculating the distance between the state data and the initial average value, the remaining life prediction device compares the distance with a second threshold, and if the distance is greater than the second threshold, classifies the state data corresponding to the distance into the abnormal state cluster. The second threshold is set according to practical experience, and the embodiment of the present invention is not limited.
S2014, if the distance is judged to be smaller than or equal to the second threshold, classifying the state data corresponding to the distance into a normal state cluster corresponding to the state data closest to the m-1 state data;
specifically, the remaining life predicting means compares the distance between the state data and the initial average value with a second threshold value after calculating the distance, if the distance is less than or equal to the second threshold, calculating the distance between the state data corresponding to the distance and each of the m-1 state data, i.e. the distance from the state data corresponding to the distance to the cluster center of each normal state cluster is calculated, m-1 distances can be obtained, comparing m-1 distances to obtain the minimum value of m-1 distances, classifying the state data corresponding to the distances into the normal state cluster of the cluster center corresponding to the minimum value of m-1 distances, namely, the state data corresponding to the distance is classified into the normal state cluster corresponding to the state data with the shortest distance in the m-1 state data.
S2015, after the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model are clustered for the first time, recalculating the clustering centers of m-1 normal state clusters according to the state data of m-1 normal state clusters, and taking the average value of the recalculating clustering centers of m-1 normal state clusters as the current average value;
specifically, the remaining life predicting apparatus performs step S2013 or step S2014 on each state data of the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model, that is, the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model is once clustered, and the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model can be classified into m-1 normal state clusters and one abnormal state cluster, respectively. The residual life prediction device can recalculate the cluster centers of the m-1 normal state clusters according to the state data of the m-1 normal state clusters obtained by the primary clustering, and take the average value of the cluster centers of the m-1 normal state clusters obtained by recalculation as the current average value.
For example, the remaining life predicting apparatus randomly selects 7 pieces of the state data, and after one-time clustering, obtains 7 pieces of the normal state clusters. The remaining life predicting device may recalculate all the state data of the 7 normal state clusters obtained by clustering to obtain the clustering centers of the 7 normal state clusters, that is, sum all the state data of one normal state cluster, and divide the sum by the number of the state data of the normal state cluster, thereby obtaining the clustering center of the normal state cluster, and then calculate the average value of the recalculated clustering centers of the 7 normal state clusters, and take the average value as the current average value.
S2016, if the current average value is judged to be equal to the initial average value, outputting m-1 normal state clusters and one abnormal state cluster as the first preset number of clusters; otherwise, updating the initial average value to be the current average value, updating the clustering centers of the m-1 normal state clusters, and clustering again until the current average value obtained after clustering again is equal to the initial average value of clustering again.
Specifically, the remaining life predicting means compares the current average value with the initial average value after obtaining the current average value, and outputs m-1 normal state clusters and one abnormal state cluster as the first preset number of clusters if the current average value is equal to the initial average value. If the current average value is not equal to the initial average value, updating the initial average value to the current average value, that is, replacing the initial average value with the current average value as an initial average value for re-clustering, respectively calculating an average value of state data of each normal state cluster as a clustering center for re-clustering according to the obtained state data of m-1 normal state clusters, repeating steps S2012, S2013, S2014 and S2015, re-clustering once, and re-obtaining the current average value. The remaining life predicting means compares the current average value obtained by re-clustering with the initial average value of re-clustering, and if the current average value obtained after re-clustering is not equal to the initial average value of re-clustering, repeats the above re-clustering process, and re-clustering until the current average value obtained after re-clustering is equal to the initial average value of re-clustering. If the current average value obtained after the re-clustering is equal to the initial average value of the re-clustering, the remaining life prediction apparatus takes m-1 normal state clusters and abnormal state clusters obtained by the re-clustering as the first preset number of clusters.
On the basis of the above embodimentsFurther, H is obtained according to the number of the state data of the first cluster with the preset number and the number of the abnormal state dataiThe method comprises the following steps:
according to the formula
Figure BDA0001812386470000141
Calculating to obtain HiWherein Q isDifferent from each otherFor said number of abnormal state data, QGeneral assemblyThe number of state data for the first predetermined number of clusters.
Specifically, the remaining life predicting means obtains the abnormal state data amount QDifferent from each otherAnd obtaining the state data quantity Q of the first cluster with the preset quantityGeneral assemblyThen, can be according to the formula
Figure BDA0001812386470000142
Calculating to obtain Hi。
Fig. 4 is a schematic flowchart of a remaining life prediction method for a mechanical device according to still another embodiment of the present invention, and as shown in fig. 4, the step of building the remaining life prediction back propagation neural network model includes:
s401, obtaining historical state data corresponding to each state parameter, and obtaining a second preset number of groups of initial training data according to the historical state data corresponding to each state parameter, wherein each group of initial training data comprises the historical state data corresponding to each state parameter in the preset time period;
in particular, sufficient training data is required in order to build the residual life prediction back propagation neural network model. By recording the state data corresponding to each state parameter of the mechanical equipment, the historical state data corresponding to each state parameter can be obtained. The remaining life predicting device may obtain historical state data corresponding to each of the state parameters, and then divide the historical state data corresponding to each of the state parameters into a second preset number of groups of initial training data, where each group of initial training data includes historical state data corresponding to each of the state parameters in the preset time period. The second preset number is set according to practical experience, and the embodiment of the invention is not limited.
For example, the preset time period is 1 hour, the second preset number is 5000, and the remaining life predicting apparatus obtains 5000 sets of the initial training data from the respective historical state data corresponding to each of the state parameters of the mechanical device, where each set of the initial training data includes the respective historical state data corresponding to each of the state parameters within 1 hour.
S402, obtaining a second preset number of groups of residual life prediction back propagation neural network model training data according to each group of initial training data and the single-layer perceptron model;
specifically, after obtaining the second preset number of groups of initial training data, the remaining life prediction apparatus inputs each group of initial training data to the single-layer sensor model, so as to obtain an output result of the second preset number of groups of single-layer sensors, and obtains the second preset number of groups of remaining life prediction back propagation neural network model training data by using the output result of the second preset number of groups of single-layer sensors as training data of the remaining life prediction back propagation neural network model.
S403, training initial back propagation neural network model training data based on the second preset number of groups of back propagation neural network model training data for predicting the remaining life to obtain a back propagation neural network model for predicting the remaining life; wherein the initial back propagation neural network model includes a hidden layer.
Specifically, after obtaining the second preset number of groups of residual life prediction back propagation neural network model training data, the residual life prediction device inputs the second preset number of groups of residual life prediction back propagation neural network model training data into an initial residual life prediction back propagation neural network model one by one, trains the initial back propagation neural network model until the second preset number of training is completed or a global error obtained by calculation according to an output result of the initial back propagation neural network model and a preset expected result is smaller than an expected error, and obtains the residual life prediction back propagation neural network model. The specific training process of the initial back propagation neural network model is the prior art, and can be implemented by using Matlab or Python software, which is not described herein again; the expected error is set according to actual experience, and the embodiment of the invention is not limited.
The initial back propagation neural network model includes an input layer, a hidden layer, and an output layer. Taking the output result of the single-layer perceptron as the input of the input layer of the initial back propagation neural network model, wherein the number of the input layer nodes of the input layer of the initial back propagation neural network model is equal to the number of the output layer nodes of the output layer of the single-layer perceptron, and is equal to j, the number of the output layer nodes of the output layer of the initial back propagation neural network model is set to be k, and then the number of the hidden layer nodes of the hidden layer of the initial back propagation neural network model is set to be k
Figure BDA0001812386470000161
Wherein, alpha is (0, 10). Setting the input vector of the input layer of the initial back propagation neural network model as { y1，y2，y3，…yj}，yρRepresenting an input value of a rho-th input layer node of an input layer of the initial back propagation neural network model, rho being a positive integer and rho ≦ j, an expected output vector of an output layer of the initial back propagation neural network model being { l ≦ j1，l2，l3，…lk}，ltRepresenting an expected output value of a t-th output layer node of an output layer of the initial back propagation neural network model, t being a positive integer and t ≦ k, a net input vector of the output layer of the initial back propagation neural network model being { L ≦ k1，L2，L3，…Lk}，LtRepresenting the output of the initial back-propagation neural network modelNet input value of t-th output layer node of the layer, actual output vector of output layer of the initial back propagation neural network model is { c }1，c2，c3，…ck}，ctRepresenting the actual output value of the t-th output layer node of the output layer of the initial back propagation neural network model, the net input vector of the hidden layer being { s }1，s2，s3，…sβ}，spA net input value representing a p-th hidden layer node of the hidden layer, the net output vector of the hidden layer being { b }1，b2，b3，…bβ}，bpRepresenting a net output value of a pth hidden layer node of the hidden layer, where p ≦ β; setting the connection weight from the input layer of the initial back propagation neural network model to the hidden layer as w ═ wρpWhere ρ is a positive integer and ρ is not more than j, p is a positive integer and p is not more than β, and the connection weight of the hidden layer to the output layer of the initial back propagation neural network model is v ═ vptWherein t is a positive integer and t is less than or equal to k; setting the threshold value of each hidden layer node of the hidden layer as thetapThe threshold value corresponding to each output layer node of the output layer of the initial back propagation neural network model is γ ═ γt. The expected output value, the threshold of each hidden layer node of the hidden layer, and the threshold corresponding to each output layer node of the output layer of the initial back propagation neural network model are preset and are set according to actual experience, which is not limited in the embodiment of the invention. Wherein, wρp、vpt、γtAnd thetapThe initial value setting takes [ -1, 1]Random value of interval.
The net input value of the pth of the hidden layer node may be expressed as:
Figure BDA0001812386470000171
the net output value of the pth of the hidden layer node may be expressed as:
Figure BDA0001812386470000172
the net input value of the k output layer node of the output layer of the initial back propagation neural network model may be expressed as:
Figure BDA0001812386470000173
the actual output value of the kth output layer node of the output layer of the initial back propagation neural network model may be expressed as:
Figure BDA0001812386470000174
calculating a corrected error d for a k-th output layer node of an output layer of the initial back-propagation neural network modeltComprises the following steps:
dt＝(lt 2-ct 2)f′(Lt)
obtaining a global error E according to the expected output values of the output layer nodes of the output layer of the initial back propagation neural network model and the actual output values of the output layer nodes of the output layer of the initial back propagation neural network model as follows:
Figure BDA0001812386470000181
the connection weight adjustment amount from the hidden layer to the output layer of the initial back propagation neural network model is as follows:
Figure BDA0001812386470000182
wherein η is a learning rate, and is set according to actual experience, and the embodiment of the present invention is not limited.
The threshold adjustment amount corresponding to the tth output layer node of the output layer of the initial back propagation neural network model is as follows:
Δγt＝τdt
wherein τ is a learning rate, and is set according to practical experience, which is not limited in the embodiments of the present invention.
Adjusting the connection weight from the hidden layer to the output layer of the initial back propagation neural network model by adopting a momentum increasing method, wherein the connection weight adjustment amount from the hidden layer to the output layer of the initial back propagation neural network model is as follows:
Figure BDA0001812386470000183
wherein, λ (n) is momentum coefficient, n is positive integer.
Figure BDA0001812386470000184
Calculating a corrected error e for a pth hidden layer node of the hidden layerpComprises the following steps:
Figure BDA0001812386470000191
adjusting the threshold value corresponding to the pth hidden layer node of the hidden layer by the following amount:
Figure BDA0001812386470000192
wherein,
Figure BDA0001812386470000193
the learning rate is set according to actual experience, and the embodiment of the invention is not limited.
The method for predicting the remaining life of the mechanical equipment provided by the invention is illustrated by combining a specific embodiment. In this example, the method for predicting the remaining life of the mechanical equipment provided by the invention is used for predicting the remaining life of the wind driven generator, and the gear box, the stator winding, the rotor winding and the frequency converter of the wind driven generator are important components of the wind driven generator, so that the method has an important influence on the remaining life of the wind driven generator. The state parameters are set as: gearbox vibration frequency, rotor winding temperature, converter current, converter voltage, engine temperature, engine speed, engine vibration frequency, and gearbox lubrication oil temperature. Acquiring state data corresponding to the 8 state parameters within 1 hour of the wind driven generator, and then inputting the state data corresponding to the 8 state parameters into a single-layer sensor model of the wind driven generator to obtain an intermediate prediction result of the wind driven generator, wherein the single-layer sensor model of the wind driven generator is preset, and the number of output layer nodes of an output layer of the single-layer sensor model of the wind driven generator is set to be 6. And then inputting the intermediate prediction result of the wind driven generator into a residual life prediction back propagation neural network model of the wind driven generator, and outputting the residual life of the wind driven generator. The number of nodes of an input layer of the residual life prediction back propagation neural network model of the wind driven generator is 6, the number of nodes of an output layer of the residual life prediction back propagation neural network model of the wind driven generator is 12, a residual life interval corresponding to each output layer node is set to be {3 days or less, 3-5 days, 5-7 days, 7-10 days, 10-13 days, 13-15 days, 15-17 days, 17-20 days, 20-23 days, 23-25 days, 25-27 days and 27-30 days }, an implied layer of the residual life prediction back propagation neural network model of the wind driven generator is one, and the number of nodes of the implied layer is set to be 10 implied.
Fig. 5 is a schematic structural diagram of a remaining life predicting apparatus for a mechanical device according to an embodiment of the present invention, and as shown in fig. 5, the remaining life predicting apparatus for a mechanical device according to the present invention includes an obtaining unit 501, an obtaining unit 502, and a predicting unit 503, where:
the acquiring unit 501 is configured to acquire state data corresponding to each state parameter of the mechanical device in a preset time period; wherein the state parameter is preset; the obtaining unit 502 is configured to obtain an intermediate prediction result according to the state data and the single-layer sensor model corresponding to each state parameter; wherein the single-layer perceptron model is preset; the prediction unit 503 is configured to predict the remaining life of the mechanical device according to the intermediate prediction result and the remaining life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
Specifically, when the mechanical device is running, the sensor arranged on the mechanical device may be used to obtain the state data corresponding to each state parameter of the mechanical device in the preset time period, where the state parameter may be a state parameter of a main component that affects the running of the mechanical device, and may affect the remaining life of the mechanical device. The obtaining unit 501 may obtain state data corresponding to each state parameter of the mechanical device in the preset time period. The state parameters are preset and are set according to actual experience, and the embodiment of the invention is not limited; the preset time period is set according to actual needs, and the embodiment of the invention is not limited.
After obtaining the state data corresponding to each state parameter, the obtaining unit 502 uses the state data corresponding to each state parameter as the input of the single-layer perceptron, and may obtain the output result of the single-layer perceptron, where the output result is the intermediate prediction result. Wherein the single-layer perceptron model is preset.
After obtaining the intermediate prediction result, the prediction unit 503 may obtain an output result of the remaining life prediction back propagation neural network model by using the intermediate prediction result as an input of the remaining life prediction back propagation neural network model, where the output result of the remaining life prediction back propagation neural network model is the remaining life of the mechanical device, so as to realize the remaining life prediction of the mechanical device. Wherein the remaining life prediction back propagation neural network model is pre-established.
According to the residual life prediction device for the mechanical equipment, the state data corresponding to each state parameter of the mechanical equipment in the preset time period can be obtained, the intermediate prediction result is obtained according to the state data corresponding to each state parameter and the single-layer sensor model, then the back propagation neural network model is predicted according to the intermediate prediction result and the residual life, and the residual life of the mechanical equipment is predicted.
On the basis of the foregoing embodiments, further, the transfer function of the output layer of the single-layer sensor model is:
Figure BDA0001812386470000211
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure BDA0001812386470000212
wherein, CiThe connection weight of the ith input layer node of the input layer of the single-layer perceptron model, n is the number of the input layer nodes of the input layer of the single-layer perceptron model, HiIs obtained according to the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron modeljThe connection weight and the first threshold are preset as a first threshold of a jth output layer node of an output layer of the single-layer perceptron model.
Specifically, the single-layer perceptron is composed of two layers of neurons, an input layer and an output layer, wherein the input layer comprises a plurality of input layer nodes, and the output layer comprises at least one output layer node. In the embodiment of the present application, the state parameters correspond to the input layer nodes one to one, and the states are determined by the input layer nodesThe state data corresponding to the parameters are input into the single-layer perceptron from the corresponding input layer nodes, the state data input into the input layer nodes of the single-layer perceptron is the state data corresponding to the input layer nodes of the single-layer perceptron model, and the number of the output layer nodes is set according to actual experience. The transfer function of the output layer node is a Sigmoid function:
Figure BDA0001812386470000221
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure BDA0001812386470000222
wherein, CiThe connection weight of the ith input layer node of the input layer of the single-layer perceptron model is preset and is set according to actual experience, and the embodiment of the invention is not limited; hiObtained according to state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model, HiThe specific obtaining process is described in the following; n is the number of input layer nodes of the input layer of the single-layer perceptron model, thetajThe first threshold is a first threshold of a jth output layer node of an output layer of the single-layer perceptron model, and the first threshold is preset and set according to practical experience, and the embodiment of the invention is not limited. Wherein i and j are positive integers.
The embodiment of the apparatus provided in the present invention may be specifically configured to execute the processing flows of the above method embodiments, and the functions of the apparatus are not described herein again, and refer to the detailed description of the above method embodiments.
Fig. 6 is a schematic physical structure diagram of an electromechanical device according to an embodiment of the present invention, as shown in fig. 6, the electromechanical device includes a processor (processor)601, a memory (memory)602, and a communication bus 603;
the processor 601 and the memory 602 complete communication with each other through a communication bus 603;
processor 601 is configured to call program instructions in memory 602 to perform the methods provided by the above-described method embodiments, including, for example: acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset; obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset; predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
The present embodiment discloses a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform the method provided by the above-mentioned method embodiments, for example, comprising: acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset; obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset; predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
The present embodiments provide a non-transitory computer-readable storage medium storing computer instructions that cause the computer to perform the methods provided by the above method embodiments, for example, including: acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset; obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset; predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the remaining life prediction back propagation neural network model is pre-established.
In addition, the logic instructions in the memory may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer mechanical device (which may be a personal computer, an apparatus, or a network mechanical device) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer mechanical device (which may be a personal computer, a server, or a network mechanical device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (7)

1. A method for predicting the remaining life of a mechanical device, comprising:
acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset;
obtaining an intermediate prediction result according to the state data and the single-layer perceptron model corresponding to each state parameter; wherein the single-layer perceptron model is preset;
predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the residual life prediction back propagation neural network model is pre-established;
wherein the transfer function of the output layer of the single-layer perceptron model is as follows:
Figure FDA0002935485970000011
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure FDA0002935485970000012
wherein, CiThe connection weight of the ith input layer node of the input layer of the single-layer perceptron model, n is the number of the input layer nodes of the input layer of the single-layer perceptron model, HiIs obtained according to the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron modeljThe connection weight and the first threshold are preset, wherein the first threshold is a first threshold of a jth output layer node of an output layer of the single-layer perceptron model;
wherein, the HiThe obtaining of the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model comprises the following steps:
performing cluster analysis on state data corresponding to an ith input layer node of an input layer of the single-layer sensor model to obtain a first preset number of clusters, wherein the first preset number of clusters comprises m-1 normal state clusters and one abnormal state cluster; wherein m is the first preset number;
obtaining the average value of the number of the state data of the normal clusters according to the number of the state data of each normal cluster in m-1 normal clusters, and obtaining the result of the product of the average value of the number of the state data of the normal clusters and a preset value;
comparing the state data quantity of each normal state cluster of m-1 normal state clusters with the result of the product to obtain a normal state cluster of which the state data quantity of the normal state cluster is smaller than the result of the product;
obtaining the quantity of abnormal state data according to the quantity of the state data of the abnormal state cluster and the quantity of the state data of the normal state cluster, wherein the quantity of the state data of the normal state cluster is smaller than the quantity of the state data of the normal state cluster of the product result;
obtaining H according to the state data quantity and the abnormal state data quantity of the first cluster with the preset quantityi。
2. The method according to claim 1, wherein performing cluster analysis on the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model to obtain a first preset number of clusters, wherein the first preset number of clusters comprises m-1 normal state clusters and one abnormal state cluster, and comprises:
randomly selecting m-1 state data from state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model, respectively taking each state data in the m-1 state data as a clustering center of the normal state cluster, and calculating an average value of the clustering centers of the m-1 normal state clusters as an initial average value;
calculating the distance between each state data in the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model and the initial average value;
if the distance is judged to be larger than a second threshold value, the state data corresponding to the distance is classified into the abnormal state cluster;
if the distance is judged to be smaller than or equal to the second threshold value, the state data corresponding to the distance is classified into a normal state cluster corresponding to the state data with the shortest distance in the m-1 state data;
after the state data corresponding to the ith input layer node of the input layer of the single-layer sensor model are clustered once, recalculating the clustering centers of m-1 normal state clusters according to the state data of m-1 normal state clusters, and taking the average value of the recalculating cluster centers of m-1 normal state clusters as the current average value;
if the current average value is judged to be equal to the initial average value, outputting m-1 normal state clusters and one abnormal state cluster as the first preset number of clusters; otherwise, updating the initial average value to be the current average value, updating the clustering centers of the m-1 normal state clusters, and clustering again until the current average value obtained after clustering again is equal to the initial average value of clustering again.
3. The method of claim 1, whichCharacterized in that H is obtained according to the state data quantity and the abnormal state data quantity of the first cluster with the preset quantityiThe method comprises the following steps:
according to the formula
Figure FDA0002935485970000031
Calculating to obtain HiWherein Q isDifferent from each otherFor said number of abnormal state data, QGeneral assemblyThe number of state data for the first predetermined number of clusters.
4. The method of any one of claims 1 to 3, wherein the step of building the model of the residual life prediction back propagation neural network comprises:
acquiring historical state data corresponding to each state parameter, and acquiring a second preset number of groups of initial training data according to the historical state data corresponding to each state parameter, wherein each group of initial training data comprises the historical state data corresponding to each state parameter in the preset time period;
obtaining a second preset number of groups of residual life prediction back propagation neural network model training data according to each group of the initial training data and the single-layer perceptron model;
training initial back propagation neural network model training data based on the second preset number of groups of the residual life prediction back propagation neural network model training data to obtain the residual life prediction back propagation neural network model; wherein the initial back propagation neural network model includes a hidden layer.
5. A remaining life prediction apparatus for a mechanical device, comprising:
the acquisition unit is used for acquiring state data corresponding to each state parameter of the mechanical equipment in a preset time period; wherein the state parameter is preset;
the obtaining unit is used for obtaining an intermediate prediction result according to the state data and the single-layer sensor model corresponding to each state parameter; wherein the single-layer perceptron model is preset;
the prediction unit is used for predicting the residual life of the mechanical equipment according to the intermediate prediction result and the residual life prediction back propagation neural network model; wherein the residual life prediction back propagation neural network model is pre-established;
wherein the transfer function of the output layer of the single-layer perceptron model is as follows:
Figure FDA0002935485970000041
wherein, yjIs the output value of the j output layer node of the output layer of the single-layer perceptron model,
Figure FDA0002935485970000042
wherein, CiThe connection weight of the ith input layer node of the input layer of the single-layer perceptron model, n is the number of the input layer nodes of the input layer of the single-layer perceptron model, HiIs obtained according to the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron modeljThe connection weight and the first threshold are preset, wherein the first threshold is a first threshold of a jth output layer node of an output layer of the single-layer perceptron model;
wherein, the HiThe obtaining of the state data corresponding to the ith input layer node of the input layer of the single-layer perceptron model comprises the following steps:
performing cluster analysis on state data corresponding to an ith input layer node of an input layer of the single-layer sensor model to obtain a first preset number of clusters, wherein the first preset number of clusters comprises m-1 normal state clusters and one abnormal state cluster; wherein m is the first preset number;
obtaining the average value of the number of the state data of the normal clusters according to the number of the state data of each normal cluster in m-1 normal clusters, and obtaining the result of the product of the average value of the number of the state data of the normal clusters and a preset value;
comparing the state data quantity of each normal state cluster of m-1 normal state clusters with the result of the product to obtain a normal state cluster of which the state data quantity of the normal state cluster is smaller than the result of the product;
obtaining the quantity of abnormal state data according to the quantity of the state data of the abnormal state cluster and the quantity of the state data of the normal state cluster, wherein the quantity of the state data of the normal state cluster is smaller than the quantity of the state data of the normal state cluster of the product result;
obtaining H according to the state data quantity and the abnormal state data quantity of the first cluster with the preset quantityi。
6. An electromechanical device, comprising: a processor, a memory, and a communication bus, wherein:
the processor and the memory are communicated with each other through the communication bus;
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 4.
7. A non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the method of any one of claims 1 to 4.
CN201811125938.0A 2018-09-26 2018-09-26 Method and device for predicting residual life of mechanical equipment Active CN110415835B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811125938.0A CN110415835B (en) 2018-09-26 2018-09-26 Method and device for predicting residual life of mechanical equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811125938.0A CN110415835B (en) 2018-09-26 2018-09-26 Method and device for predicting residual life of mechanical equipment

Publications (2)

Publication Number Publication Date
CN110415835A CN110415835A (en) 2019-11-05
CN110415835B true CN110415835B (en) 2021-04-23

Family

ID=68358107

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811125938.0A Active CN110415835B (en) 2018-09-26 2018-09-26 Method and device for predicting residual life of mechanical equipment

Country Status (1)

Country Link
CN (1) CN110415835B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113127984A (en) * 2019-12-31 2021-07-16 中移(上海)信息通信科技有限公司 Method, device, equipment and storage medium for equipment maintenance
CN113919965A (en) * 2020-07-10 2022-01-11 上海电动工具研究所（集团）有限公司 Model training method and system, energy storage inverter and prediction method based on energy storage inverter
CN116882065A (en) * 2023-08-02 2023-10-13 广州汽车集团股份有限公司 Engine spark plug life prediction method and device, readable medium and electronic equipment

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1482453A (en) * 2003-07-11 2004-03-17 华东理工大学 Machine olfaction odor distinguishing method based on modularized composite neural net
CN102402727A (en) * 2011-11-10 2012-04-04 中联重科股份有限公司 System and method for predicting remaining life of component of construction machine
CN102737508A (en) * 2012-06-19 2012-10-17 银江股份有限公司 Urban road traffic state detection method combined with support vector machine (SVM) and back propagation (BP) neural network
CN102750824A (en) * 2012-06-19 2012-10-24 银江股份有限公司 Urban road traffic condition detection method based on voting of network sorter
CN104732411A (en) * 2015-03-27 2015-06-24 中国农业科学院农业信息研究所 Agricultural product consumption guide method based on BP neural network
CN106821337A (en) * 2017-04-13 2017-06-13 南京理工大学 A kind of sound of snoring source title method for having a supervision
CN106886660A (en) * 2017-03-23 2017-06-23 哈尔滨理工大学 EEMD Hilbert envelope spectrums are with Rolling Bearing Status recognition methods under the varying load that DBN is combined
CN108549769A (en) * 2018-04-13 2018-09-18 中国人民解放军火箭军工程大学 A kind of maintenance of equipment method for predicting residual useful life and system

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1482453A (en) * 2003-07-11 2004-03-17 华东理工大学 Machine olfaction odor distinguishing method based on modularized composite neural net
CN102402727A (en) * 2011-11-10 2012-04-04 中联重科股份有限公司 System and method for predicting remaining life of component of construction machine
CN102737508A (en) * 2012-06-19 2012-10-17 银江股份有限公司 Urban road traffic state detection method combined with support vector machine (SVM) and back propagation (BP) neural network
CN102750824A (en) * 2012-06-19 2012-10-24 银江股份有限公司 Urban road traffic condition detection method based on voting of network sorter
CN104732411A (en) * 2015-03-27 2015-06-24 中国农业科学院农业信息研究所 Agricultural product consumption guide method based on BP neural network
CN106886660A (en) * 2017-03-23 2017-06-23 哈尔滨理工大学 EEMD Hilbert envelope spectrums are with Rolling Bearing Status recognition methods under the varying load that DBN is combined
CN106821337A (en) * 2017-04-13 2017-06-13 南京理工大学 A kind of sound of snoring source title method for having a supervision
CN108549769A (en) * 2018-04-13 2018-09-18 中国人民解放军火箭军工程大学 A kind of maintenance of equipment method for predicting residual useful life and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Gear fault diagnosis based on support vector machine optimized by artificial bee colony algorithm;Yang Dalian 等;《Mechanism and Machine Theory》;20151231;第1-2页 *
智慧生产系统管理与优化学术研究进展;王坚;《中国机械工程学会机械自动化分会&中国自动化学会制造技术专委会学术工作进展报告2017》;20170801;第1-2页 *

Also Published As

Publication number Publication date
CN110415835A (en) 2019-11-05

Similar Documents

Publication Publication Date Title
CN110415835B (en) Method and device for predicting residual life of mechanical equipment
Sahu et al. Evaluating the impact of prediction techniques: Software reliability perspective.
CN110007652B (en) Hydroelectric generating set degradation trend interval prediction method and system
Zang et al. Long-term traffic speed prediction based on multiscale spatio-temporal feature learning network
CN111047085B (en) Hybrid vehicle working condition prediction method based on meta-learning
CN106448151B (en) Short-term traffic flow prediction method
US20170091675A1 (en) Production equipment including machine learning system and assembly and test unit
US20110066579A1 (en) Neural network system for time series data prediction
CN109001640B (en) Data processing method and device for power battery
CN108537327B (en) Neural network prediction method and device based on time series BP
CN113420509A (en) Wind turbine state evaluation method and device and storage medium
CN110794308B (en) Method and device for predicting train battery capacity
CN110309537B (en) Intelligent health prediction method and system for aircraft
CN110415036B (en) User grade determining method, device, computer equipment and storage medium
Vargas et al. Automobile spare-parts forecasting: A comparative study of time series methods
CN110334865B (en) Power equipment fault rate prediction method and system based on convolutional neural network
CN111523727B (en) Method for predicting remaining life of battery by considering recovery effect based on uncertain process
CN108761346A (en) A kind of vehicle lithium battery method for predicting residual useful life
CN110414748A (en) Photovoltaic power prediction technique
CN112099345A (en) Fuzzy tracking control method, system and medium based on input hysteresis
CN112149976B (en) Power grid accurate investment project decision method based on artificial intelligence
CN115902642A (en) Battery state of charge estimation method and device, electronic equipment and storage medium
CN115166534A (en) Method and system for predicting remaining service life of lithium ion battery
CN104318053B (en) Telephone traffic prediction method for calculating weighted moving average weight on the basis of univariate regression
CN111400964A (en) Fault occurrence time prediction method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant