CN110298497A - Manufacturing forecast maintenance system and its application method based on big data - Google Patents
Manufacturing forecast maintenance system and its application method based on big data Download PDFInfo
- Publication number
- CN110298497A CN110298497A CN201910501389.0A CN201910501389A CN110298497A CN 110298497 A CN110298497 A CN 110298497A CN 201910501389 A CN201910501389 A CN 201910501389A CN 110298497 A CN110298497 A CN 110298497A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- sample
- fault
- training
- 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.)
- Pending
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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a kind of manufacturing forecast maintenance system and its application method based on big data obtains several training samples according to the history data, the training sample includes fault sample and normal sample this method comprises: obtaining history data;Breadth and depth Wide And Deep model is trained based on several training samples, obtains Fault Model;The normal sample is inputted into the Fault Model, obtains several first output valves;According to several first output valves, fault threshold is determined;The current operating data of machinery to be detected is inputted into the Fault Model, obtains the second output valve;Detect whether second output valve is greater than the fault threshold;If second output valve is greater than the fault threshold, outputting alarm prompt.Through the invention, the malfunction of maintenance personal's discovering device can be guided in time, provide technical guarantee for the safe and highly efficient operation of equipment.
Description
Technical field
The present invention relates to technology for mechanical fault diagnosis fields, more particularly to the manufacturing forecast maintenance system based on big data and
Its application method.
Background technique
Petrochemical plant has the characteristics of long period, consecutive production, and pump and unit are even more key in petrochemical equipment
Equipment, once key equipment breaks down or fails and will lead to grave consequences in process units operation, not only equipment is impaired, whole
A process units can also stop comprehensively, or even will appear safety accident.Therefore for the safety and reliability of equipment operation
It is required that also higher and higher.
Existing breakdown maintenance or planned maintenance belong to preventative maintenance, are usually tieed up after machinery breaks down
It repairs, or machinery is repaired in the set time.If just being repaired after machinery breaks down, it may occur that improper fortune
Capable situation;If being repaired in the set time to machinery, and it is likely to occur and repairs excessive situation, the damage of meeting acceleration mechanical
Consumption.
Summary of the invention
The main purpose of the present invention is to provide a kind of manufacturing forecast maintenance system and its application method based on big data,
Aim to solve the problem that when unpredictable equipment is needing the technical issues of repairing in the prior art.
To achieve the above object, the present invention provides a kind of application method of manufacturing forecast maintenance system based on big data,
The application method of the manufacturing forecast maintenance system based on big data the following steps are included:
History data is obtained, several training samples, the training sample packet are obtained according to the history data
Include fault sample and normal sample;
Breadth and depth Wide And Deep model is trained based on several training samples, obtains failure inspection
Survey model;
The normal sample is inputted into the Fault Model, obtains several first output valves;
According to several first output valves, fault threshold is determined;
The current operating data of machinery to be detected is inputted into the Fault Model, obtains the second output valve;
Detect whether second output valve is greater than the fault threshold;
If second output valve is greater than the fault threshold, outputting alarm prompt.
Optionally, the acquisition history data obtains several training samples according to the history data, described
Training sample includes the steps that fault sample and normal sample includes:
Obtain history data, and the mechanical operation data mark for malfunction being in the history data
It is denoted as fault data, the mechanical operation data that normal condition is in the history data is labeled as normal data;
Using the fault data as fault sample, using the normal data as normal sample, and the failure is set
The label of sample is the first label, and the label that the normal sample is arranged is the second label.
Optionally, the history data includes:
Mechanical DCS data and/or Bentley system data.
Optionally, the breadth and depth Wide And Deep model includes deep neural network DNN model and linear
Model, it is described that breadth and depth Wide And Deep model is trained based on several training samples, obtain failure inspection
Survey model the step of include:
Individualized training sample is inputted into breadth and depth Wide And Deep model, it is corresponding to obtain individualized training sample
DNN model output value and linear model output valve;
It is superimposed the DNN model output value and linear model output valve, obtains superposition value, based on described in sigma function determination
The corresponding classification of superposition value, and judge whether classification label corresponding with the individualized training sample is consistent;
If consistent, it is denoted as prediction correctly, if inconsistent, is denoted as prediction error;
Using new individualized training sample as individualized training sample, and executes and individualized training sample is inputted into breadth and depth
Wide And Deep model obtains the step of the corresponding DNN model output value of individualized training sample and linear model output valve
Suddenly, until traversing all training samples;
Statistics obtains predicting correct number a, the number b of prediction error;
Pass through formula:
Training error value e is calculated;
Detect whether the training error value is less than or equal to preset error value;
If the training error value be less than or equal to preset error value, using current Wide And Deep model as
Fault Model;
If the training error value is greater than preset error value, DNN model and line are updated based on the training error value
The parameter of property model, obtains new Wide And Deep model;
Using the new Wide And Deep model as Wide And Deep model, and execute described by individualized training
Sample inputs breadth and depth Wide And Deep model, obtains the corresponding DNN model output value of individualized training sample and line
The step of property model output value.
Optionally, described according to several first output valves, the step of determining fault threshold, includes:
The maximum value in several first output valves is chosen, using the maximum value as fault threshold.
In addition, to achieve the above object, the present invention also provides a kind of manufacturing forecast maintenance system based on big data is described
Manufacturing forecast maintenance system based on big data includes: memory, processor and is stored on the memory and can be described
The computer program run on processor is realized when the computer program is executed by processor and is based on big data as described above
Manufacturing forecast maintenance system application method the step of.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
It is stored with computer program on storage medium, is realized when the computer program is executed by processor as described above based on big number
According to manufacturing forecast maintenance system application method the step of.
In the present invention, history data is obtained, several training samples, the instruction are obtained according to the history data
Practicing sample includes fault sample and normal sample;Based on several training samples to breadth and depth Wide And Deep
Model is trained, and obtains Fault Model;The normal sample is inputted into the Fault Model, obtains several first
Output valve;According to several first output valves, fault threshold is determined;Described in current operating data input by machinery to be detected
Fault Model obtains the second output valve;Detect whether second output valve is greater than the fault threshold;If described second
Output valve is greater than the fault threshold, then outputting alarm prompt.Through the invention, based on the magnanimity accumulated in enterprise production process
Historical data trains Wide And Deep model, obtains fault prediction model, then based on equipment by fault prediction model
Current operating data carries out failure predication, can guide the malfunction of maintenance personal's discovering device in time, is that the safety of equipment is high
Effect operation provides technical guarantee.
Detailed description of the invention
Fig. 1 is the manufacturing forecast maintenance system based on big data for the hardware running environment that the embodiment of the present invention is related to
Structural schematic diagram;
Fig. 2 is that the present invention is based on the signals of the process of the application method first embodiment of the manufacturing forecast maintenance system of big data
Figure.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The industry based on big data for the hardware running environment being related to as shown in FIG. 1, FIG. 1 is the embodiment of the present invention is pre-
Survey maintenance system structural schematic diagram.
As shown in Figure 1, should manufacturing forecast maintenance system based on big data may include: processor 1001, such as CPU,
Network interface 1004, user interface 1003, memory 1005, communication bus 1002.Wherein, communication bus 1002 is for realizing this
Connection communication between a little components.User interface 1003 may include display screen (Display), input unit such as keyboard
(Keyboard), optional user interface 1003 can also include standard wireline interface and wireless interface.Network interface 1004 is optional
May include standard wireline interface and wireless interface (such as WI-FI interface).Memory 1005 can be high speed RAM memory,
It is also possible to stable memory (non-volatile memory), such as magnetic disk storage.Memory 1005 optionally may be used also
To be independently of the storage device of aforementioned processor 1001.
It will be understood by those skilled in the art that the manufacturing forecast maintenance system structure shown in Fig. 1 based on big data is simultaneously
The restriction to the manufacturing forecast maintenance system based on big data is not constituted, may include components more more or fewer than diagram, or
Person combines certain components or different component layouts.
As shown in Figure 1, as may include that operating system, network are logical in a kind of memory 1005 of computer storage medium
Believe module, Subscriber Interface Module SIM and computer program.
In manufacturing forecast maintenance system based on big data shown in Fig. 1, after network interface 1004 is mainly used for connection
Platform server carries out data communication with background server;User interface 1003 is mainly used for connecting client (user terminal), with visitor
Family end carries out data communication;And processor 1001 can be used for calling the computer program stored in memory 1005, and execute
It operates below:
History data is obtained, several training samples, the training sample packet are obtained according to the history data
Include fault sample and normal sample;
Breadth and depth Wide And Deep model is trained based on several training samples, obtains failure inspection
Survey model;
The normal sample is inputted into the Fault Model, obtains several first output valves;
According to several first output valves, fault threshold is determined;
The current operating data of machinery to be detected is inputted into the Fault Model, obtains the second output valve;
Detect whether second output valve is greater than the fault threshold;
If second output valve is greater than the fault threshold, outputting alarm prompt.
Further, processor 1001 can call the computer program stored in memory 1005, also execute following behaviour
Make:
Obtain history data, and the mechanical operation data mark for malfunction being in the history data
It is denoted as fault data, the mechanical operation data that normal condition is in the history data is labeled as normal data;
Using the fault data as fault sample, using the normal data as normal sample, and the failure is set
The label of sample is the first label, and the label that the normal sample is arranged is the second label.
Further, processor 1001 can call the computer program stored in memory 1005, also execute following behaviour
Make:
The history data includes: mechanical DCS data and/or Bentley system data.
Further, processor 1001 can call the computer program stored in memory 1005, also execute following behaviour
Make:
Individualized training sample is inputted into breadth and depth Wide And Deep model, it is corresponding to obtain individualized training sample
DNN model output value and linear model output valve;
It is superimposed the DNN model output value and linear model output valve, obtains superposition value, based on described in sigma function determination
The corresponding classification of superposition value, and judge whether classification label corresponding with the individualized training sample is consistent;
If consistent, it is denoted as prediction correctly, if inconsistent, is denoted as prediction error;
Using new individualized training sample as individualized training sample, and executes and individualized training sample is inputted into breadth and depth
Wide And Deep model obtains the step of the corresponding DNN model output value of individualized training sample and linear model output valve
Suddenly, until traversing all training samples;
Statistics obtains predicting correct number a, the number b of prediction error;
Pass through formula:
Training error value e is calculated;
Detect whether the training error value is less than or equal to preset error value;
If the training error value be less than or equal to preset error value, using current Wide And Deep model as
Fault Model;
If the training error value is greater than preset error value, DNN model and line are updated based on the training error value
The parameter of property model, obtains new Wide And Deep model;
Using the new Wide And Deep model as Wide And Deep model, and execute described by individualized training
Sample inputs breadth and depth Wide And Deep model, obtains the corresponding DNN model output value of individualized training sample and line
The step of property model output value.
Further, processor 1001 can call the computer program stored in memory 1005, also execute following behaviour
Make:
The maximum value in several first output valves is chosen, using the maximum value as fault threshold.
It is that the present invention is based on the application method first embodiments of the manufacturing forecast maintenance system of big data referring to Fig. 2, Fig. 2
Flow diagram.
In one embodiment, the application method of the manufacturing forecast maintenance system based on big data includes:
Step S10 obtains history data, obtains several training samples, the instruction according to the history data
Practicing sample includes fault sample and normal sample;
In the present embodiment, operation data mechanical under each state accumulated in enterprise production process is obtained, is transported as history
Row data.
In an alternative embodiment, step S10 includes:
Obtain history data, and the mechanical operation data mark for malfunction being in the history data
It is denoted as fault data, the mechanical operation data that normal condition is in the history data is labeled as normal data;
Using the fault data as fault sample, using the normal data as normal sample, and the mark of the fault sample is set
Label are the first label, and the label that the normal sample is arranged is the second label.Wherein, history data includes: mechanical DCS
Data and/or Bentley system data.DCS data include temperature data, pressure data etc., Bentley system data, that is, equipment wave
Shape frequency spectrum data.
Include: in the present embodiment, in the history data of acquisition mechanical operation data in malfunction and
Mechanical operation data in normal condition.Mechanical operation data in malfunction is labeled as fault data, and
Using fault data as fault sample;Mechanical operation data in normal condition is labeled as normal data, and will be normal
Data are as normal sample.The first label, such as " 1 " are set by the label of fault sample, indicates the corresponding machinery of the sample
State is failure, sets the second label, such as " 0 " for the label of normal sample, indicates that the corresponding machine performance of the sample is
Normally.
Step S20 is trained breadth and depth Wide And Deep model based on several training samples, obtains
To Fault Model;
In the present embodiment, breadth and depth Wide And Deep model includes deep neural network DNN model and linear
Model, step S20 include:
Individualized training sample is inputted breadth and depth Wide And Deep model, obtains individualized training sample by step S201
This corresponding DNN model output value and linear model output valve;
In the present embodiment, it is assumed that training sample is training sample 1 to training sample 1000, then first that training sample 1 is defeated
Enter breadth and depth Wide And Deep model, obtains the corresponding DNN model output value X1 of training sample 1 and linear model
Output valve Y1.
Step S202 is superimposed the DNN model output value and linear model output valve, obtains superposition value, is based on sigma function
It determines the corresponding classification of the superposition value, and judges whether classification label corresponding with the individualized training sample is consistent;
In the present embodiment, X1 and Y1 is superimposed, superposition value Z is obtained, is then based on sigma function and determine that the superposition value is corresponding
Classification (failure or normal), then judge whether classification label corresponding with training sample 1 consistent.Wherein, sigma function activates
Function is used for binary classification.
Step S203, if unanimously, being denoted as prediction correctly, if inconsistent, being denoted as prediction error;
In the present embodiment, if training sample 1 is normal sample, label is the second label, and it is corresponding to represent the sample
Machine performance is normal, if the classification determined based on sigma function be it is normal, prediction is correct, if the classification that sigma function determines is event
Hinder, then prediction error.If training sample 1 is fault sample, label is the first label, represents the corresponding mechanical-like of the sample
State is failure, if the classification determined based on sigma function is normal, prediction error, if the classification that sigma function determines is failure, in advance
It surveys correct.
Step S204 using new individualized training sample as individualized training sample, and is executed and is inputted individualized training sample
Breadth and depth Wide And Deep model, obtains the corresponding DNN model output value of individualized training sample and linear model is defeated
The step of being worth out, until traversing all training samples;
In the present embodiment, after having executed above-mentioned steps S201 to step S203, using new individualized training sample as list
A training sample, repeat the above steps S201 to step S203, until traversing all training samples.
Step S205, statistics obtain predicting correct number a, the number b of prediction error;
In the present embodiment, S201 to step S204, can be obtained the corresponding prediction of all training samples through the above steps
Situation.
Step S206, passes through formula:
Training error value e is calculated;
In the present embodiment, by above-mentioned formula, the number of prediction error ratio shared in total frequency of training can be calculated
Example, such as total number of training are 1000, if prediction error number is 100, training error value e is 10%.
Step S207, detects whether the training error value is less than or equal to preset error value;
In the present embodiment, preset error value is judged according to actual needs, such as is set as 5%.
Step S208, if the training error value is less than or equal to preset error value, by current Wide And Deep
Model is as Fault Model;
In the present embodiment, if training error value is less than or equal to preset error value, illustrate current Wide And Deep
The precision of prediction of model is higher, meets actual demand, therefore, can be using current Wide And Deep model as fault detection
Model.
Step S209 updates DNN mould based on the training error value if the training error value is greater than preset error value
The parameter of type and linear model obtains new Wide And Deep model;
In the present embodiment, if training error value is greater than preset error value, illustrate that current Wide And Deep model is pre-
It is lower to survey precision, therefore, need to be updated, be obtained new based on parameter of the training error value to DNN model and linear model
Wide And Deep model.
Step S210, using the new Wide And Deep model as Wide And Deep model, and described in execution
Individualized training sample is inputted into breadth and depth Wide And Deep model, it is defeated to obtain the corresponding DNN model of individualized training sample
Value and the step of linear model output valve out.
It in the present embodiment, needs to be trained new Wide And Deep model, by new Wide And Deep mould
Type jumps as Wide And Deep model and executes step S201.
In the present embodiment, Wide and Deep deep learning model is used.It is linear model that wherein the end Wide is corresponding,
Input feature vector can be continuous feature, be also possible to sparse discrete features.It, can by L1 regularization in linear model training
It is converged in effective feature combination quickly.The corresponding end Deep is DNN model, the real number of the corresponding low-dimensional of each feature to
Amount, we term it the embedding of feature.DNN model adjusts the weight of hidden layer, and more new feature by backpropagation
Embedding.The output of the entire model of Wide and Deep is that linear model output is superimposed with what DNN model exported.Model
Using joint training, the training error of model, which can be fed back simultaneously in linear model and DNN model, carries out parameter more for training
Newly.Stand-alone training is carried out compared to model single in integration trainingt, the fusion of model only carries out in the stage that finally gives a forecast, joint
The fusion of model was carried out in the training stage in training, and the weight update of single model will receive the end Wide and the end Deep to mould
The joint effect of type training error.Therefore in the characteristic Design stage of model, the end Wide model and the end Deep model only need point
It is not absorbed in the aspect being good at, the end Wide model is remembered by the combined crosswise of discrete features, and the end Deep model passes through spy
The embedding progress of sign is extensive, and the size and complexity of model single in this way can also be controlled, and the performance of overall model
It remains to be improved.The model is developed on deep learning frame tensorflow, constructs the Wide& of a n-layer m node
Deep model, wherein Wide partially uses logistic regression, and Deep uses DNN network in part, using csv format as sample input format,
Model, which is saved, passes through tensorflow-serving using the GPU of P40 as hardware foundation at the pb format of tensorflow
Inference service is provided.
The normal sample is inputted the Fault Model, obtains several first output valves by step S30;
In the present embodiment, after obtaining Fault Model, by normal sample input fault detection model.Such as it will be normal
500 the first output valves can be obtained to 500 input fault detection model of normal sample in sample 1.
Step S40 determines fault threshold according to several first output valves;
It, can be using the average value of several first output valves as fault threshold in one alternative embodiment.
In another alternative embodiment, step S40 includes:
The maximum value in several first output valves is chosen, using the maximum value as fault threshold.
In the present embodiment, the maximum value in several first output valves is chosen, and using maximum value as fault threshold.
The current operating data of machinery to be detected is inputted the Fault Model, obtains the second output by step S50
Value;
In the present embodiment, the current DCS data of the current operating data of machinery to be detected, the i.e. machinery and/or Bentley
System data.DCS data include temperature data, pressure data etc., Bentley system data, that is, equipment waveform frequency spectrum data.It will be to
Mechanical current operating data input fault detection model is detected, the Fault Model can be obtained and be based on current operating data
Predicted value, i.e. the second output valve.
Step S60, detects whether second output valve is greater than the fault threshold;
Step S70, if second output valve is greater than the fault threshold, outputting alarm prompt.
In the present embodiment, if the second output valve is greater than fault threshold, and fault threshold is obtained based on several normal samples
The first output valve in maximum value, then illustrate the machinery that the current operating data of Current mechanical is not belonging to run under normal condition
Operation data, that is, illustrate it is to be detected it is mechanical there may be failure or will break down greatly very much, therefore, outputting alarm prompt,
So that maintenance personal in time overhauls the machinery.
In the present embodiment, history data is obtained, several training samples are obtained according to the history data, it is described
Training sample includes fault sample and normal sample;Based on several training samples to breadth and depth Wide And
Deep model is trained, and obtains Fault Model;The normal sample is inputted into the Fault Model, is obtained several
First output valve;According to several first output valves, fault threshold is determined;The current operating data of machinery to be detected is inputted
The Fault Model obtains the second output valve;Detect whether second output valve is greater than the fault threshold;If described
Second output valve is greater than the fault threshold, then outputting alarm prompt.Through this embodiment, based on being accumulated in enterprise production process
Mass historical data training Wide And Deep model, obtain fault prediction model, be then based on by fault prediction model
The current operating data of equipment carries out failure predication, can guide the malfunction of maintenance personal's discovering device in time, is equipment
Safe and highly efficient operation provides technical guarantee.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with computer program, when the computer program is executed by processor realize as above based on big data manufacturing forecast dimension
The step of repairing the application method each embodiment of system.
The specific embodiment of computer readable storage medium of the present invention and the above-mentioned manufacturing forecast based on big data repair and are
Each embodiment of the application method of system is essentially identical, and this will not be repeated here.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the system that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or system institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or system.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that terminal device (it can be mobile phone,
Computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (9)
1. a kind of application method of the manufacturing forecast maintenance system based on big data, which is characterized in that described based on big data
The application method of manufacturing forecast maintenance system the following steps are included:
History data is obtained, several training samples are obtained according to the history data, the training sample includes event
Hinder sample and normal sample;
Breadth and depth Wide And Deep model is trained based on several training samples, obtains fault detection mould
Type;
The normal sample is inputted into the Fault Model, obtains several first output valves;
According to several first output valves, fault threshold is determined;
The current operating data of machinery to be detected is inputted into the Fault Model, obtains the second output valve;
Detect whether second output valve is greater than the fault threshold;
If second output valve is greater than the fault threshold, outputting alarm prompt.
2. the application method of the manufacturing forecast maintenance system based on big data as described in claim 1, which is characterized in that described
History data is obtained, several training samples are obtained according to the history data, the training sample includes failure sample
The step of this and normal sample includes:
History data is obtained, and the mechanical operation data for being in malfunction in the history data is labeled as
The mechanical operation data that normal condition is in the history data is labeled as normal data by fault data;
Using the fault data as fault sample, using the normal data as normal sample, and the fault sample is set
Label be the first label, be arranged the normal sample label be the second label.
3. the application method of the manufacturing forecast maintenance system based on big data as claimed in claim 2, which is characterized in that described
History data includes:
Mechanical DCS data and/or Bentley system data.
4. the application method of the manufacturing forecast maintenance system based on big data as claimed in claim 2, which is characterized in that described
Breadth and depth Wide And Deep model includes deep neural network DNN model and linear model, if described based on described
The step of dry training sample is trained to breadth and depth Wide And Deep model, obtains Fault Model include:
Individualized training sample is inputted into breadth and depth Wide And Deep model, obtains the corresponding DNN mould of individualized training sample
Type output valve and linear model output valve;
It is superimposed the DNN model output value and linear model output valve, obtains superposition value, the superposition is determined based on sigma function
It is worth corresponding classification, and judges whether classification label corresponding with the individualized training sample is consistent;
If consistent, it is denoted as prediction correctly, if inconsistent, is denoted as prediction error;
Using new individualized training sample as individualized training sample, and executes and individualized training sample is inputted into breadth and depth Wide
And Deep model, the step of obtaining the corresponding DNN model output value of individualized training sample and linear model output valve, until
Traverse all training samples;
Statistics obtains predicting correct number a, the number b of prediction error;
Pass through formula:
Training error value e is calculated;
Detect whether the training error value is less than or equal to preset error value;
If the training error value is less than or equal to preset error value, using current Wide And Deep model as failure
Detection model;
If the training error value is greater than preset error value, DNN model and linear mould are updated based on the training error value
The parameter of type obtains new Wide And Deep model;
Using the new Wide And Deep model as Wide And Deep model, and execute described by individualized training sample
Breadth and depth Wide And Deep model is inputted, the corresponding DNN model output value of individualized training sample and linear mould are obtained
The step of type output valve.
5. the application method of the manufacturing forecast maintenance system based on big data as described in claim 1, which is characterized in that described
According to several first output valves, the step of determining fault threshold, includes:
The maximum value in several first output valves is chosen, using the maximum value as fault threshold.
6. a kind of manufacturing forecast maintenance system based on big data, which is characterized in that the manufacturing forecast dimension based on big data
The system of repairing includes: memory, processor and is stored in the computer journey that can be run on the memory and on the processor
Sequence, the computer program realize following steps when being executed by the processor:
History data is obtained, several training samples are obtained according to the history data, the training sample includes event
Hinder sample and normal sample;
Breadth and depth Wide And Deep model is trained based on several training samples, obtains fault detection mould
Type;
The normal sample is inputted into the Fault Model, obtains several first output valves;
According to several first output valves, fault threshold is determined;
The current operating data of machinery to be detected is inputted into the Fault Model, obtains the second output valve;
Detect whether second output valve is greater than the fault threshold;
If second output valve is greater than the fault threshold, outputting alarm prompt.
7. the manufacturing forecast maintenance system based on big data as claimed in claim 6, which is characterized in that the computer program
Following steps are also realized when being executed by the processor:
History data is obtained, and the mechanical operation data for being in malfunction in the history data is labeled as
The mechanical operation data that normal condition is in the history data is labeled as normal data by fault data;
Using the fault data as fault sample, using the normal data as normal sample, and the fault sample is set
Label be the first label, be arranged the normal sample label be the second label.
8. the manufacturing forecast maintenance system based on big data as claimed in claim 6, which is characterized in that the computer program
Following steps are also realized when being executed by the processor:
Individualized training sample is inputted into breadth and depth Wide And Deep model, obtains the corresponding DNN mould of individualized training sample
Type output valve and linear model output valve;
It is superimposed the DNN model output value and linear model output valve, obtains superposition value, the superposition is determined based on sigma function
It is worth corresponding classification, and judges whether classification label corresponding with the individualized training sample is consistent;
If consistent, it is denoted as prediction correctly, if inconsistent, is denoted as prediction error;
Using new individualized training sample as individualized training sample, and executes and individualized training sample is inputted into breadth and depth Wide
And Deep model, the step of obtaining the corresponding DNN model output value of individualized training sample and linear model output valve, until
Traverse all training samples;
Statistics obtains predicting correct number a, the number b of prediction error;
Pass through formula:
Training error value e is calculated;
Detect whether the training error value is less than or equal to preset error value;
If the training error value is less than or equal to preset error value, using current Wide And Deep model as failure
Detection model;
If the training error value is greater than preset error value, DNN model and linear mould are updated based on the training error value
The parameter of type obtains new Wide And Deep model;
Using the new Wide And Deep model as Wide And Deep model, and execute described by individualized training sample
Breadth and depth Wide And Deep model is inputted, the corresponding DNN model output value of individualized training sample and linear mould are obtained
The step of type output valve.
9. the manufacturing forecast maintenance system based on big data as claimed in claim 6, which is characterized in that the computer program
Making for the manufacturing forecast maintenance system based on big data as claimed in claim 3 or 5 is also realized when being executed by the processor
The step of with method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910501389.0A CN110298497A (en) | 2019-06-11 | 2019-06-11 | Manufacturing forecast maintenance system and its application method based on big data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910501389.0A CN110298497A (en) | 2019-06-11 | 2019-06-11 | Manufacturing forecast maintenance system and its application method based on big data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110298497A true CN110298497A (en) | 2019-10-01 |
Family
ID=68027850
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910501389.0A Pending CN110298497A (en) | 2019-06-11 | 2019-06-11 | Manufacturing forecast maintenance system and its application method based on big data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110298497A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110830515A (en) * | 2019-12-13 | 2020-02-21 | 支付宝(杭州)信息技术有限公司 | Flow detection method and device and electronic equipment |
CN110988563A (en) * | 2019-12-23 | 2020-04-10 | 厦门理工学院 | UPS (uninterrupted Power supply) fault detection method, device, equipment and storage medium |
CN111144639A (en) * | 2019-12-24 | 2020-05-12 | 国电南京自动化股份有限公司 | Subway equipment fault prediction method and system based on ALLN algorithm |
CN111651505A (en) * | 2020-06-05 | 2020-09-11 | 中国民用航空厦门空中交通管理站 | Data-driven equipment operation situation analysis and early warning method and system |
CN111708343A (en) * | 2019-10-31 | 2020-09-25 | 中国科学院沈阳自动化研究所 | Method for detecting abnormal behavior of field process behavior in manufacturing industry |
CN112163618A (en) * | 2020-09-27 | 2021-01-01 | 珠海格力电器股份有限公司 | Equipment fault detection method and detection system |
CN112465244A (en) * | 2020-12-04 | 2021-03-09 | 重庆忽米网络科技有限公司 | TensorFlow-based industrial equipment pre-inspection and pre-repair model training method and device |
CN113065679A (en) * | 2019-12-27 | 2021-07-02 | 北京国双科技有限公司 | Equipment maintenance performance monitoring method and device |
CN113255977A (en) * | 2021-05-13 | 2021-08-13 | 江西鑫铂瑞科技有限公司 | Intelligent factory production equipment fault prediction method and system based on industrial internet |
CN117272032A (en) * | 2023-11-22 | 2023-12-22 | 青岛埃恩斯信息技术科技有限公司 | Air compressor fault diagnosis method and device based on vibration detection |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108427658A (en) * | 2018-03-12 | 2018-08-21 | 北京奇艺世纪科技有限公司 | A kind of data predication method, device and electronic equipment |
CN108596645A (en) * | 2018-03-13 | 2018-09-28 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment of information recommendation |
CN109196527A (en) * | 2016-04-13 | 2019-01-11 | 谷歌有限责任公司 | Breadth and depth machine learning model |
CN109284866A (en) * | 2018-09-06 | 2019-01-29 | 安吉汽车物流股份有限公司 | Goods orders prediction technique and device, storage medium, terminal |
US20190050750A1 (en) * | 2017-08-11 | 2019-02-14 | Linkedln Corporation | Deep and wide machine learned model for job recommendation |
CN109747685A (en) * | 2019-01-15 | 2019-05-14 | 北京交大思诺科技股份有限公司 | Responder system fault pre-alarming platform |
-
2019
- 2019-06-11 CN CN201910501389.0A patent/CN110298497A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109196527A (en) * | 2016-04-13 | 2019-01-11 | 谷歌有限责任公司 | Breadth and depth machine learning model |
US20190050750A1 (en) * | 2017-08-11 | 2019-02-14 | Linkedln Corporation | Deep and wide machine learned model for job recommendation |
CN108427658A (en) * | 2018-03-12 | 2018-08-21 | 北京奇艺世纪科技有限公司 | A kind of data predication method, device and electronic equipment |
CN108596645A (en) * | 2018-03-13 | 2018-09-28 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment of information recommendation |
CN109284866A (en) * | 2018-09-06 | 2019-01-29 | 安吉汽车物流股份有限公司 | Goods orders prediction technique and device, storage medium, terminal |
CN109747685A (en) * | 2019-01-15 | 2019-05-14 | 北京交大思诺科技股份有限公司 | Responder system fault pre-alarming platform |
Non-Patent Citations (3)
Title |
---|
Z.B. ZHENG 等: "Wide and Deep Convolutional Neural Networks for Electricity-Theft Detection to Secure Smart Grids", 《IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS》 * |
张米露 等: "一种融合分形与BP网络PCA的海流机故障检测方法", 《电机与控制学报》 * |
杜江 等: "深广神经网络在变压器故障诊断中的应用", 《中北大学学报(自然科学版)》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111708343A (en) * | 2019-10-31 | 2020-09-25 | 中国科学院沈阳自动化研究所 | Method for detecting abnormal behavior of field process behavior in manufacturing industry |
CN111708343B (en) * | 2019-10-31 | 2021-11-09 | 中国科学院沈阳自动化研究所 | Method for detecting abnormal behavior of field process behavior in manufacturing industry |
CN110830515A (en) * | 2019-12-13 | 2020-02-21 | 支付宝(杭州)信息技术有限公司 | Flow detection method and device and electronic equipment |
CN110988563A (en) * | 2019-12-23 | 2020-04-10 | 厦门理工学院 | UPS (uninterrupted Power supply) fault detection method, device, equipment and storage medium |
CN110988563B (en) * | 2019-12-23 | 2022-04-01 | 厦门理工学院 | UPS (uninterrupted Power supply) fault detection method, device, equipment and storage medium |
CN111144639A (en) * | 2019-12-24 | 2020-05-12 | 国电南京自动化股份有限公司 | Subway equipment fault prediction method and system based on ALLN algorithm |
CN113065679A (en) * | 2019-12-27 | 2021-07-02 | 北京国双科技有限公司 | Equipment maintenance performance monitoring method and device |
CN111651505A (en) * | 2020-06-05 | 2020-09-11 | 中国民用航空厦门空中交通管理站 | Data-driven equipment operation situation analysis and early warning method and system |
CN111651505B (en) * | 2020-06-05 | 2023-05-16 | 中国民用航空厦门空中交通管理站 | Equipment operation situation analysis and early warning method and system based on data driving |
CN112163618A (en) * | 2020-09-27 | 2021-01-01 | 珠海格力电器股份有限公司 | Equipment fault detection method and detection system |
CN112465244A (en) * | 2020-12-04 | 2021-03-09 | 重庆忽米网络科技有限公司 | TensorFlow-based industrial equipment pre-inspection and pre-repair model training method and device |
CN113255977A (en) * | 2021-05-13 | 2021-08-13 | 江西鑫铂瑞科技有限公司 | Intelligent factory production equipment fault prediction method and system based on industrial internet |
CN117272032A (en) * | 2023-11-22 | 2023-12-22 | 青岛埃恩斯信息技术科技有限公司 | Air compressor fault diagnosis method and device based on vibration detection |
CN117272032B (en) * | 2023-11-22 | 2024-02-13 | 青岛埃恩斯信息技术科技有限公司 | Air compressor fault diagnosis method and device based on vibration detection |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110298497A (en) | Manufacturing forecast maintenance system and its application method based on big data | |
US10671060B2 (en) | Data-driven model construction for industrial asset decision boundary classification | |
US11159018B2 (en) | Method and system for online decision making of generator start-up | |
KR102574016B1 (en) | Methods and devices for condition classification of power network assets | |
Abbasghorbani et al. | Reliability‐centred maintenance for circuit breakers in transmission networks | |
Langeron et al. | A modeling framework for deteriorating control system and predictive maintenance of actuators | |
CN109120451A (en) | Equipment evaluation method, equipment and computer readable storage medium based on Internet of Things | |
CN110334948B (en) | Power equipment partial discharge severity evaluation method and system based on characteristic quantity prediction | |
CN107679576A (en) | The fault monitoring method and device of vehicle | |
Al‐Garni et al. | Artificial neural network application of modeling failure rate for Boeing 737 tires | |
CN113343581B (en) | Transformer fault diagnosis method based on graph Markov neural network | |
CN104462842A (en) | Excavating diagnosis method of failure data based on bayesian network | |
CN104598736A (en) | Roller bearing service life predicting model of self-adaptive multi-kernel combination relevance vector machine | |
CN110334865A (en) | A kind of electrical equipment fault rate prediction technique and system based on convolutional neural networks | |
US20170356349A1 (en) | System and method to enhance lean blowout monitoring | |
CN103309342A (en) | Safety verification scheme aiming at industrial control system | |
US9719881B2 (en) | Scalable framework for managing civil structure monitoring devices | |
CN102768641A (en) | Webpage testing factor selecting device and webpage testing factor selecting method | |
Ragab et al. | Pattern‐based prognostic methodology for condition‐based maintenance using selected and weighted survival curves | |
Taleb-Berrouane et al. | Dynamic RAMS analysis using advanced probabilistic approach | |
Kröger | Achieving resilience of large-scale engineered infrastructure systems | |
Shirmohammadi et al. | A computational model for determining the optimal preventive maintenance policy with random breakdowns and imperfect repairs | |
Kang et al. | Risk assessment of FPSO topside based on generalized Stochastic Petri Net | |
Seo et al. | A study on modeling using big data and deep learning method for failure diagnosis of system | |
CN110336280B (en) | Power system cascading failure analysis method based on dictionary set acceleration |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191001 |
|
RJ01 | Rejection of invention patent application after publication |