CN108090608A - A kind of gantry crane trend prediction method based on BP neural network - Google Patents

A kind of gantry crane trend prediction method based on BP neural network Download PDF

Info

Publication number
CN108090608A
CN108090608A CN201711328788.9A CN201711328788A CN108090608A CN 108090608 A CN108090608 A CN 108090608A CN 201711328788 A CN201711328788 A CN 201711328788A CN 108090608 A CN108090608 A CN 108090608A
Authority
CN
China
Prior art keywords
mrow
data
msub
gantry crane
layer
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
Application number
CN201711328788.9A
Other languages
Chinese (zh)
Inventor
唐刚
杨辉
黄婉娟
顾邦平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Maritime University
Original Assignee
Shanghai Maritime University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Maritime University filed Critical Shanghai Maritime University
Priority to CN201711328788.9A priority Critical patent/CN108090608A/en
Publication of CN108090608A publication Critical patent/CN108090608A/en
Pending legal-status Critical Current

Links

Classifications

    • 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
    • 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
    • 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
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/083—Shipping

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Computational Linguistics (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

A kind of gantry crane trend prediction method based on BP neural network, it is characterised in that：Pass through the acceleration transducer in gantry crane and strain transducer gathered data first, the load cluster centre of the vibration data collected is thirdly determined using the method for K averages, then the data of cluster centre point are classified, finally the gantry crane status data collected is analyzed using BP neural network algorithm and predicts the data of its future state, realizes the purpose of opposite bank bridge status predication.By the state of data analysis gantry crane, show whether gantry crane needs the conclusion maintained or repaired whithin a period of time.Method proposed by the present invention is that the Practical Project data of gantry crane are analyzed, and goes prediction gantry crane future running state data using BP neural network algorithm on this basis, and prediction result is accurately and reliably.

Description

A kind of gantry crane trend prediction method based on BP neural network
Technical field
The present invention relates to harbour machinery fields, are specifically a kind of gantry crane trend prediction method based on BP neural network.
Background technology
Due to global trade rapid development, Container Transport is fast-developing, the utilization rate of gantry crane in port logistics transport It is higher and higher.It can ensure the normal operation of gantry crane, directly affect the work efficiency and economic benefit at harbour.Therefore it is more and more Harbour the machine performance of gantry crane is detected and assessed, sensor is installed by the key position in gantry crane, is acquired big The data message of amount is simultaneously stored in database.The various information of these data messages under cover equipment running status, but by In the amount of data is very huge, disorderly and unsystematic and individual data is present with burst, processing has very with analyzing these mass datas Big challenge.
The conclusion for carrying out finishing analysis by monitoring obtained data in real time to gantry crane and drawing gantry crane state, is gantry crane Working condition provides very big reference value.Therefore the data of bridge carry out analysis and evaluation and prediction on the opposite bank, can not only sentence The service life of disconnected gantry crane and service condition, and work efficiency of the gantry crane in port traffic can be improved, it is of great significance.
The method of gantry crane data processing is mainly included at this stage：Self-organizing feature map (self-organizing Feature map, SOFM) neural network algorithm classification, K average gantry cranes data classification.These methods are only by data Processing is classified, there is no predicting data, the state during gantry crane works in future can not be predicted, it is impossible to well Suitably care and maintenance is done to gantry crane, to reduce the risk of gantry crane at work.
It is using the thought of BP neural network：It, can be accurate using the forward-propagating of neutral net error-duration model and signal The data for predicting gantry crane following a period of time, whithin a period of time, have good judgement to the state of gantry crane.
Due to importance of the gantry crane in port logistics transport, predicting the working condition of gantry crane has highly important meaning Justice using the method for BP neural network prediction data, the extension data of gantry crane vibration can be predicted according to available data, are analyzed Extension data can accurately judge the state of gantry crane, and appropriate maintenance is carried out to gantry crane, ensure that gantry crane normally works, Reduce gantry crane existing hidden danger at work simultaneously.
The content of the invention
The present invention provides a kind of bank based on BP neural network for deficiency existing for existing gantry crane status assessment technology Bridge trend prediction method can be realized and gantry crane data are accurately predicted, and the data for passing through prediction judge the state of gantry crane, Show whether gantry crane needs the conclusion maintained or repaired whithin a period of time.
In order to achieve the above object, the present invention proposes a kind of gantry crane trend prediction method based on BP neural network, in fact Applying step is：
Step 1：Obtain the data of a gantry crane measuring point whithin a period of time in the database, which is using installing Sensor on gantry crane bridge motor extracted real-time vibration severity data at interval of 10 to 20 seconds, and stores in the database, institute The vibration severity data stated are used for assessing gantry crane loaded-up condition；Acquired data V={ v1,v2,…,vl,…,vL, wherein, L Represent the number of the test sample, vlIt represents to obtain the l articles vibration data in data；
Step 2：The mode of the data of acquisition temporally decile is divided into K group data, and every group of data K- average is gathered The mode of class, calculates 5 clustering clusters, and one of which cluster centre data are P={ P1,P2,P3,P4,P5, it will be per in group cluster Heart point falls into 5 types from small to large according to data, and forms new data group with tag along sort data Q={ 1,2,3,4,5 }, owns It is X={ x that data group, which forms new sample data,1,x2,…,xs,…,xS, wherein, tag along sort data 1,2,3,4,5 are distinguished Represent five kinds of gantry crane zero load, underloading, middle load, heavy duty, super-heavy load states；S is the number of sample data, and S=5K, K are just whole Number, xsRepresent the s articles characteristic in sample data；
Step 3：Arrangement classification is carried out to the time-based sequencing of sample data and different classifications label, obtains data Group is D={ d1,d2,…,do,…,dK}；Wherein, doFor the vibration data of 5 kinds of states within o-th of period, and there is do= {do1,do2,do3,do4,do5, do1Represent the data of light condition, do2Represent the data of light condition, do3State is carried in expression Data, do4Represent the data of heavy condition, do5Represent the data of super-heavy load state；
By data group D={ d1,d2,…,do,…,dKIt is divided into two class of test sample data and training sample data, wherein The vibration data of the test sample is Y={ y1,y2,…,ym,…,yM, wherein M represents the number of the test sample, ymTable Show the m datas in the test sample；The vibration data of training sample is Z={ z1,z2,…,zn,…,zN, wherein N tables Show the number of the training sample, znRepresent the nth bar data in the training sample；
Step 4：Test sample data and training sample data are normalized, generation normalization characteristic data Y ' ={ y1', y2' ..., ym' ... yM' and Z '={ z1', z2' ..., zn' ... zN', wherein, ym' expression the normalization characteristic M datas in data Y ', znNth bar data in the ' expression normalization characteristic data Z '；
Normalizing calculation is：
Wherein：Xmax, XminThe respectively maximum and minimum value of actual sample data, X* are the value after normalization, and X is real Actual value；
After handling data using BP neural network, renormalization calculating is carried out to data, just obtains actual data value；Instead Normalizing equation is：
X=(Xmax-Xmin)×X*+Xmin
Step 5：Establish BP neural network model；The topological structure of the BP neural network model includes input layer, implies Layer and output layer；Wherein, the neuron number of input layer is f, and the neuron number of hidden layer is g, and the neuron of output layer is a Number is h；Arbitrary input layer is fi, i ∈ (1,2 ... p)；Arbitrary hidden layer neuron is gj, j ∈ (1,2 ... q)；It is arbitrary defeated Go out layer neuron for hk, k ∈ (1,2 ... r)；
Step 6：The basic parameter of BP neural network model is initialized, including：Learning rate μ, input layer to hidden layer Weight wij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to output layer biasing number bkAnd excitation function R (x)；Wherein, input layer is to the weight w of hidden layerij, hidden layer to output layer weight wjk, input layer To the biasing number a of hidden layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Input layer is to the weight w of hidden layerijMeaning is：Arbitrary input layer fiTo arbitrary hidden layer Neuron gjBetween weight；Hidden layer is to the weight w of output layerjkMeaning is：Arbitrary hidden layer neuron gjTo arbitrary Output layer neuron hkBetween weight；Input layer is to the biasing number a of hidden layerjMeaning is：Each input layer is to arbitrarily Hidden layer neuron gjBiasing number；Hidden layer is to the biasing number b of output layerkMeaning is：Each hidden layer neuron is to arbitrary defeated Go out a layer neuron hkBiasing number；
Step 7：Using MATLAB softwares, by training sample data Z={ z1,z2,…,zn,…,zNImport BP nerve nets Network model is trained BP neural network model；
Step 8：In MATLAB softwares, the test data set Y={ y that will obtain in step 31,y2,…,ym,…,yM} It imports among BP neural network model, predicts the vibration data group T={ t at the point of following U period1,t2,…, tu,…,tU}.Wherein, tuFor 5 vibration datas within u-th of period, and there is tu={ tu1,tu2,tu3,tu4,tu5, tu1 Represent the data of light condition, tu2Represent the data of light condition, tu3The data of state, t are carried in expressionu4Represent heavy condition Data, tu5Represent the data of super-heavy load state；
Further, step 7：BP neural network model is trained, is comprised the steps of：
Step 7.1：Input layer includes three neurons, and hidden layer neuron g is calculated using the following formulajOutput valve：
Output layer neuron h is calculated using the following formula againkOutput valve：
Define error function：
Wherein：ykFor the desired output of output layer neuron, initial value is given history training sample data；E is Deviation；
Step 7.2：The output layer neuron h being calculatedkOutput valve substitute into loss function, deviation E is calculated； Whether judgment bias value E meets the requirements, if meeting the requirements, goes to step 7.9；If being unsatisfactory for requiring, 7.3 are gone to step；
Step 7.3：Use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(c+1)=(1- γ) gjek+γΔwjkc
Wherein：
Wherein：γ be weights inertia coeffeicent, ecAnd ec-1Respectively c and c-1 training error；Δwjk(c)For the c times instruction Hidden layer neuron g when practicingjTo output layer neuron hkWeight adjustment amount；Δwjk(c+1)For the c+1 times it is trained when hidden layer god Through first gjTo output layer neuron hkWeight adjustment amount；
Step 7.4：Use again following formula calculate input layer to hidden layer weight adjustment amount for：
Wherein：Δwij(c)For the c times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；Δ wij(c+1)For the c+1 times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；
Step 7.5：Biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 7.6：Biasing number a is calculated using following formulajUpdated value：
Step 7.7：Therefore, using the weight adjustment amount of the hidden layer that step 7.3 is calculated to output layer, step 7.4 The biasing number b that the input layer being calculated is calculated to the weight adjustment amount of hidden layer, step 7.5kUpdated value and step The corresponding ginseng for the BP neural network model that once training obtains before the updated value of the rapid 7.6 biasing number a being calculated is optimized and revised Number, thus obtains updated BP neural network model；
Step 7.8：Based on the updated BP neural network model that step 7.7 obtains, return to step 7.1,
Step 7.9：BP neural network model after being trained；
State following at the measuring point may determine that according to the vibration data of prediction gained, shaken according to the gantry crane measuring point of prediction The size of dynamic data, can predict whether gantry crane can work normally in different states under loads, if vibration data changes It is larger, then it needs to carry out gantry crane repair in advance；If vibration data variation is little, daily dimension can be carried out to gantry crane Shield.
It is provided by the invention that the classification Forecasting Methodology of gantry crane data is had the following advantages based on BP neural network：
1st, the Practical Project data that the method for the present invention is worked based on gantry crane, can more embody the working condition of gantry crane, obtain As a result it is more accurate.
2nd, the present invention is based on BP neural network algorithm, and BP neural network algorithm has good robustness, self-organizing certainly The characteristics of adaptability, ensure that data prediction has the characteristics that good redundancy and accuracy.
3rd, the present invention has carried out accurate prediction to the data of gantry crane, preferably judges the working condition of gantry crane, so as to Be conducive to utilization of the gantry crane on the direction of harbour, ensure production work safety.
Description of the drawings
Fig. 1 is the overall flow figure of the present invention.
Fig. 2 is sensor scheme of installation in gantry crane.
Fig. 3 is the flow chart of BP neural network model training.
Specific embodiment
The attached drawing in the embodiment of the present invention will be combined first below, the technical solution in the embodiment of the present invention is carried out clear Chu is fully described by；Then, technical scheme is introduced by a specific case history.Obviously, described reality It is only part of the embodiment of the present invention to apply example, and instead of all the embodiments, based on the embodiments of the present invention, this field is general Logical technical staff all other embodiments obtained without making creative work belong to what the present invention protected Scope.
A kind of method based on BP neural network algorithm opposite bank bridge status predication：
Fig. 1 is the overall flow figure of the present invention.Implementation steps according to the flow diagram present invention are first to obtain gantry crane Data are being carried out clustering processing and the cluster centre of data are being classified, are then being normalized by data；Initialization nerve The basic parameter of network model, and train neural network model using training sample；After neural network model is obtained, it will test Sample is substituted among model, obtains prediction data；The state of gantry crane can accurately be predicted.
Step 1：Obtain the data of a gantry crane measuring point whithin a period of time in the database, which is using installing Sensor on gantry crane bridge motor extracted real-time vibration severity data at interval of 10 to 20 seconds, and stores in the database, institute The vibration severity data stated are used for assessing gantry crane loaded-up condition；Acquired data V={ v1,v2,…,vl,…,vL, wherein, L Represent the number of the test sample, vlIt represents to obtain the l articles vibration data in data；
Such as Fig. 2, sensor is mounted on position on gantry crane crossbeam and track.The measurement of these measuring points is shaking for crossbeam or track Dynamic earthquake intensity.Wherein, Z1H represents the vibration of gantry crane A-frame top large vehicle walking direction；Z1V represents gantry crane A-frame top Vertical Square To vibration；Z1A represents that gantry crane represents the horizontal direction vibration of gantry crane A-frame top；Z2H represents gantry crane crossbeam top large vehicle walking Direction vibrates；Z2V represents gantry crane crossbeam top vertical vibration；Z2A represents the horizontal direction vibration of gantry crane crossbeam top；Z3V Represent vertical vibration at the front tension bar of gantry crane crossbeam left side；Z4V represents vertical vibration at the front tension bar of gantry crane crossbeam right side； Z5V represents vertical vibration at the main hinge in gantry crane crossbeam left side；Z6V represents vertical vibration at the main hinge in gantry crane crossbeam right side； Z7H represents the vibration of gantry crane sea side crossbeam large vehicle walking direction；Z8H represents the vibration of gantry crane trackside crossbeam large vehicle walking direction；
Step 2：The mode of the data of acquisition temporally decile is divided into K group data, and every group of data K- average is gathered The mode of class, calculates 5 clustering clusters, and one of which cluster centre data are P={ P1,P2,P3,P4,P5, it will be per in group cluster Heart point falls into 5 types from small to large according to data, and forms new data group with tag along sort data Q={ 1,2,3,4,5 }, owns It is X={ x that data group, which forms new sample data,1,x2,…,xs,…,xS, wherein, tag along sort data 1,2,3,4,5 are distinguished Represent five kinds of gantry crane zero load, underloading, middle load, heavy duty, super-heavy load states；S is the number of sample data, and S=5K, K are just whole Number, xsRepresent the s articles characteristic in sample data；
Step 3：Arrangement classification is carried out to the time-based sequencing of sample data and different classifications label, obtains data Group is D={ d1,d2,…,do,…,dK}；Wherein, doFor the vibration data of 5 kinds of states within o-th of period, and there is do= {do1,do2,do3,do4,do5, do1Represent the data of light condition, do2Represent the data of light condition, do3State is carried in expression Data, do4Represent the data of heavy condition, do5Represent the data of super-heavy load state；
By data group D={ d1,d2,…,do,…,dKIt is divided into two class of test sample data and training sample data, wherein The vibration data of the test sample is Y={ y1,y2,…,ym,…,yM, wherein M represents the number of the test sample, ymTable Show the m datas in the test sample；The vibration data of training sample is Z={ z1,z2,…,zn,…,zN, wherein N tables Show the number of the training sample, znRepresent the nth bar data in the training sample；
Step 4：Test sample data and training sample data are normalized, generation normalization characteristic data Y ' ={ y1', y2' ..., ym' ... yM' and Z '={ z1', z2' ..., zn' ... zN', wherein, ym' expression the normalization characteristic M datas in data Y ', znNth bar data in the ' expression normalization characteristic data Z '；
Normalizing calculation is：
Wherein：Xmax, XminThe respectively maximum and minimum value of actual sample data, X* are the value after normalization, and X is real Actual value；
After handling data using BP neural network, renormalization calculating is carried out to data, just obtains actual data value；Instead Normalizing equation is：
X=(Xmax-Xmin)×X*+Xmin
Step 5：Establish BP neural network model；The topological structure of the BP neural network model includes input layer, implies Layer and output layer；Wherein, the neuron number of input layer is f, and the neuron number of hidden layer is g, and the neuron of output layer is a Number is h；Arbitrary input layer is fi, i ∈ (1,2 ... p)；Arbitrary hidden layer neuron is gj, j ∈ (1,2 ... q)；It is arbitrary defeated Go out layer neuron for hk, k ∈ (1,2 ... r)；
Step 6：The basic parameter of BP neural network model is initialized, including：Learning rate μ, input layer to hidden layer Weight wij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to output layer biasing number bkAnd excitation function R (x)；Wherein, input layer is to the weight w of hidden layerij, hidden layer to output layer weight wjk, input layer To the biasing number a of hidden layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Input layer is to the weight w of hidden layerijMeaning is：Arbitrary input layer fiTo arbitrary hidden layer Neuron gjBetween weight；Hidden layer is to the weight w of output layerjkMeaning is：Arbitrary hidden layer neuron gjTo arbitrary Output layer neuron hkBetween weight；Input layer is to the biasing number a of hidden layerjMeaning is：Each input layer is to arbitrarily Hidden layer neuron gjBiasing number；Hidden layer is to the biasing number b of output layerkMeaning is：Each hidden layer neuron is to arbitrary defeated Go out a layer neuron hkBiasing number；
Step 7：Using MATLAB softwares, by training sample data Z={ z1,z2,…,zn,…,zNImport BP nerve nets Network model is trained BP neural network model；
Step 8：In MATLAB softwares, the test data set Y={ y that will obtain in step 31,y2,…,ym,…,yM} It imports among BP neural network model, predicts the vibration data group T={ t at the point of following U period1,t2,…, tu,…,tU}.Wherein, tuFor 5 vibration datas within u-th of period, and there is tu={ tu1,tu2,tu3,tu4,tu5, tu1 Represent the data of light condition, tu2Represent the data of light condition, tu3The data of state, t are carried in expressionu4Represent heavy condition Data, tu5Represent the data of super-heavy load state；
Fig. 3 is the overall flow figure of neural network model；Training neural network model, is mainly missed by BP neural network Poor backpropagation and the thought of signal forward-propagating, by constantly comparing the error of reality output and desired output, to adjust Weights between each neuron make the error between desired output data and real data constantly reduce, until meeting training step Number or error precision, the neural network model after just being trained.By the training of real data, make in BP neural network model Parameter it is more accurate, finally test sample is predicted, obtained data are more accurate.
Further, step 7：BP neural network model is trained, is comprised the steps of：
Step 7.1：Input layer includes three neurons, and hidden layer neuron g is calculated using the following formulajOutput valve：
Output layer neuron h is calculated using the following formula againkOutput valve：
Define error function：
Wherein：ykFor the desired output of output layer neuron, initial value is given history training sample data；E is Deviation；
Step 7.2：The output layer neuron h being calculatedkOutput valve substitute into loss function, deviation E is calculated； Whether judgment bias value E meets the requirements, if meeting the requirements, goes to step 7.9；If being unsatisfactory for requiring, 7.3 are gone to step；
Step 7.3：Use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(c+1)=(1- γ) gjek+γΔwjkc
Wherein：
Wherein：γ be weights inertia coeffeicent, ecAnd ec-1Respectively c and c-1 training error；Δwjk(c)For the c times instruction Hidden layer neuron g when practicingjTo output layer neuron hkWeight adjustment amount；Δwjk(c+1)For the c+1 times it is trained when hidden layer god Through first gjTo output layer neuron hkWeight adjustment amount；
Step 7.4：Use again following formula calculate input layer to hidden layer weight adjustment amount for：
Wherein：Δwij(c)For the c times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；Δ wij(c+1)For the c+1 times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；
Step 7.5：Biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 7.6：Biasing number a is calculated using following formulajUpdated value：
Step 7.7：Therefore, using the weight adjustment amount of the hidden layer that step 7.3 is calculated to output layer, step 7.4 The biasing number b that the input layer being calculated is calculated to the weight adjustment amount of hidden layer, step 7.5kUpdated value and step The corresponding ginseng for the BP neural network model that once training obtains before the updated value of the rapid 7.6 biasing number a being calculated is optimized and revised Number, thus obtains updated BP neural network model；
Step 7.8：Based on the updated BP neural network model that step 7.7 obtains, return to step 7.1,
Step 7.9：BP neural network model after being trained；
State following at the measuring point may determine that according to the vibration data of prediction gained, shaken according to the gantry crane measuring point of prediction The size of dynamic data, can predict whether gantry crane can work normally in different states under loads, if vibration data changes It is larger, then it needs to carry out gantry crane repair in advance；If vibration data variation is little, daily dimension can be carried out to gantry crane Shield.
The beneficial effects of the invention are as follows：Gantry crane is predicted using BP neural network, compensates for the prior art not to gantry crane The deficiency that state is predicted.By can accurately predict the prediction of data the state of gantry crane, and can be effectively to gantry crane Care and maintenance is carried out, significantly reduces the risk of gantry crane in use.
Illustrate that the present invention is predicting vibration data in future using gantry crane data with existing followed by a case history Accuracy, and judge gantry crane state and make corresponding measure according to data, ensure the normal work of gantry crane.
Case history：
Vibration signal of the experimental data on gantry crane crane shaft chooses a measuring point (the gantry crane land side horizontal stroke of gantry crane Beam large vehicle walking direction vibrate (Z8H)) on data, in data take out on December 28 0 when by 01 month 03 day 23 when vibration Data are as training sample, and vibration data when taking out at 18 days 14 January by 2 months 7 days 23 is as test sample.Due to acquisition Sensor one data of acquisition in every 10 to 20 seconds of system, so 8,000 or so data can be gathered in one day, in order to analyze The gantry crane loaded-up condition of 3 weeks by a definite date, it is necessary first to K- mean clusters be carried out according to step 2 to data, taken out more representative Data.
By observing these data, it is known that gantry crane is much like in daily Vibration Condition, it is known that gantry crane is daily Handling situations it is identical, according to gantry crane different conditions by the state of gantry crane be divided into super-heavy load, heavy duty, middle load, underloading, zero load five Kind of classification, by the data of every day according to size order obtain it is different be divided into five classes, be shown in Table 1.
Table 1：The result that 21 day data of gantry crane is clustered
December 28 December 29 December 30 December 31 January 1 January 2 January 3
0.0586 3 0.0352 2 0.0186 1 0.0825 4 0.0335 2 0.1255 5 0.0228 1
0.0268 1 0.0528 3 0.0332 2 0.0244 1 0.0491 3 0.024 1 0.0506 3
0.081 4 0.124 5 0.0724 4 0.0382 2 0.1346 5 0.0519 3 0.0346 2
0.0399 2 0.0226 1 0.1242 5 0.0568 3 0.0749 4 0.0756 4 0.1254 5
0.1257 5 0.0757 4 0.051 3 0.1355 5 0.0215 1 0.0353 2 0.073 4
January 18 January 19 January 20 January 21 January 22 January 23 January 24
0.1259 5 0.0189 1 0.1117 5 0.0611 3 0.052 3 0.0224 1 0.0723 4
0.0506 3 0.0516 3 0.032 2 0.1794 5 0.0191 1 0.0357 2 0.0497 3
0.0325 2 0.1258 5 0.0703 4 0.0362 2 0.078 4 0.0749 4 0.0332 2
0.0768 4 0.0773 4 0.0475 3 0.1021 4 0.1334 5 0.0521 3 0.1199 5
0.0208 1 0.0315 2 0.0194 1 0.0195 1 0.0328 2 0.1285 5 0.0222 1
January 25 January 26 January 27 January 28 January 29 January 30 January 31
0.0353 2 0.0465 3 0.1154 5 0.0971 5 0.0297 2 0.0699 4 0.0799 4
0.0509 3 0.0651 4 0.0301 2 0.016 1 0.0649 4 0.0203 1 0.0175 1
0.0233 1 0.0312 2 0.0698 4 0.0604 4 0.0185 1 0.109 5 0.0503 3
0.1073 5 0.0182 1 0.0166 1 0.0289 2 0.0444 3 0.0313 2 0.1395 5
0.0702 4 0.1028 5 0.0464 3 0.0419 3 0.1059 5 0.0486 3 0.0307 2
2 months 1 2 months 2 2 months 3 2 months 4 2 months 5 2 months 6 2 months 7
0.0698 4 0.0202 1 0.1113 5 0.0433 3 0.0315 2 0.0472 3 0.0352 2
0.0223 1 0.0336 2 0.0183 1 0.0192 1 0.0471 3 0.0319 2 0.0775 4
0.1106 5 0.0515 3 0.0449 3 0.1128 5 0.1146 5 0.0695 4 0.0526 3
0.05 3 0.0735 4 0.0294 2 0.0288 2 0.0193 1 0.119 5 0.1235 5
0.0345 2 0.1174 5 0.0663 4 0.0645 4 0.0674 4 0.0203 1 0.0227 1
Wherein：It is carried in 1- zero loads, 2- underloadings, 3-, 4- is heavily loaded, 5- super-heavy loads
Various types of data is arranged to sort out according to the order of time respectively and obtains table 2, then to BP nerve nets in Matlab Network model is trained, and is made training sample training neural network model using the data in December 28 to 03 day 01 month, is instructed BP neural network model after white silk.Test sample data of the January 25 to 7 days 2 months are analyzed and processed.
Table 2：Gantry crane cluster centre is arranged by different classifications
December 28 December 29 December 30 December 31 January 1 January 2 January 3
0.0268 1 0.0226 1 0.0186 1 0.0244 1 0.0215 1 0.024 1 0.023 1
0.0399 2 0.0352 2 0.0332 2 0.0382 2 0.0335 2 0.035 2 0.035 2
0.0586 3 0.0528 3 0.051 3 0.0568 3 0.0491 3 0.052 3 0.051 3
0.081 4 0.0757 4 0.0724 4 0.0825 4 0.0749 4 0.076 4 0.073 4
0.1257 5 0.124 5 0.1242 5 0.1355 5 0.1346 5 0.126 5 0.125 5
January 18 January 19 January 20 January 21 January 22 January 23 January 24
0.0208 1 0.0189 1 0.0194 1 0.0195 1 0.0191 1 0.022 1 0.022 1
0.0325 2 0.0315 2 0.032 2 0.0362 2 0.0328 2 0.036 2 0.033 2
0.0506 3 0.0516 3 0.0475 3 0.0611 3 0.052 3 0.052 3 0.05 3
0.0768 4 0.0773 4 0.0703 4 0.1021 4 0.078 4 0.075 4 0.072 4
0.1259 5 0.1258 5 0.1117 5 0.1794 5 0.1334 5 0.129 5 0.12 5
January 25 January 26 January 27 January 28 January 29 January 30 January 31
0.0233 1 0.0182 1 0.0166 1 0.016 1 0.0185 1 0.02 1 0.018 1
0.0353 2 0.0312 2 0.0301 2 0.0289 2 0.0297 2 0.031 2 0.031 2
0.0509 3 0.0465 3 0.0464 3 0.0419 3 0.0444 3 0.049 3 0.05 3
0.0702 4 0.0651 4 0.0698 4 0.0604 4 0.0649 4 0.07 4 0.08 4
0.1073 5 0.1028 5 0.1154 5 0.0971 5 0.1059 5 0.109 5 0.14 5
2 months 1 2 months 2 2 months 3 2 months 4 2 months 5 2 months 6 2 months 7
0.0223 1 0.0202 1 0.0183 1 0.0192 1 0.0193 1 0.02 1 0.023 1
0.0345 2 0.0336 2 0.0294 2 0.0288 2 0.0315 2 0.032 2 0.035 2
0.05 3 0.0515 3 0.0449 3 0.0433 3 0.0471 3 0.047 3 0.053 3
0.0698 4 0.0735 4 0.0663 4 0.0645 4 0.0674 4 0.07 4 0.078 4
0.1106 5 0.1174 5 0.1113 5 0.1128 5 0.1146 5 0.119 5 0.124 5
Wherein：It is carried in 1- zero loads, 2- underloadings, 3-, 4- is heavily loaded, 5- super-heavy loads
Test sample is imported into BP neural network model, is shaken using BP neural network model prediction gantry crane is several days following Dynamic data.It is accurate that verification herein predicts whether, using the vibration data of last two days as verification data, then utilizes BP neural network Obtained prediction data is as shown in table 3.
Table 3：Utilize BP neural network prediction result and actual result
Understand that the relative error of prediction result and actual result, can be accurate in real process between 2%~7% Predict the data of gantry crane future within these few days.Whether gantry crane state can extremely be judged by data.In this example, lead to Cross the data that prediction obtains, it can be determined that gantry crane is in the state of normal work, so carrying out operational maintenance to gantry crane；If There is big fluctuation in gantry crane data, then need to carry out maintenance gantry crane, ensure that gantry crane works normally at harbour.
Using BP neural network algorithm opposite bank bridge status predication, it can predict data of the gantry crane in future, pass through vibration Trend and size of the data under different conditions, judge gantry crane state, more efficiently gantry crane are safeguarded.Pass through engineering The result of instance analysis, it can be appreciated that the method for the present invention can accurately predict gantry crane data, judge the shape of gantry crane State.
The method of the present invention is obtained extension data, is overcome existing method and be not previously predicted bank by gantry crane available data, prediction A kind of deficiency of bridge like state, it is proposed that method of Accurate Prediction gantry crane state；Improve work effect of the gantry crane in port traffic Rate reduces the risk during gantry crane works at harbour, is of great significance in harbour field.
Although the present invention is described in detail with reference to the foregoing embodiments, for those skilled in the art, It still can modify to the technical solution recorded in foregoing embodiments or which part technical characteristic is carried out etc. With replacing, within the spirit and principles of the invention, any modifications, equivalent replacements and improvements are made should be included in this Within the protection domain of invention.

Claims (1)

1. a kind of gantry crane trend prediction method based on BP neural network, it is characterised in that comprise the following steps：
Step 1：The data of a gantry crane measuring point whithin a period of time are obtained in the database, which utilized in installation gantry crane Sensor on bridge motor extracted real-time vibration severity data at interval of 10 to 20 seconds, and stores in the database, described Vibration severity data are used for assessing gantry crane loaded-up condition；Acquired data V={ v1,v2,···,vl,···,vL, In, L represents the number of the test sample, vlIt represents to obtain the l articles vibration data in data；
Step 2：The mode of the data of acquisition temporally decile is divided into K group data, and to every group of data K- mean cluster Mode, calculates 5 clustering clusters, and one of which cluster centre data are P={ P1,P2,P3,P4,P5, it will be per group cluster central point It falls into 5 types from small to large according to data, and new data group, all data is formed with tag along sort data Q={ 1,2,3,4,5 } It is X={ x that group, which forms new sample data,1,x2,···,xs,···,xS, wherein, tag along sort data 1,2,3,4,5 Five kinds of gantry crane zero load, underloading, middle load, heavy duty, super-heavy load states are represented respectively；S is the number of sample data, and S=5K, K are Positive integer, xsRepresent the s articles characteristic in sample data；
Step 3：Arrangement classification is carried out to the time-based sequencing of sample data and different classifications label, obtaining data group is D={ d1,d2,…,do,…,dK}；Wherein, doFor the vibration data of 5 kinds of states within o-th of period, and there is do={ do1, do2,do3,do4,do5, do1Represent the data of light condition, do2Represent the data of light condition, do3The number of state is carried in expression According to do4Represent the data of heavy condition, do5Represent the data of super-heavy load state；
By data group D={ d1,d2,…,do,…,dKIt is divided into two class of test sample data and training sample data, wherein the survey The vibration data of sample sheet is Y={ y1,y2,···,ym,···,yM, wherein M represents the number of the test sample, ym Represent the m datas in the test sample；The vibration data of training sample is Z={ z1,z2,···,zn,···, zN, wherein N represents the number of the training sample, znRepresent the nth bar data in the training sample；
Step 4：Test sample data and training sample data are normalized, generation normalization characteristic data Y '= {y1', y2', ym', yM' and Z '={ z1', z2', zn', zN', wherein, ym' represent M datas in the normalization characteristic data Y ', znNth bar data in the ' expression normalization characteristic data Z '；
Normalizing calculation is：
<mrow> <msup> <mi>X</mi> <mo>*</mo> </msup> <mo>=</mo> <mfrac> <mrow> <mi>X</mi> <mo>-</mo> <msub> <mi>X</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> <mrow> <msub> <mi>X</mi> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>X</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> </mrow>
Wherein：Xmax, XminThe respectively maximum and minimum value of actual sample data, X* are the value after normalization, and X is actual Value；
After handling data using BP neural network, renormalization calculating is carried out to data, just obtains actual data value；Anti- normalizing Changing equation is：
X=(Xmax-Xmin)×X*+Xmin
Step 5：Establish BP neural network model；The topological structure of the BP neural network model include input layer, hidden layer and Output layer；Wherein, the neuron number of input layer is f, and the neuron number of hidden layer is g, and the neuron number of output layer is h；Arbitrary input layer is fi, i ∈ (1,2 ... p)；Arbitrary hidden layer neuron is gj, j ∈ (1,2 ... q)；Arbitrary output layer Neuron is hk, k ∈ (1,2 ... r)；
Step 6：The basic parameter of BP neural network model is initialized, including：The weight of learning rate μ, input layer to hidden layer wij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to output layer biasing number bkWith And excitation function R (x)；Wherein, input layer is to the weight w of hidden layerij, hidden layer to output layer weight wjk, input layer is to hidden Biasing number a containing layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Input layer is to the weight w of hidden layerijMeaning is：Arbitrary input layer fiTo arbitrary hidden layer nerve First gjBetween weight；Hidden layer is to the weight w of output layerjkMeaning is：Arbitrary hidden layer neuron gjTo arbitrary output Layer neuron hkBetween weight；Input layer is to the biasing number a of hidden layerjMeaning is：Each input layer is implied to arbitrary Layer neuron gjBiasing number；Hidden layer is to the biasing number b of output layerkMeaning is：Each hidden layer neuron is to arbitrary output layer Neuron hkBiasing number；
Step 7：Using MATLAB softwares, by training sample data Z={ z1,z2,···,zn,···,zNImport BP nerves Network model is trained BP neural network model；
Step 8：In MATLAB softwares, the test data set Y={ y that will obtain in step 31,y2,…,ym,…,yMImport Among BP neural network model, the vibration data group T={ t at the point of following U period are predicted1,t2,…,tu,…, tU}；Wherein, tuFor 5 vibration datas within u-th of period, and there is tu={ tu1,tu2,tu3,tu4,tu5, tu1Represent empty The data of load state, tu2Represent the data of light condition, tu3The data of state, t are carried in expressionu4Represent the data of heavy condition, tu5Represent the data of super-heavy load state；
Step 7 comprises the steps of：
Step 7.1：Input layer includes three neurons, and hidden layer neuron g is calculated using the following formulajOutput valve：
<mrow> <msub> <mi>g</mi> <mi>j</mi> </msub> <mo>=</mo> <mi>&amp;mu;</mi> <mrow> <mo>(</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>p</mi> </munderover> <msub> <mi>w</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <msub> <mi>f</mi> <mi>i</mi> </msub> <mo>+</mo> <msub> <mi>a</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> </mrow>
Output layer neuron h is calculated using the following formula againkOutput valve：
<mrow> <msub> <mi>h</mi> <mi>k</mi> </msub> <mo>=</mo> <munderover> <mi>&amp;Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>q</mi> </munderover> <msub> <mi>g</mi> <mi>j</mi> </msub> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <mo>+</mo> <msub> <mi>b</mi> <mi>k</mi> </msub> </mrow>
Define error function：
<mrow> <mi>E</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mi>h</mi> <mi>k</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow>
Wherein：ykFor the desired output of output layer neuron, initial value is given history training sample data；E is deviation Value；
Step 7.2：The output layer neuron h being calculatedkOutput valve substitute into loss function, deviation E is calculated；Judge Whether deviation E meets the requirements, if meeting the requirements, goes to step 7.9；If being unsatisfactory for requiring, 7.3 are gone to step；
Step 7.3：Use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(c+1)=(1- γ) gjek+γΔwjkc
Wherein：
<mrow> <mi>&amp;gamma;</mi> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mn>1</mn> <mo>,</mo> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>c</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>c</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> <mo>&lt;</mo> <mn>1</mn> </mtd> </mtr> <mtr> <mtd> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mn>5</mn> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>c</mi> </msub> <mo>|</mo> </mrow> </mrow> <mrow> <mn>6</mn> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>c</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mrow> </mfrac> <mo>,</mo> <mn>1</mn> <mo>&amp;le;</mo> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>c</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>c</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> <mo>&amp;le;</mo> <mn>2</mn> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> <mo>,</mo> <mn>2</mn> <mo>&lt;</mo> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>c</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>c</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> </mtd> </mtr> </mtable> </mfenced> </mrow>
Wherein：γ be weights inertia coeffeicent, ecAnd ec-1Respectively c and c-1 training error；Δwjk(c)For the c times it is trained when Hidden layer neuron gjTo output layer neuron hkWeight adjustment amount；Δwjk(c+1)For the c+1 times it is trained when hidden layer neuron gjTo output layer neuron hkWeight adjustment amount；
Step 7.4：Use again following formula calculate input layer to hidden layer weight adjustment amount for：
<mrow> <msub> <mi>&amp;Delta;w</mi> <mrow> <mi>j</mi> <mi>k</mi> <mrow> <mo>(</mo> <mi>c</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> </msub> <mo>=</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>&amp;gamma;</mi> <mo>)</mo> </mrow> <msub> <mi>&amp;mu;g</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msub> <mi>g</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <msub> <mi>f</mi> <mi>i</mi> </msub> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>r</mi> </munderover> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <msub> <mi>e</mi> <mi>k</mi> </msub> <mo>+</mo> <msub> <mi>&amp;gamma;&amp;Delta;w</mi> <mrow> <mi>j</mi> <mi>k</mi> <mi>c</mi> </mrow> </msub> </mrow>
Wherein：Δwij(c)For the c times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；Δ wij(c+1)For the c+1 times it is trained when input layer fiTo hidden layer neuron gjWeight adjustment amount；
Step 7.5：Biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 7.6：Biasing number a is calculated using following formulajUpdated value：
<mrow> <msub> <mi>a</mi> <mi>j</mi> </msub> <mo>=</mo> <msub> <mi>a</mi> <mi>j</mi> </msub> <mo>+</mo> <msub> <mi>&amp;mu;g</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msub> <mi>g</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <msub> <mi>e</mi> <mi>k</mi> </msub> </mrow>
Step 7.7：Therefore, calculated using the weight adjustment amount of the hidden layer that step 7.3 is calculated to output layer, step 7.4 The biasing number b that obtained input layer is calculated to the weight adjustment amount of hidden layer, step 7.5kUpdated value and step 7.6 The updated value for the biasing number a being calculated once trains the correspondence parameter of obtained BP neural network model before optimizing and revising, by This obtains updated BP neural network model；
Step 7.8：Based on the updated BP neural network model that step 7.7 obtains, return to step 7.1,
Step 7.9：BP neural network model after being trained；
State following at the measuring point may determine that according to the vibration data of prediction gained, according to the gantry crane measuring point vibration number of prediction According to size, can predict whether gantry crane can work normally in different state under loads, if vibration data changes greatly, It then needs to carry out gantry crane repair in advance；If vibration data variation is little, daily maintenance can be carried out to gantry crane.
CN201711328788.9A 2017-12-13 2017-12-13 A kind of gantry crane trend prediction method based on BP neural network Pending CN108090608A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711328788.9A CN108090608A (en) 2017-12-13 2017-12-13 A kind of gantry crane trend prediction method based on BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711328788.9A CN108090608A (en) 2017-12-13 2017-12-13 A kind of gantry crane trend prediction method based on BP neural network

Publications (1)

Publication Number Publication Date
CN108090608A true CN108090608A (en) 2018-05-29

Family

ID=62175028

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711328788.9A Pending CN108090608A (en) 2017-12-13 2017-12-13 A kind of gantry crane trend prediction method based on BP neural network

Country Status (1)

Country Link
CN (1) CN108090608A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109117576A (en) * 2018-08-29 2019-01-01 上海海事大学 A kind of method of determining gantry crane structural loads and real-time stress field
CN109685140A (en) * 2018-12-25 2019-04-26 上海海事大学 A kind of DBSCAN algorithm gantry crane state classification method based on principal component analysis
CN110020712A (en) * 2019-03-26 2019-07-16 浙江大学 A kind of optimization population BP neural network forecast method and system based on cluster
CN110032555A (en) * 2019-04-16 2019-07-19 上海建科工程咨询有限公司 A kind of neural network tower crane Risk Forecast Method and system
CN111160528A (en) * 2019-12-28 2020-05-15 浙江大学 Method for predicting service performance degradation of reinforced concrete bridge
CN111860839A (en) * 2020-07-28 2020-10-30 上海海事大学 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
CN112434929A (en) * 2020-11-19 2021-03-02 广西电网有限责任公司电力科学研究院 Method for evaluating risk resistance of distribution transformer
CN116957366A (en) * 2023-09-20 2023-10-27 中交第四航务工程勘察设计院有限公司 Port data acquisition and evaluation method and system based on Internet of things

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103778276A (en) * 2013-12-27 2014-05-07 河海大学 Reliability degree predication method for composite material based on fatigue life distribution
CN107194625A (en) * 2017-07-25 2017-09-22 国家电网公司 Wind power plant based on neutral net abandons wind-powered electricity generation amount appraisal procedure

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103778276A (en) * 2013-12-27 2014-05-07 河海大学 Reliability degree predication method for composite material based on fatigue life distribution
CN107194625A (en) * 2017-07-25 2017-09-22 国家电网公司 Wind power plant based on neutral net abandons wind-powered electricity generation amount appraisal procedure

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
张冰倩等: "基于K-均值聚类算法的岸桥载荷分类识别", 《东华大学学报（自然科学版）》 *
王志欣: "岸桥工况状态的神经网络识别技术的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109117576A (en) * 2018-08-29 2019-01-01 上海海事大学 A kind of method of determining gantry crane structural loads and real-time stress field
CN109685140A (en) * 2018-12-25 2019-04-26 上海海事大学 A kind of DBSCAN algorithm gantry crane state classification method based on principal component analysis
CN110020712A (en) * 2019-03-26 2019-07-16 浙江大学 A kind of optimization population BP neural network forecast method and system based on cluster
CN110032555A (en) * 2019-04-16 2019-07-19 上海建科工程咨询有限公司 A kind of neural network tower crane Risk Forecast Method and system
CN111160528A (en) * 2019-12-28 2020-05-15 浙江大学 Method for predicting service performance degradation of reinforced concrete bridge
CN111160528B (en) * 2019-12-28 2021-01-08 浙江大学 Method for predicting service performance degradation of reinforced concrete bridge
CN111860839A (en) * 2020-07-28 2020-10-30 上海海事大学 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
CN112434929A (en) * 2020-11-19 2021-03-02 广西电网有限责任公司电力科学研究院 Method for evaluating risk resistance of distribution transformer
CN116957366A (en) * 2023-09-20 2023-10-27 中交第四航务工程勘察设计院有限公司 Port data acquisition and evaluation method and system based on Internet of things
CN116957366B (en) * 2023-09-20 2024-01-16 中交第四航务工程勘察设计院有限公司 Port data acquisition and evaluation method and system based on Internet of things

Similar Documents

Publication Publication Date Title
CN108090608A (en) A kind of gantry crane trend prediction method based on BP neural network
Eftekhar Azam et al. Damage detection in structural systems utilizing artificial neural networks and proper orthogonal decomposition
CN105973594B (en) A kind of rolling bearing fault Forecasting Methodology based on continuous depth confidence network
CN104712542B (en) A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults
CN106204392A (en) Environment risk source early warning system
CN109063308A (en) A kind of health evaluating method based on depth quantum learning
CN109636171A (en) A kind of comprehensive diagnos and risk evaluating method that regional vegetation restores
CN108345666A (en) A kind of vehicle abnormality track-detecting method based on time-space isolated point
CN102539098A (en) Bridge dynamic load testing method based on neural network technology
CN114072825A (en) Monitoring, predicting and maintaining condition of railway elements using digital twinning
CN106209829A (en) A kind of network security management system based on warning strategies
Kumar Artificial neural network as a tool for reservoir characterization and its application in the petroleum engineering
Marinelli et al. Earthmoving trucks condition level prediction using neural networks
CN115953252A (en) Method for determining construction safety liability insurance premium
CN103413207A (en) Scientific and technological talent evaluation method based on qualitative standardization and impact factor evaluation method
CN102252861A (en) Progressive method for identifying support generalized displacement of damaged cable based on space coordinate monitoring
CN114757365A (en) High-speed railway roadbed settlement prediction and early warning method based on deep learning
CN109784586A (en) The prediction technique and system of the situation of being in danger of vehicle insurance
CN117291072B (en) Bridge damage identification method
Akintunde et al. Singular value decomposition and unsupervised machine learning for virtual strain sensing: Application to an operational railway bridge
Kyzym et al. Simulating Development of Science in a Country with the Use of the Cognitive Approach
Gao et al. Gas outburst prediction based on the intelligent Dempster-Shafer evidence theory
CN115906669A (en) Dense residual error network landslide susceptibility evaluation method considering negative sample selection strategy
Bahamon-Blanco et al. Recognition of track defects through measured acceleration using a recurrent neural network
Bhosale et al. Use of Artificial Neural Network in Construction Management

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20180529