CN110415835A - A kind of method for predicting residual useful life and device of mechanical equipment - Google Patents
A kind of method for predicting residual useful life and device of mechanical equipment Download PDFInfo
- Publication number
- CN110415835A CN110415835A CN201811125938.0A CN201811125938A CN110415835A CN 110415835 A CN110415835 A CN 110415835A CN 201811125938 A CN201811125938 A CN 201811125938A CN 110415835 A CN110415835 A CN 110415835A
- Authority
- CN
- China
- Prior art keywords
- cluster
- status data
- layer
- single layer
- preset
- 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.)
- Granted
Links
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 And Monitoring For Control Systems (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The present invention discloses the method for predicting residual useful life and device of a kind of mechanical equipment.Wherein, which comprises obtain the corresponding status data of each state parameter of mechanical equipment in preset time period；Wherein, the state parameter is preset；According to the corresponding status data of each state parameter and single layer perceptron model, medium range forecast result is obtained；Wherein, the single layer perceptron model is preset；According to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, the remaining life of the mechanical equipment is predicted；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.Described device is for executing the above method.The method for predicting residual useful life and device of mechanical equipment provided by the invention improve the accuracy of mechanical equipment predicting residual useful life due to combining single layer perceptron model and back propagation artificial neural network model to predict the remaining life of mechanical equipment.
Description
Technical field
The present invention relates to mechanical equipment technical fields, and in particular to a kind of method for predicting residual useful life and dress of mechanical equipment
It sets.
Background technique
Mechanical equipment is made of all parts, mechanical equipment in During Process of Long-term Operation, component can gradually wear out and
Aging, remaining life can be gradually reduced, and mechanical equipment even safety accident out of service be eventually led to, without reasonably replacing
Component will cause waste, therefore, the remaining life of correctly predicted mechanical equipment, for ensureing mechanical equipment safe operation, improving
Economic benefit makes great sense.
In the prior art, it can be divided into three classes to the remaining life estimation method of mechanical equipment: surplus based on physical model
Remaining life estimation, the remaining life estimation of knowledge based model and the remaining life estimation based on data-driven model.Based on object
The remaining life estimation of reason model is normally based on the differential equation to describe the various working conditions of mechanical equipment, however machinery is set
Standby physical model is generally difficult to obtain, and the mechanical equipment remaining life estimation method based on physical model is without general
Property.In the remaining life estimation of knowledge based model, through frequently with expert system, however the knowledge model of expert system is used
Qualification and expert for expert have very high requirement for the Grasping level of the domain knowledge, and it is suitable to be difficult to find
Expert come guarantee mechanical equipment remaining life estimation accuracy.Remaining life estimation based on data-driven model is to be based on
The method of Principle of Statistics captures the information and knowledge implied in the data of magnanimity, often currently by acquiring the data of magnanimity
Data-driven model is established by artificial neural network or support vector machines, by establish neural network model or
Person's supporting vector machine model, then by corresponding algorithm, such as genetic algorithm, particle swarm algorithm or gradient descent algorithm, divide
The penalty coefficient of the other connection weight between layers to neural network, hidden layer node number and support vector machines carries out
Optimization, to obtain data-driven model.However since the data of acquisition often have the nonlinear characteristic of height, data-driven
Model is generally difficult to restrain, or is easily trapped into the situation of local optimum, and the prediction result of acquisition is not accurate enough.
Therefore, the method for predicting residual useful life for how proposing a kind of mechanical equipment can be improved the remaining longevity of mechanical equipment
The accuracy of life prediction becomes industry important topic urgently to be resolved.
Summary of the invention
For the defects in the prior art, the present invention provides the method for predicting residual useful life and device of a kind of mechanical equipment.
On the one hand, the present invention proposes a kind of method for predicting residual useful life of mechanical equipment, comprising:
Obtain the corresponding status data of each state parameter of mechanical equipment in preset time period；Wherein, the shape
State parameter is preset；
According to the corresponding status data of each state parameter and single layer perceptron model, medium range forecast knot is obtained
Fruit；Wherein, the single layer perceptron model is preset；
According to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, predict that the machinery is set
Standby remaining life；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
On the other hand, the present invention provides a kind of residual service life prediction device of mechanical equipment, comprising:
Acquiring unit, for obtaining the corresponding status number of each state parameter of mechanical equipment in preset time period
According to；Wherein, the state parameter is preset；
Obtaining unit, for according to the corresponding status data of each state parameter and single layer perceptron model,
Obtain medium range forecast result；Wherein, the single layer perceptron model is preset；
Predicting unit is used for according to the medium range forecast result and predicting residual useful life back propagation artificial neural network model,
Predict the remaining life of the mechanical equipment；Wherein, the predicting residual useful life back propagation artificial neural network model is to build in advance
Vertical.
In another aspect, the present invention provides a kind of electro mechanical devices, comprising: processor, memory and communication bus,
In:
The processor and the memory complete mutual communication by the communication bus；
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to refer to
Enable the method for predicting residual useful life that the mechanical equipment provided such as the various embodiments described above is provided.
Another aspect, the present invention provide a kind of non-transient computer readable storage medium, and the non-transient computer is readable
Storage medium stores computer instruction, and the computer instruction makes the computer execute the machinery provided such as the various embodiments described above
The method for predicting residual useful life of equipment.
The method for predicting residual useful life and device of mechanical equipment provided by the invention, due to that can obtain in preset time period
The corresponding status data of each state parameter of mechanical equipment, and according to the corresponding status data of each state parameter
With single layer perceptron model, medium range forecast is obtained as a result, then according to medium range forecast result and predicting residual useful life backpropagation
Neural network model predicts the remaining life of mechanical equipment, due to combining single layer perceptron model and reverse transmittance nerve network
Model predicts the remaining life of mechanical equipment, improves the accuracy of mechanical equipment predicting residual useful life.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that one embodiment of the invention provides；
Fig. 2 be another embodiment of the present invention provides mechanical equipment method for predicting residual useful life flow diagram；
Fig. 3 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that further embodiment of this invention provides；
Fig. 4 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that yet another embodiment of the invention provides；
Fig. 5 is the structural schematic diagram of the residual service life prediction device for the mechanical equipment that one embodiment of the invention provides；
Fig. 6 is the entity structure schematic diagram for the electro mechanical devices that one embodiment of the invention provides.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached in the embodiment of the present invention
Figure, technical solution in the embodiment of the present invention are explicitly described, it is clear that described embodiment is a part of the invention
Embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making wound
Every other embodiment obtained under the premise of the property made labour, shall fall within the protection scope of the present invention.
Fig. 1 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that one embodiment of the invention provides, such as Fig. 1
It is shown, the method for predicting residual useful life of mechanical equipment provided by the invention, comprising:
S101, the corresponding status data of each state parameter for obtaining mechanical equipment in preset time period；Wherein, institute
It is preset for stating state parameter；
Specifically, mechanical equipment at runtime, can obtain preset time by the sensor being arranged on mechanical equipment
The corresponding status data of each state parameter of mechanical equipment in section, the state parameter, which can be, influences equipment operation
The state parameter of main component can influence the remaining life of the mechanical equipment.The predicting residual useful life of mechanical equipment fills
Set (hereinafter referred to as residual service life prediction device) it is available in the preset time period mechanical equipment it is each described
The corresponding status data of state parameter.Wherein, the state parameter is preset, is configured based on practical experience, this
Inventive embodiments are without limitation；The preset time period is configured according to actual needs, and the embodiment of the present invention is without limitation.
For example, the mechanical equipment is wind-driven generator, each state parameter is respectively the lubricating oil temperature of gear-box
Degree, the vibration frequency of gear-box, the temperature of rotor windings, inverter current, frequency converter voltage, the temperature of engine, engine
Revolving speed and engine vibration frequency.
S102, according to the corresponding status data of each state parameter and single layer perceptron model, obtain intermediate
Prediction result；The single layer perceptron model is preset；
Specifically, the residual service life prediction device obtain each corresponding status data of state parameter it
Afterwards, using each corresponding status data of state parameter as the input of single layer perceptron, the single layer can be obtained
The output of perceptron is as a result, the output result is medium range forecast result.Wherein, the single layer perceptron model is default
's.
S103, according to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, described in prediction
The remaining life of mechanical equipment；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
Specifically, the residual service life prediction device is after obtaining the medium range forecast result, by the medium range forecast
As a result the input as predicting residual useful life back propagation artificial neural network model can obtain the predicting residual useful life and reversely pass
The output of neural network model is broadcast as a result, the output of the predicting residual useful life back propagation artificial neural network model is as a result, be
The remaining life of the mechanical equipment, to realize the predicting residual useful life to the mechanical equipment.Wherein, the remaining life
Prediction back propagation artificial neural network model pre-establishes.
For example, the residual service life prediction device obtains each corresponding historical state data of state parameter,
And according to each corresponding historical state data of state parameter, the second preset quantity group initial training data are obtained,
Initial training data described in every group include each state parameter corresponding historic state in the preset time period
Data；Then according to initial training data described in each group and the single layer perceptron model, the second preset quantity group is obtained
Predicting residual useful life back propagation artificial neural network model training data；It is pre- to be then based on the second preset quantity group remaining life
It surveys back propagation artificial neural network model training data to be trained initial back propagation artificial neural network model training data, obtain
Obtain the predicting residual useful life back propagation artificial neural network model；Wherein, the initial back propagation artificial neural network model packet
Include a hidden layer.
The method for predicting residual useful life of mechanical equipment provided by the invention is set due to that can obtain machinery in preset time period
The standby corresponding status data of each state parameter, and according to the corresponding status data of each state parameter and single layer
Perceptron model obtains medium range forecast as a result, then according to medium range forecast result and predicting residual useful life backpropagation neural network
Network model predicts the remaining life of mechanical equipment, due to combining single layer perceptron model and back propagation artificial neural network model pair
The remaining life of mechanical equipment is predicted, the accuracy of mechanical equipment predicting residual useful life is improved.
On the basis of the various embodiments described above, further, the transfer function of the output layer of the single layer perceptron model
Are as follows:
Wherein, yjFor the output valve of j-th of output layer output node layer of the single layer perceptron model,Wherein, CiFor the company of i-th of input layer of the input layer of the single layer perceptron model
Weight is connect, n is the input layer quantity of the input layer of the single layer perceptron model, HiIt is according to the single layer perceptron mould
The corresponding status data of i-th of input layer of the input layer of type obtains, θjFor the output of the single layer perceptron model
The first threshold of j-th of output node layer of layer, the connection weight and the first threshold are preset.
Specifically, the single layer perceptron is made of two layers of neuron of input layer and output layer, and the input layer includes more
A input layer node, the output layer include at least one output node layer.In the embodiment of the present application, the state parameter with
The input layer corresponds, and the corresponding status data of the state parameter is input to institute from corresponding input layer
It states in single layer perceptron, the status data for being input to the input layer of the single layer perceptron is the single layer perceptron model
The corresponding status data of input layer, the quantity of the output node layer is configured based on practical experience, and the present invention is real
Apply example without limitation.The transfer function of the output node layer is Sigmoid function:Wherein, yjIt is described
The output valve of j-th of output node layer of the output layer of single layer perceptron model,Wherein, CiFor
The connection weight of i-th of input layer of the input layer of the single layer perceptron model, the connection weight is preset, according to
Practical experience is configured, and the embodiment of the present invention is without limitation；HiIt is i-th according to the input layer of the single layer perceptron model
The corresponding status data of a input layer obtains, HiSpecific acquisition process see below it is described；N is single layer perception
The input layer quantity of the input layer of machine model, θjFor the .j output layer section of the output layer of the single layer perceptron model
Point first threshold, the first threshold be it is preset, be configured based on practical experience, the embodiment of the present invention is without limitation.
Wherein, i and j is positive integer.
Fig. 2 be another embodiment of the present invention provides mechanical equipment method for predicting residual useful life flow diagram, such as
Shown in Fig. 2, on the basis of the various embodiments described above, further, the HiIt is the input according to the single layer perceptron model
What the corresponding status data of i-th of input layer of layer obtained includes:
S201, the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model is carried out
Clustering obtains the first preset quantity cluster, and the first preset quantity cluster includes that m-1 normal condition cluster and one are different
Normal manifold of states；Wherein, m is first preset quantity；
Specifically, when the corresponding status data of each state parameter inputs the single layer perceptron, i-th defeated
Enter the corresponding status data of the corresponding state parameter of node layer input, inputs the status number of i-th of input layer
According to as the corresponding status data of i-th of input layer.The residual service life prediction device is to the single layer perceptron model
The corresponding status data of i-th of input layer of input layer carry out clustering, the first preset quantity cluster can be obtained,
The first preset quantity cluster includes m-1 normal condition cluster and an abnormality cluster；Wherein, first preset quantity
It is configured according to the actual situation, the embodiment of the present invention is without limitation.It will be appreciated that m is first preset quantity.
For example, i-th input layer section of the residual service life prediction device from the input layer of the single layer perceptron model
The m-1 status datas are randomly selected in the corresponding status data of point, respectively as in the cluster of m-1 normal condition cluster
The heart, and the average value of the m-1 status datas is calculated as initial average output value.Then, the single layer perceptron model is calculated
Input layer the corresponding status data of i-th of input layer between each status data and the initial average output value
Distance.Secondly, after judgement knows that the distance is more than or equal to second threshold, by described apart from corresponding status number
According to being included into the abnormality cluster；After judgement knows that the distance is less than the second threshold, by described apart from corresponding
Status data is included into normal condition cluster corresponding with the nearest status data of distance in the m-1 status datas.Again, right
After the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model once clusters completion, root
It is calculated according to the status data of the m-1 normal condition clusters and obtains current average.Then, know in judgement described current average
Value is equal to after the initial average output value, described in m-1 normal condition clusters of output and an abnormality cluster conduct
First preset quantity cluster；Otherwise, updating the initial average output value is the current average, and it is a described normal to update m-1
The cluster centre of manifold of states re-starts cluster, and the current average obtained until clustering again is equal to last poly-
The initial average output value of class.
S202, according to the status data quantity of each normal condition cluster in the m-1 normal condition cluster, obtain institute
The average value of the status data quantity of normal clusters is stated, and obtains the average value and preset value of the status data quantity of the normal clusters
Product result；
Specifically, the residual service life prediction device can count acquisition after obtaining the m-1 normal condition clusters
The quantity for the status data for including in each normal condition cluster in the m-1 normal condition clusters, the normal condition cluster
In include status data the i.e. described normal condition cluster of quantity status data quantity, then according to each normal condition cluster
Status data quantity calculates the average value for obtaining the status data quantity of the normal clusters, then by the status data of the normal clusters
The average value of quantity is multiplied with preset value, obtains the average value of the status data quantity of the normal clusters and the product of preset value
As a result.Wherein, the preset value is configured based on practical experience, and the embodiment of the present invention is without limitation.
For example, the residual service life prediction device obtains 7 normal condition clusters, wrapped in 7 normal condition clusters
The quantity of the status data included is respectively q1、q2、q3、q4、q5、q6And q7, the average value of the status data quantity of the normal clustersAssuming that the preset value is 30%, then described normal
The result of the product of the average value and preset value of the status data quantity of cluster
S203, the status data quantity of each of the m-1 normal condition clusters normal condition cluster is multiplied with described
Long-pending result is compared, obtain the normal condition cluster status data quantity be less than the product result normal condition
Cluster；
Specifically, the residual service life prediction device is in the average value that obtains the status data quantity of the normal clusters and pre-
If after the result of the product of value, by the status data quantity of each of the m-1 normal condition clusters normal condition cluster
It is compared with the result of the product, the status data quantity that can obtain the normal condition cluster is less than the knot of the product
The normal condition cluster of fruit.
For example, the result that the residual service life prediction device obtains the product is Q, wrapped in 7 normal condition clusters
The quantity of the status data included is respectively q1、q2、q3、q4、q5、q6And q7.The residual service life prediction device is by q1、q2、q3、q4、
q5、q6And q7It is compared respectively with Q, if q1Greater than Q, q2Less than Q, q3Greater than Q, q4Greater than Q, q5Greater than Q, q6Greater than Q, q7
Less than Q, then the status data quantity of the normal condition cluster is less than the normal condition cluster of the result of the product are as follows: q2It is corresponding
The normal clusters and q7The corresponding normal clusters.
It is S204, small according to the status data quantity of the abnormality cluster and the status data quantity of the normal condition cluster
In the status data quantity of the normal condition cluster of the result of the product, abnormality data bulk is obtained；
Specifically, the residual service life prediction device is less than described in the status data quantity for obtaining the normal condition cluster
After the normal condition cluster of the result of product, the status data quantity of each normal condition cluster is less than to the knot of the product
The status data quantity of the normal condition cluster of fruit is added with the status data quantity of the abnormality cluster obtains abnormality number
Data bulk.
For example, the residual service life prediction device obtains the status data quantity of two normal condition clusters less than described
The normal condition cluster of the result of product, corresponding status data quantity is respectively q2And q7, the status data of the abnormality cluster
Quantity is q ', the abnormality data bulk QIt is different=q2+q7+q′。
S205, it is obtained according to the status data quantity of the first preset quantity cluster and the abnormality data bulk
Hi。
Specifically, the residual service life prediction device, which can count, obtains in m cluster, a normal condition cluster of m-1
The status data quantity of status data quantity and the abnormality cluster, by the status data quantity of each normal condition cluster
And status data quantity of the sum of the status data quantity of abnormality cluster as the first preset quantity cluster, then use
The abnormality data bulk obtains H divided by the status data quantity of the first preset quantity clusteri。
For example, the residual service life prediction device obtains the quantity for the status data for including in 7 normal condition clusters
Respectively q1、q2、q3、q4、q5、q6And q7And the status data quantity of the abnormality cluster is q ', first present count
Measure the status data quantity Q of a clusterAlways=q1+q2+q3+q4+q5+q6+q7+ q ', the abnormality data bulk QIt is different, then
Fig. 3 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that further embodiment of this invention provides, such as
Shown in Fig. 3, on the basis of the various embodiments described above, further, the i-th of the input layer to the single layer perceptron model
The corresponding status data of a input layer carries out clustering, obtains the first preset quantity cluster, first preset quantity
A cluster includes m-1 normal condition cluster and an abnormality cluster includes:
S2011, from the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model with
Machine chooses m-1 status datas, using each status data in the m-1 status datas as described normal
The cluster centre of manifold of states, and calculate the m-1 normal condition cluster cluster centre average value as initial average output value；
Specifically, when the corresponding status data of each state parameter inputs the single layer perceptron, i-th defeated
Enter the corresponding status data of the corresponding state parameter of node layer input, inputs the status number of i-th of input layer
According to as the corresponding status data of i-th of input layer.The residual service life prediction device is from the single layer perceptron model
Input layer the corresponding status data of i-th of input layer in randomly select m-1 status datas, by m-1 institute
Cluster centre of each of status data status data as a normal condition cluster is stated, is obtained described in m-1
The cluster centre of normal condition cluster.In cluster of the residual service life prediction device thus to obtain the m-1 normal condition clusters
The heart, using the average value of the cluster centre of the m-1 normal condition clusters as initial average output value.
S2012, the input layer for calculating the single layer perceptron model the corresponding status data of i-th of input layer in
The distance between each status data and the initial average output value；
Specifically, the residual service life prediction device can calculate described in acquisition after obtaining the initial average output value
In the corresponding status data of i-th of input layer of the input layer of single layer perceptron model each status data with it is described
The distance between initial average output value.
If S2013, judgement know that the distance is greater than second threshold, institute is included into apart from corresponding status data by described
State abnormality cluster；
Specifically, the residual service life prediction device is calculating between the acquisition status data and the initial average output value
Distance after, the distance is compared with second threshold, if the distance be greater than the second threshold, by institute
It states and is included into the abnormality cluster apart from corresponding status data.Wherein, the second threshold is configured based on practical experience,
The embodiment of the present invention is without limitation.
If S2014, judgement know that the distance is less than or equal to the second threshold, by described apart from corresponding shape
State data are included into normal condition cluster corresponding with the nearest status data of distance in the m-1 status datas；
Specifically, the residual service life prediction device is calculating between the acquisition status data and the initial average output value
Distance after, the distance is compared with second threshold, if the distance be less than or equal to the second threshold,
So calculate it is described in the corresponding status data and m-1 status datas between each status data away from
From calculating the distance of the cluster centre apart from corresponding status data to each normal condition cluster, can obtain
M-1 distance is obtained, m-1 distance is compared, the minimum value in m-1 distance can be obtained, by described apart from corresponding status number
According to the normal condition cluster being included into the m-1 corresponding cluster centre of minimum value in, i.e., by described apart from corresponding state
Data are included into normal condition cluster corresponding with the nearest status data of distance in the m-1 status datas.
S2015, the corresponding status data one of i-th of input layer in the input layer to the single layer perceptron model
After secondary cluster is completed, the m-1 normal condition clusters are recalculated according to the status data of the m-1 normal condition clusters
Cluster centre, and the average value of the cluster centre for a normal condition cluster of m-1 that will recalculate acquisition is as current average
Value；
Specifically, i-th input layer of the residual service life prediction device to the input layer of the single layer perceptron model
Each status data of the corresponding status data of node carries out step S2013 or step S2014, i.e., to the single layer sense
Know that the corresponding status data of i-th of input layer of the input layer of machine model once clusters, it can be by the single layer perceptron
The corresponding status data of i-th of input layer of the input layer of model is included into the m-1 normal condition clusters and one respectively
The abnormality cluster.The m-1 normal condition clusters that the residual service life prediction device is obtained according to the primary cluster
Status data, the cluster centre for obtaining m-1 normal condition clusters, and the m- that acquisition will be recalculated can be recalculated
The average value of the cluster centre of 1 normal condition cluster is as current average.
It is obtained after once clustering for example, the residual service life prediction device randomly selects 7 status datas
Obtain 7 normal condition clusters.7 normal condition clusters that the residual service life prediction device can will be obtained according to cluster
All status datas recalculate obtain 7 normal condition clusters cluster centre, i.e., by a normal condition cluster
The summation of all status datas, then divided by the quantity of the status data of normal condition cluster described in this, so that it is described normal to obtain this
Then the cluster centre of manifold of states averages to the cluster centre for 7 normal condition clusters for recalculating acquisition, will be upper
Average value is stated as the current average.
If S2016, judgement know that the current average is equal to the initial average output value, it is a described normal to export m-1
Manifold of states and an abnormality cluster are as the first preset quantity cluster；Otherwise, updating the initial average output value is
The current average, and the cluster centre for updating the m-1 normal condition clusters re-starts cluster, until clustering it again
The current average obtained afterwards is equal to the initial average output value clustered again.
Specifically, the residual service life prediction device is after obtaining the current average, by the current average
It is compared with the initial average output value, if the current average is equal with the initial average output value, output m-1
The normal condition cluster and an abnormality cluster are as the first preset quantity cluster.If the current average
It is unequal with the initial average output value, then updating the initial average output value is the current average, i.e., with described current flat
Mean value substitutes the initial average output value as the initial average output value clustered again, and according to the m-1 of the acquisition normal conditions
The status data of cluster calculates separately the average value of the status data of each normal condition cluster as in the cluster clustered again
The heart repeats step S2012, S2013, S2014 and S2015, re-starts primary cluster, can regain current average.
The initial average output value that the residual service life prediction device is clustered by the current average for clustering acquisition again and again
It is compared, if the current average obtained after cluster again is not equal to the initial average output value clustered again,
The above-mentioned process clustered again is so repeated, then is clustered again, the current average etc. obtained until clustering again
In the initial average output value clustered again.It is clustered again if the current average obtained after cluster again is equal to
The initial average output value, the residual service life prediction device will cluster the m-1 normal condition clusters of acquisition and described again
Abnormality cluster is as the first preset quantity cluster.
On the basis of the various embodiments described above, further, the status number according to the first preset quantity cluster
Data bulk and the abnormality data bulk obtain HiInclude:
According to formulaIt calculates and obtains Hi, wherein QIt is differentFor the abnormality data bulk, QAlwaysIt is described
The status data quantity of one preset quantity cluster.
Specifically, the residual service life prediction device obtains the abnormality data bulk QIt is different, and obtain described first
The status data quantity Q of preset quantity clusterAlwaysIt later, can be according to formulaIt calculates and obtains Hi。
Fig. 4 is the flow diagram of the method for predicting residual useful life for the mechanical equipment that yet another embodiment of the invention provides, such as
Shown in Fig. 4, the step of establishing the predicting residual useful life back propagation artificial neural network model, includes:
S401, each corresponding historical state data of state parameter is obtained, and is joined according to each state
The corresponding historical state data of number obtains the second preset quantity group initial training data, initial training data packet described in every group
Include each corresponding historical state data of state parameter of the preset time period；
Specifically, in order to establish the predicting residual useful life back propagation artificial neural network model, enough trained numbers are needed
According to.By recording each corresponding status data of state parameter of the mechanical equipment, can obtain each described
The corresponding historical state data of state parameter.The available each state parameter of residual service life prediction device is each
Then it is pre- to be divided into second by self-corresponding historical state data for each corresponding historical state data of state parameter
If sets of numbers initial training data, initial training data described in every group include each state parameter in the preset time period
Interior corresponding historical state data.Wherein, second preset quantity is configured based on practical experience, and the present invention is implemented
Example is without limitation.
For example, the preset time period is 1 hour, second preset quantity is 5000, the predicting residual useful life dress
It sets from each corresponding historical state data of state parameter of the mechanical equipment, it is initial described in 5000 groups of acquisition
Training data, initial training data described in every group include each state parameter corresponding historic state in 1 hour
Data.
S402, the initial training data according to each group and the single layer perceptron model obtain second present count
Amount group predicting residual useful life back propagation artificial neural network model training data；
Specifically, the residual service life prediction device obtain initial training data described in the second preset quantity group it
Afterwards, initial training data described in each group are inputted into the single layer perceptron model respectively, the second preset quantity group institute can be obtained
The output of single layer perceptron is stated as a result, using the output result of single layer perceptron described in the second preset quantity group as the remaining longevity
The training data of life prediction back propagation artificial neural network model, it is anti-to obtain predicting residual useful life described in the second preset quantity group
To Propagation Neural Network model training data.
S403, it is based on the second preset quantity group predicting residual useful life back propagation artificial neural network model training data pair
Initial back propagation artificial neural network model training data is trained, and obtains the predicting residual useful life backpropagation neural network
Network model；Wherein, the initial back propagation artificial neural network model includes a hidden layer.
Specifically, the residual service life prediction device is reversely passed in acquisition the second preset quantity group predicting residual useful life
After broadcasting neural network model training data, one by one by the second preset quantity group predicting residual useful life backpropagation neural network
Network model training data are input in initial predicting residual useful life back propagation artificial neural network model, to described initial reversed
Propagation Neural Network model is trained, until completing second preset quantity time training or according to described initial reversed
The output result of Propagation Neural Network model and default expected result calculate the global error obtained and are less than anticipation error, obtain institute
State predicting residual useful life back propagation artificial neural network model.Wherein, the initial back propagation artificial neural network model is specific
Training process is the prior art, can use Matlab Python software realization, herein without repeating；The expectation misses
Difference is configured based on practical experience, and the embodiment of the present invention is without limitation.
The initial back propagation artificial neural network model includes an input layer, a hidden layer and an output layer.
Using the output result of the single layer perceptron as the input of the input layer of the initial back propagation artificial neural network model, institute
State the quantity of the input layer of the input layer of initial back propagation artificial neural network model and the output of the single layer perceptron
The quantity of the output node layer of layer is equal, is equal to j, the output layer of the initial back propagation artificial neural network model is arranged
The quantity for exporting node layer is k, then the hidden layer node of the hidden layer of the initial back propagation artificial neural network model
QuantityWherein, (0,10) α ∈.If the input layer of the initial back propagation artificial neural network model
Input vector be { y1, y2, y3... yj, yρIndicate ρ of the input layer of the initial back propagation artificial neural network model
The input value of input layer, ρ are positive integer and ρ≤j, the phase of the output layer of the initial back propagation artificial neural network model
Prestige output vector is { l1, l2, l3... lk, ltIndicate t-th of the output layer of the initial back propagation artificial neural network model
The desired output of node layer is exported, t is positive integer and t≤k, the output layer of the initial back propagation artificial neural network model
Net input vector be { L1, L2, L3... Lk, LtIndicate the t of the output layer of the initial back propagation artificial neural network model
The net input value of a output node layer, the reality output vector of the output layer of the initial back propagation artificial neural network model are
{c1, c2, c3... ck, ctIndicate t-th of output node layer of the output layer of the initial back propagation artificial neural network model
Real output value, the net input vector of the hidden layer are { s1, s2, s3... sβ, spIndicate p-th of the hidden layer it is implicit
The net input value of node layer, the net output vector of the hidden layer are { b1, b2, b3... bβ, bpIndicate the pth of the hidden layer
The net output valve of a hidden layer node, wherein p≤β；If the input layer of the initial back propagation artificial neural network model is to institute
The connection weight for stating hidden layer is w=wρp, wherein ρ is positive integer and ρ≤j, p are positive integer and p≤β, the hidden layer to institute
The connection weight for stating the output layer of initial back propagation artificial neural network model is v=vpt, wherein t is positive integer and t≤k；If
The threshold value of each hidden layer node of the hidden layer is θ=θp, the output of the initial back propagation artificial neural network model
The corresponding threshold value of each output node layer of layer is γ=γt.Wherein, the desired output, the hidden layer it is each hidden
Each output node layer of the output layer of threshold value and the initial back propagation artificial neural network model containing node layer is corresponding
Threshold value be it is preset, be configured based on practical experience, the embodiment of the present invention is without limitation.Wherein, wρp、vpt、γtAnd θpInitially
The random value for taking [- 1,1] section is arranged in value.
The net input value of p-th of the hidden layer hidden layer node can be indicated are as follows:
The net output value table of p-th of the hidden layer hidden layer node can be shown as:
It can be by the net input of k-th of output node layer of the output layer of the initial back propagation artificial neural network model
Value indicates are as follows:
It can be defeated by the reality of k-th of output node layer of the output layer of the initial back propagation artificial neural network model
Value indicates out are as follows:
Calculate the correction error of k-th of output node layer of the output layer of the initial back propagation artificial neural network model
dtAre as follows:
dt=(lt 2-ct 2)f′(Lt)
According to the desired output of each output node layer of the output layer of the initial back propagation artificial neural network model
The real output value of each output node layer of the output layer of value and the initial back propagation artificial neural network model obtains complete
Office's error E are as follows:
The hidden layer to the initial back propagation artificial neural network model output layer connection weight adjustment amount are as follows:
Wherein, η is learning rate, is configured based on practical experience, and the embodiment of the present invention is without limitation.
The corresponding adjusting thresholds of t-th of output node layer of the output layer of the initial back propagation artificial neural network model
Amount are as follows:
Δγt=τ dt
Wherein, τ is learning rate, is configured based on practical experience, and the embodiment of the present invention is without limitation.
Using the method for increase momentum to the hidden layer to the output of the initial back propagation artificial neural network model
The connection weight of layer is adjusted, the connection of the output layer of the hidden layer to the initial back propagation artificial neural network model
Weight adjustment amount are as follows:
Wherein, λ (n) is momentum coefficient, and n is positive integer.
Calculate the correction error e of p-th of hidden layer node of the hidden layerpAre as follows:
By the corresponding adjusting thresholds amount of p-th of the hidden layer hidden layer node are as follows:
Wherein,It for learning rate, is configured based on practical experience, the embodiment of the present invention is without limitation.
It is carried out below with reference to method for predicting residual useful life of the specific embodiment to mechanical equipment provided by the invention
It illustrates.In this example, using the method for predicting residual useful life of mechanical equipment provided by the invention to wind-driven generator
Remaining life is predicted, since the gear-box of the wind-driven generator, generator unit stator winding, rotor windings and frequency converter are
The important component of the wind-driven generator has a major impact the remaining life of wind-driven generator.The state parameter setting are as follows:
Gear-box vibration frequency, inverter current, frequency converter voltage, engine temperature, engine speed, is started at rotor windings temperature
Machine vibration frequency and gearbox lubrication oil temperature.It is respectively right to obtain above-mentioned 8 state parameters in the wind-driven generator nearly 1 hour
Then 8 corresponding status datas of state parameter are input to the list of the wind-driven generator by the status data answered
Layer perceptron model in, obtain the medium range forecast of the wind-driven generator as a result, the wind-driven generator single layer perceptron mould
Type be it is preset, the quantity of the output node layer of the output layer of the single layer perceptron model of the wind-driven generator is set as 6.
The medium range forecast result of the wind-driven generator is then input to the predicting residual useful life backpropagation of the wind-driven generator
In neural network model, the remaining life of the wind-driven generator is exported.Wherein, the predicting residual useful life of the wind-driven generator
The node of the input layer of the input layer of back propagation artificial neural network model is 6, the predicting residual useful life of the wind-driven generator
The output node layer of the output layer of back propagation artificial neural network model is set as 12, and above-mentioned each output node layer is corresponding surplus
Remaining service life section be respectively set to 3 days hereinafter, 3-5 days, 5-7 days, 7-10 days, 10-13 days, 13-15 days, 15-17 days, 17-20
It, 20-23 days, 23-25 days, 25-27 days, 27-30 days }, the predicting residual useful life Back propagation neural of the wind-driven generator
The hidden layer of network model is one, and the quantity of the hidden layer node of the hidden layer is set as 10.
Fig. 5 is the structural schematic diagram of the residual service life prediction device for the mechanical equipment that one embodiment of the invention provides, such as Fig. 5
Shown, the residual service life prediction device of mechanical equipment of the present invention includes acquiring unit 501, obtaining unit 502 and predicting unit
503, in which:
Acquiring unit 501 is used to obtain the corresponding status number of each state parameter of mechanical equipment in preset time period
According to；Wherein, the state parameter is preset；Obtaining unit 502 is used for according to each corresponding shape of state parameter
State data and single layer perceptron model obtain medium range forecast result；Wherein, the single layer perceptron model is preset；Prediction
Unit 503 is used to predict the machine according to the medium range forecast result and predicting residual useful life back propagation artificial neural network model
The remaining life of tool equipment；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
Specifically, mechanical equipment at runtime, can obtain preset time by the sensor being arranged on mechanical equipment
The corresponding status data of each state parameter of mechanical equipment in section, the state parameter, which can be, influences equipment operation
The state parameter of main component can influence the remaining life of the mechanical equipment.Acquiring unit 501 is available to described
Each corresponding status data of state parameter of the mechanical equipment in preset time period.Wherein, the state ginseng
Number be it is preset, be configured based on practical experience, the embodiment of the present invention is without limitation；The preset time period is according to practical need
It is configured, the embodiment of the present invention is without limitation.
Obtaining unit 502 is after obtaining each corresponding status data of state parameter, by each shape
Input of the corresponding status data of state parameter as single layer perceptron, can obtain the output knot of the single layer perceptron
Fruit, the output result is medium range forecast result.Wherein, the single layer perceptron model is preset.
After obtaining the medium range forecast result, predicting unit 503 is using the medium range forecast result as remaining life
The input for predicting back propagation artificial neural network model, can obtain the predicting residual useful life back propagation artificial neural network model
It exports as a result, the output of the predicting residual useful life back propagation artificial neural network model is as a result, remaining for the as described mechanical equipment
The remaining service life, to realize the predicting residual useful life to the mechanical equipment.Wherein, the predicting residual useful life Back propagation neural
Network model pre-establishes.
The residual service life prediction device of mechanical equipment provided by the invention is set due to that can obtain machinery in preset time period
The standby corresponding status data of each state parameter, and according to the corresponding status data of each state parameter and single layer
Perceptron model obtains medium range forecast as a result, then according to medium range forecast result and predicting residual useful life backpropagation neural network
Network model predicts the remaining life of mechanical equipment, due to combining single layer perceptron model and back propagation artificial neural network model pair
The remaining life of mechanical equipment is predicted, the accuracy of mechanical equipment predicting residual useful life is improved.
On the basis of the various embodiments described above, further, the transfer function of the output layer of the single layer perceptron model
Are as follows:
Wherein, yjFor the output valve of j-th of output layer output node layer of the single layer perceptron model,Wherein, CiFor the company of i-th of input layer of the input layer of the single layer perceptron model
Weight is connect, n is the input layer quantity of the input layer of the single layer perceptron model, HiIt is according to the single layer perceptron mould
The corresponding status data of i-th of input layer of the input layer of type obtains, θjFor the output of the single layer perceptron model
The first threshold of j-th of output node layer of layer, the connection weight and the first threshold are preset.
Specifically, the single layer perceptron is made of two layers of neuron of input layer and output layer, and the input layer includes more
A input layer node, the output layer include at least one output node layer.In the embodiment of the present application, the state parameter with
The input layer corresponds, and the corresponding status data of the state parameter is input to institute from corresponding input layer
It states in single layer perceptron, the status data for being input to the input layer of the single layer perceptron is the single layer perceptron model
The corresponding status data of input layer, the quantity of the output node layer is configured based on practical experience, and the present invention is real
Apply example without limitation.The transfer function of the output node layer is Sigmoid function:Wherein, yjIt is described
The output valve of j-th of output node layer of the output layer of single layer perceptron model,Wherein, CiFor
The connection weight of i-th of input layer of the input layer of the single layer perceptron model, the connection weight is preset, according to
Practical experience is configured, and the embodiment of the present invention is without limitation；HiIt is i-th according to the input layer of the single layer perceptron model
The corresponding status data of a input layer obtains, HiSpecific acquisition process see below it is described；N is single layer perception
The input layer quantity of the input layer of machine model, θjFor j-th of output layer section of the output layer of the single layer perceptron model
Point first threshold, the first threshold be it is preset, be configured based on practical experience, the embodiment of the present invention is without limitation.
Wherein, i and j is positive integer.
The embodiment of device provided by the invention specifically can be used for executing the process flow of above-mentioned each method embodiment,
Details are not described herein for function, is referred to the detailed description of above method embodiment.
Fig. 6 is the entity structure schematic diagram for the electro mechanical devices that one embodiment of the invention provides, as shown in fig. 6, described
Electro mechanical devices include processor (processor) 601, memory (memory) 602 and communication bus 603；
Wherein, processor 601, memory 602 complete mutual communication by communication bus 603；
Processor 601 is used to call the program instruction in memory 602, to execute provided by above-mentioned each method embodiment
Method, for example, obtain the corresponding status data of each state parameter of mechanical equipment in preset time period；Wherein,
The state parameter is preset；According to the corresponding status data of each state parameter and single layer perceptron model,
Obtain medium range forecast result；Wherein, the single layer perceptron model is preset；According to the medium range forecast result and remaining longevity
Life prediction back propagation artificial neural network model, predicts the remaining life of the mechanical equipment；Wherein, the predicting residual useful life is anti-
It is pre-established to Propagation Neural Network model.
The present embodiment discloses a kind of computer program product, and the computer program product includes being stored in non-transient calculating
Computer program on machine readable storage medium storing program for executing, the computer program include program instruction, when described program instruction is calculated
When machine executes, computer is able to carry out method provided by above-mentioned each method embodiment, for example, obtains in preset time period
The corresponding status data of each state parameter of mechanical equipment；Wherein, the state parameter is preset；According to each institute
The corresponding status data of state parameter and single layer perceptron model are stated, medium range forecast result is obtained；Wherein, the single layer sense
Know that machine model is preset；According to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, prediction
The remaining life of the mechanical equipment；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
The present embodiment provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage medium
Computer instruction is stored, the computer instruction makes the computer execute method provided by above-mentioned each method embodiment, example
It such as include: the corresponding status data of each state parameter for obtaining mechanical equipment in preset time period；Wherein, the state
Parameter is preset；According to the corresponding status data of each state parameter and single layer perceptron model, obtain intermediate
Prediction result；Wherein, the single layer perceptron model is preset；It is anti-according to the medium range forecast result and predicting residual useful life
To Propagation Neural Network model, the remaining life of the mechanical equipment is predicted；Wherein, the predicting residual useful life backpropagation mind
It is pre-established through network model.
In addition, the logical order in above-mentioned memory can be realized and as independence by way of SFU software functional unit
Product when selling or using, can store in a computer readable storage medium.Based on this understanding, of the invention
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer mechanical equipment (can be personal computer, device or network mechanical equipment etc.) executes each reality of the present invention
Apply all or part of the steps of the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory
(ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk
Etc. the various media that can store program code.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer mechanical equipment (can be personal computer, server or network mechanical equipment etc.) executes
Method described in certain parts of each embodiment or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of method for predicting residual useful life of mechanical equipment characterized by comprising
Obtain the corresponding status data of each state parameter of mechanical equipment in preset time period；Wherein, the state ginseng
Number is preset；
According to the corresponding status data of each state parameter and single layer perceptron model, medium range forecast result is obtained；
Wherein, the single layer perceptron model is preset；
According to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, the mechanical equipment is predicted
Remaining life；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
2. the method according to claim 1, wherein the transfer function of the output layer of the single layer perceptron model
Are as follows:
Wherein, yjFor the output valve of j-th of output layer output node layer of the single layer perceptron model,Wherein, CiFor the company of i-th of input layer of the input layer of the single layer perceptron model
Weight is connect, n is the input layer quantity of the input layer of the single layer perceptron model, HiIt is according to the single layer perceptron mould
The corresponding status data of i-th of input layer of the input layer of type obtains, θjFor the output of the single layer perceptron model
The first threshold of j-th of output node layer of layer, the connection weight and the first threshold are preset.
3. according to the method described in claim 2, it is characterized in that, the HiIt is the input according to the single layer perceptron model
What the corresponding status data of i-th of input layer of layer obtained includes:
Clustering is carried out to the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model,
The first preset quantity cluster is obtained, the first preset quantity cluster includes m-1 normal condition cluster and an abnormality cluster；
Wherein, m is first preset quantity；
According to the status data quantity of each normal condition cluster in the m-1 normal condition clusters, the normal clusters are obtained
Status data quantity average value, and obtain the average value of the status data quantity of the normal clusters and the product of preset value
As a result；
By the result of the status data quantity of each of the m-1 normal condition cluster normal condition cluster and the product into
Row compare, obtain the normal condition cluster status data quantity be less than the product result normal condition cluster；
It is less than described multiply according to the status data quantity of the abnormality cluster and the status data quantity of the normal condition cluster
The status data quantity of the normal condition cluster of long-pending result obtains abnormality data bulk；
H is obtained according to the status data quantity of the first preset quantity cluster and the abnormality data bulki。
4. according to the method described in claim 3, it is characterized in that, the of the input layer to the single layer perceptron model
The corresponding status data of i input layer carries out clustering, obtains the first preset quantity cluster, first preset quantity
A cluster includes m-1 normal condition cluster and an abnormality cluster includes:
M-1 is randomly selected from the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model
A status data, using each status data in m-1 status datas as the normal condition cluster
Cluster centre, and calculate the m-1 normal condition cluster cluster centre average value as initial average output value；
Calculate each shape in the corresponding status data of i-th of input layer of the input layer of the single layer perceptron model
The distance between state data and the initial average output value；
If judgement knows that the distance is greater than second threshold, the abnormality is included into apart from corresponding status data by described
Cluster；
If judgement knows that the distance is less than or equal to the second threshold, it is included into described apart from corresponding status data
Normal condition cluster corresponding with the nearest status data of distance in the m-1 status datas；
Completion is once clustered in the corresponding status data of i-th of input layer of the input layer to the single layer perceptron model
Later, the cluster centre of the m-1 normal condition clusters is recalculated according to the status data of the m-1 normal condition clusters,
And the average value of the cluster centre for a normal condition cluster of m-1 that will recalculate acquisition is as current average；
If judgement knows that the current average is equal to the initial average output value, the m-1 normal condition clusters and one are exported
A abnormality cluster is as the first preset quantity cluster；Otherwise, it is described current flat for updating the initial average output value
Mean value, and the cluster centre for updating the m-1 normal condition clusters re-starts cluster, the institute obtained until clustering again
It states current average and is equal to the initial average output value clustered again.
5. according to the method described in claim 3, it is characterized in that, the status number according to the first preset quantity cluster
Data bulk and the abnormality data bulk obtain HiInclude:
According to formulaIt calculates and obtains Hi, wherein QIt is differentFor the abnormality data bulk, QAlwaysIt is default for described first
The status data quantity of quantity cluster.
6. method according to any one of claims 1 to 5, which is characterized in that establish the predicting residual useful life and reversely pass
The step of broadcasting neural network model include:
Each corresponding historical state data of the state parameter is obtained, and is respectively corresponded to according to each state parameter
Historical state data, obtain the second preset quantity group initial training data, initial training data described in every group include each institute
State state parameter corresponding historical state data in the preset time period；
The initial training data according to each group and the single layer perceptron model obtain the second preset quantity group remaining longevity
Life prediction back propagation artificial neural network model training data；
Based on the second preset quantity group predicting residual useful life back propagation artificial neural network model training data to initial anti-
It is trained to Propagation Neural Network model training data, obtains the predicting residual useful life back propagation artificial neural network model；
Wherein, the initial back propagation artificial neural network model includes a hidden layer.
7. a kind of residual service life prediction device of mechanical equipment characterized by comprising
Acquiring unit, for obtaining the corresponding status data of each state parameter of mechanical equipment in preset time period；Its
In, the state parameter is preset；
Obtaining unit, for obtaining according to the corresponding status data of each state parameter and single layer perceptron model
Medium range forecast result；Wherein, the single layer perceptron model is preset；
Predicting unit, for according to the medium range forecast result and predicting residual useful life back propagation artificial neural network model, prediction
The remaining life of the mechanical equipment；Wherein, the predicting residual useful life back propagation artificial neural network model pre-establishes.
8. device according to claim 7, which is characterized in that the transfer function of the output layer of the single layer perceptron model
Are as follows:
Wherein, yjFor the output valve of j-th of output layer output node layer of the single layer perceptron model,Wherein, CiFor the company of i-th of input layer of the input layer of the single layer perceptron model
Weight is connect, n is the input layer quantity of the input layer of the single layer perceptron model, HiIt is according to the single layer perceptron mould
The corresponding status data of i-th of input layer of the input layer of type obtains, θjFor the output of the single layer perceptron model
The first threshold of j-th of output node layer of layer, the connection weight and the first threshold are preset.
9. a kind of electro mechanical devices characterized by comprising processor, memory and communication bus, in which:
The processor and the memory complete mutual communication by the communication bus；
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to instruct energy
It is enough to execute such as method as claimed in any one of claims 1 to 6.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute such as method as claimed in any one of claims 1 to 6.
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 true CN110415835A (en) | 2019-11-05 |
| CN110415835B 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) |
Cited By (2)
| 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 |
| 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)
| 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 residual life of part of engineering 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 |
-
2018
- 2018-09-26 CN CN201811125938.0A patent/CN110415835B/en active Active
Patent Citations (8)
| 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 residual life of part of engineering 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)
| Title |
|---|
| YANG DALIAN 等: "Gear fault diagnosis based on support vector machine optimized by artificial bee colony algorithm", 《MECHANISM AND MACHINE THEORY》 * |
| 王坚: "智慧生产系统管理与优化学术研究进展", 《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》 * |
Cited By (2)
| 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110415835B (en) | 2021-04-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110110930B (en) | Recurrent neural network short-term power load prediction method for improving whale algorithm | |
| CN112117760A (en) | Micro-grid energy scheduling method based on double-Q-value network deep reinforcement learning | |
| CN108427986A (en) | A kind of production line electrical fault prediction technique and device | |
| CN109001640B (en) | Data processing method and device for power battery | |
| CN111047085B (en) | Hybrid vehicle working condition prediction method based on meta-learning | |
| CN109961098B (en) | Training data selection method for machine learning | |
| US9934470B2 (en) | Production equipment including machine learning system and assembly and test unit | |
| US20150134581A1 (en) | Method for training an artificial neural network | |
| CN104217258B (en) | A kind of electric load sigma-t Forecasting Methodology | |
| CN105989441A (en) | Model parameter adjustment method and device | |
| CN109324291A (en) | A kind of prediction technique for Proton Exchange Membrane Fuel Cells life prediction | |
| CN110415835A (en) | A kind of method for predicting residual useful life and device of mechanical equipment | |
| CN106779253A (en) | The term load forecasting for distribution and device of a kind of meter and photovoltaic | |
| CN110059802A (en) | For training the method, apparatus of learning model and calculating equipment | |
| CN112381359A (en) | Multi-critic reinforcement learning power economy scheduling method based on data mining | |
| CN115293052A (en) | Power system active power flow online optimization control method, storage medium and device | |
| CN111813524B (en) | Task execution method and device, electronic equipment and storage medium | |
| Razmi et al. | Neural network based on a genetic algorithm for power system loading margin estimation | |
| Khorasani | A new heuristic approach for unit commitment problem using particle swarm optimization | |
| CN116865358A (en) | Multi-time long-scale power system wind power waste and load fluctuation tracking method and equipment | |
| CN112527547A (en) | Mechanical intelligent fault prediction method based on automatic convolution neural network | |
| CN111400964A (en) | Fault occurrence time prediction method and device | |
| CN111799820A (en) | Double-layer intelligent hybrid zero-star cloud energy storage countermeasure regulation and control method for power system | |
| CN110378464A (en) | The management method and device of the configuration parameter of artificial intelligence platform | |
| CN114648178B (en) | Operation and maintenance strategy optimization method of electric energy metering device based on DDPG algorithm |
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 |