CN111860839A - Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm - Google Patents

Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm Download PDF

Info

Publication number
CN111860839A
CN111860839A CN202010735618.8A CN202010735618A CN111860839A CN 111860839 A CN111860839 A CN 111860839A CN 202010735618 A CN202010735618 A CN 202010735618A CN 111860839 A CN111860839 A CN 111860839A
Authority
CN
China
Prior art keywords
signal
shore bridge
signals
temperature
formula
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
CN202010735618.8A
Other languages
Chinese (zh)
Inventor
唐刚
常超
邵长专
胡雄
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 CN202010735618.8A priority Critical patent/CN111860839A/en
Publication of CN111860839A publication Critical patent/CN111860839A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • 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

Abstract

The invention provides a signal processing method, which fuses common signals of a shore bridge so as to obtain more fusion characteristic parameters, thereby more effectively utilizing information such as vibration, temperature, stress and the like and more accurately diagnosing and monitoring faults of the state of the shore bridge. The neural network is optimized by using the Adam algorithm, and convenience is provided for more accurately and rapidly monitoring the state of the shore bridge. The method comprises the steps of 1, collecting data; step 2, data preprocessing is carried out, and input and output are determined; step 3, constructing a neural network model; step 4, training a neural network by using an Adam algorithm in a training set to obtain a quayside crane state prediction model; step 5, checking a shore bridge state prediction model by using the test set; and 6, outputting the prediction result, and displaying the prediction result in a human-computer interaction interface. According to the method, a plurality of signals are adopted for fusion prediction, so that the fault tolerance of a certain signal is improved, and the accuracy of monitoring the operating state of the shore bridge is improved.

Description

Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
Technical Field
The invention relates to a shore bridge fault monitoring method based on multi-signal fusion and an Adam optimization algorithm, and belongs to the technology of shore bridge state prediction.
Background
The fault diagnosis of the shore bridge is always a technical difficulty, and the stability of the shore bridge has important significance for the safe operation of port transportation and trade. With the development of big data theory, a large amount of information generated in the shore bridge monitoring process is mined, and signal fusion, processing and diagnosis are carried out by using an artificial intelligence technology, so that the method becomes a new direction for state prediction and fault diagnosis of the shore bridge. The prior art mainly focuses on independently monitoring vibration and stress signals, and lacks fusion diagnosis of signals of temperature, vibration, stress and the like.
The BP neural network has strong nonlinear fitting capacity, the traditional BP neural network is low in convergence speed and difficult in super-parameter selection, real-time monitoring and diagnosis cannot be well performed, and meanwhile, when all monitoring quantities are directly used for modeling, too much monitoring quantities can cause the model to be too complex, so that the model is difficult to train due to too large calculation quantity, and the overfitting phenomenon is easy to occur.
Disclosure of Invention
Aiming at monitoring and diagnosing the operating state of the shore bridge, the invention provides a signal processing method, which fuses common signals of the shore bridge, such as vibration signals, temperature signals, stress signals and the like, so as to obtain more characteristic parameters and more effectively utilize information such as vibration temperature stress and the like, thereby providing more information for more accurately monitoring the state of the shore bridge.
The method provides a shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm, and comprises the following steps:
step 1, collecting data;
acquiring signals of vibration, temperature, stress and the like of a mechanical structure of the shore bridge in a period of time, wherein the signals of vibration, temperature, stress and the like are synchronous in time;
step 2, data preprocessing is carried out, and input and output are determined;
preprocessing the acquired signals, such as removing singular points of the signals by using a 3-probe method, performing data normalization processing and the like, taking the signals to be predicted as output signals, analyzing the correlation between the output signals and other signals by using a gray correlation analysis method, selecting the signals with large correlation as input signals according to needs, and dividing input signal samples into a training set and a test set;
step 3, constructing a neural network model;
determining training parameters including hidden layer node number, maximum iteration times, initial learning rate and the like according to the dimensionality of an input signal, namely the type of the input signal selected as required, and utilizing an Xavier initialization method to initialize a neural network weight and bias;
step 4, training a neural network by using an Adam algorithm in a training set to obtain a quayside crane state prediction model;
step 5, checking a shore bridge state prediction model by using the test set;
testing the BP neural network by using the data of the test set, and checking the network performance;
and 6, outputting the prediction result, and displaying the prediction result in a human-computer interaction interface.
Compared with the prior art, the invention has the following technical effects:
1. the invention carries out normalization processing on the data with strong correlation, so that signals with different physical meanings can be compared with each other;
2. the shore bridge state is predicted by fusing various signals, so that the accuracy of predicting the shore bridge state is greatly improved, and the safe and reliable operation of the shore bridge is guaranteed;
3. the optimization neural network based on the Adam algorithm has high prediction precision, can reliably predict the monitoring signal trend of the shore bridge, and has important significance for the safe operation of the shore bridge.
Drawings
The drawings of the invention are illustrated as follows:
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the Adam algorithm;
fig. 3 is a diagram of a BP neural network structure.
Detailed Description
The present invention is further described with reference to the accompanying drawings, and the following examples are only for better illustrating the technical solutions of the present invention, and should not be taken as limiting the scope of the present invention.
The invention discloses a shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm, which is shown in figure 1 and comprises the following processes:
step 1, collecting data;
due to the fact that the shore bridge is large in size, the whole shore bridge mechanism and structure can be monitored only by selecting special and representative measuring points. Taking a vibration signal, a stress signal and a temperature signal as examples, because the vibration and temperature change of a girder, a main motor and a trolley traction motor are most obvious in operation, a data acquisition point is determined at the position, the vibration signal can be the vibration acceleration of the girder and the like, the stress signal can be the stress condition of some important positions on a crane, such as the position with the maximum deformation of the girder and the position of a pull rod and the like, the temperature signal is monitored at the same position, and because the temperature is a slowly-varying signal, the input data can be selected to be a proper average value according to requirements.
Step 2, data preprocessing is carried out, and input and output are determined;
the data singular point refers to that data has mutation at a certain moment, the data numerical value of the data is greatly different from the data numerical value of the data at the previous moment and the data numerical value of the data at the next moment, and the data measured at the moment can be considered to be wrong, so the data needs to be removed, and the data accuracy is ensured. Monitoring data by using a 3-probe method to eliminate abnormal values and fill missing values, describing by taking a temperature signal as an example, under a normal condition, the temperature increment delta theta at a certain momentT(k) Satisfies the following conditions:
Figure BDA0002604828280000031
in the formula,. DELTA.theta.T(k) Indicates the increase of temperature at the k-th time, thetaT(k) Denotes the temperature at the k-th time, thetaT(k-1) denotes the temperature at the k-1 th time, and μ denotes the temperature increase Δ θ at all timesT(k) Mean of samples of composition, representing Δ θT(k) Standard deviation of the samples. If the abnormal points do not meet the formula, the abnormal points are regarded as collected abnormal points and removed;
normalization processing means that input quantities are different in physical meaning and dimension, normalization processing is required for enabling the input quantities to be compared with each other, and input components in network training are equivalent due to the fact that data after normalization are changed between [0 and 1 ]. The data normalization processing formula is as follows:
Figure BDA0002604828280000032
in the formula, x*For normalized values of the data, xiIs the ith data, x, of the signalmax、xminRespectively the maximum value and the minimum value of the sample data;
after the output signal is determined, the correlation between the output signal and other monitoring signals is analyzed by utilizing a grey correlation analysis method, and a signal with large correlation is selected as an input signal as required. Taking the vibration signal and the temperature signal as an example, the gray correlation calculation formula is as follows:
Figure BDA0002604828280000033
in the formula, y0(k) Value, y, representing the k-th moment of the vibration signali(k) Value, Y, representing the kth time of the temperature signal0Representing a vibration signal, YiIndicating temperature signal, xi is resolution coefficient, its value is 0-1, gamma (Y)0,Yi) Represents Y0And YiGray degree of correlation of (Y) 0. ltoreq. gamma0,Yi) 1 or less, the closer the grey relevance is to 1, the greater the relevance of the two sequences is, the less relevant variables are generally, the grey relevance is close to 0, and the more relevant variables are close to 1;
monitoring quantities related to temperature signals can be selected after the gray relevance is sorted from large to small, a monitoring quantity sequence with large relevance is selected as a training data set, and specific determination limits and the selected number are determined according to the precision requirement and the complexity of a neural network model;
the normalized input samples are divided into a training set and a test set.
Step 3, constructing a neural network model;
determining training parameters according to the dimension of an input signal, wherein the training parameters comprise hidden layer node number, maximum iteration times and initial learning rate, and the dimension of the input signal corresponds to the input layer node number;
hidden layer node number determination formula:
Figure BDA0002604828280000041
in the formula: m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and α is a constant between 1 and 10.
Step 4, training a neural network by using an Adam algorithm in a training set to obtain a quayside crane state prediction model;
the Adam algorithm trains the neural network specifically as follows:
s1, calculating a gradient:
Figure BDA0002604828280000042
s2, calculating the biased first moment estimation: s (k) ═ ρ1s(k-1)+(1-ρ1)g
S3, calculating the estimation of the biased second moment: r (k) ═ p2r(k-1)+(1-ρ2)g⊙g
S4, calculating a correction first moment:
Figure BDA0002604828280000043
s5, calculating a correction second moment:
Figure BDA0002604828280000044
s6, correction parameter value:
Figure BDA0002604828280000045
in the formula, f (x)(i)(ii) a P) represents the output function of the neural network, in the present invention particularly the diagnostic signal to be predicted, L (f (x)(i)；P),y(i)) Represents the cost function of the neural network, in the present invention, means the error function of the ideal value and the actual value of the diagnostic signal,
Figure BDA0002604828280000046
representing gradient operators, mRepresenting the size, p, of the training data set1Usually taken as 0.9, p2Usually 0.999, k is the number of iterations, representing the step size, usually 0.001, and is a small constant, usually 10-8。
Step 5, testing the neural network by using the test set data, and checking the network performance;
step 6, displaying the final prediction result on a human-computer interaction interface;
and performing inverse normalization processing on the signal output by the neural network to obtain a predicted value. And presenting the result on a human-computer interaction interface.

Claims (7)

1. A shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm is characterized by comprising the following processes:
step 1, collecting data;
acquiring signals of vibration, temperature, stress and the like of a mechanical structure of the shore bridge within a period of time, wherein the signals of vibration, temperature, stress and the like are synchronous in time;
step 2, data preprocessing is carried out, and input and output are determined;
preprocessing the acquired signals, taking the signals to be predicted as output signals, and dividing input signal samples into a training set and a test set;
step 3, constructing a neural network model;
determining training parameters including the number of hidden layer nodes, the maximum iteration times, the initial learning rate and the like according to the dimensionality of an input signal;
step 4, training a neural network by using the parameters obtained in the step 3 by using an Adam algorithm in a training set to obtain a quayside crane state prediction model;
step 5, checking the shore bridge state prediction model obtained in the step 4 by using a test set;
and 6, outputting the prediction result, and displaying the prediction result in a human-computer interaction interface.
2. The shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm according to claim 1, wherein in step 1, various monitoring signals including vibration signals, heat signals, stress signals and the like are measured.
3. The shore bridge fault monitoring method based on the multi-signal fusion and Adam optimization algorithm as claimed in claim 1, wherein in step 2, historical data of each monitored quantity is monitored by using a 3-probe method, taking a temperature signal as an example, and the formula of the 3-probe method is as follows:
Figure FDA0002604828270000011
in the formula,. DELTA.theta.T(k) Indicates the increase of temperature at the k-th time, thetaT(k) Denotes the temperature at the k-th time, thetaT(k-1) denotes the temperature at the k-1 th time, and μ denotes the temperature increase Δ θ at all timesT(k) Mean of samples of composition, representing Δ θT(k) Standard deviation of the samples.
4. The shore bridge fault monitoring method based on the multi-signal fusion and Adam optimization algorithm as claimed in claim 1, wherein in step 2, the historical data of each monitoring quantity is normalized, and the normalization formula is as follows:
Figure FDA0002604828270000012
in the formula, x*For normalized values of the data, xiIs the ith data, x, of the signalmax、xminThe maximum and minimum values of the sample data, respectively.
5. The shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm as claimed in claim 1, wherein in step 3, the hidden layer node number determination formula is as follows:
Figure FDA0002604828270000013
in the formula, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output layer nodes, and alpha is a constant between 1 and 10.
6. The shore bridge fault monitoring method based on the multi-signal fusion and Adam optimization algorithm according to claim 1, wherein a grey correlation analysis method is used for analyzing the correlation between any two monitoring signals, and taking a vibration signal and a temperature signal as an example, the grey correlation analysis method has the following formula:
Figure FDA0002604828270000021
in the formula, y0(k) Value, y, representing the k-th moment of the vibration signali(k) Value, Y, representing the kth time of the temperature signal0Representing a vibration signal, YiIndicating temperature signal, xi is resolution coefficient, its value is 0-1, gamma (Y)0,Yi) Represents Y0And YiGray degree of correlation of (Y) 0. ltoreq. gamma0,Yi) 1, the closer the grey relevance is to 1, indicating a greater correlation between the two sequences, generally less correlated variables, the closer the grey relevance is to 0, the closer the more correlated variables are to 1.
7. The shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm according to claim 1, wherein the Adam algorithm is used for optimizing a neural network, and the Adam algorithm has the following specific formula and steps:
s1, calculating a gradient:
Figure FDA0002604828270000022
s2, calculating the biased first moment estimation: s (k) ═ ρ1s(k-1)+(1-ρ1)g
S3, calculating the estimation of the biased second moment: r (k) ═ p2r(k-1)+(1-ρ2)g⊙g
S4, calculating a correction first moment:
Figure FDA0002604828270000023
s5, calculating a correction second moment:
Figure FDA0002604828270000024
s6, correction parameter value:
Figure FDA0002604828270000025
in the formula, f (x)(i)(ii) a P) represents the output function of the neural network, in the present invention particularly the diagnostic signal to be predicted, L (f (x)(i)；P),y(i)) Represents the cost function of the neural network, in the present invention, means the error function of the ideal value and the actual value of the diagnostic signal,
Figure FDA0002604828270000026
representing the gradient operator, m representing the size of the training data set, p1Usually taken as 0.9, p2Usually 0.999, k is the number of iterations, representing the step size, usually 0.001, and is a small constant, usually 10-8。
CN202010735618.8A 2020-07-28 2020-07-28 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm Pending CN111860839A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010735618.8A CN111860839A (en) 2020-07-28 2020-07-28 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010735618.8A CN111860839A (en) 2020-07-28 2020-07-28 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm

Publications (1)

Publication Number Publication Date
CN111860839A true CN111860839A (en) 2020-10-30

Family

ID=72947787

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010735618.8A Pending CN111860839A (en) 2020-07-28 2020-07-28 Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm

Country Status (1)

Country Link
CN (1) CN111860839A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112965376A (en) * 2021-02-03 2021-06-15 清华大学 Intelligent control method and device for arch dam temperature stress and transverse joint working state
CN113511592A (en) * 2021-06-30 2021-10-19 三一海洋重工有限公司 Shore bridge monitoring method and system and shore bridge
CN113723010A (en) * 2021-09-08 2021-11-30 福建中锐网络股份有限公司 Bridge damage early warning method based on LSTM temperature-displacement correlation model
CN114357594A (en) * 2022-03-08 2022-04-15 深圳市城市交通规划设计研究中心股份有限公司 Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU
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 (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015158198A1 (en) * 2014-04-17 2015-10-22 北京泰乐德信息技术有限公司 Fault recognition method and system based on neural network self-learning
CN108090608A (en) * 2017-12-13 2018-05-29 上海海事大学 A kind of gantry crane trend prediction method based on BP neural network
CN109242147A (en) * 2018-08-07 2019-01-18 重庆大学 Signal fused fan condition prediction technique based on Bp neural network
CN110502991A (en) * 2019-07-18 2019-11-26 武汉理工大学 Internal combustion engine health monitor method and system based on random convolutional neural networks structure
CN110702418A (en) * 2019-10-10 2020-01-17 山东超越数控电子股份有限公司 Aircraft engine fault prediction method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015158198A1 (en) * 2014-04-17 2015-10-22 北京泰乐德信息技术有限公司 Fault recognition method and system based on neural network self-learning
CN108090608A (en) * 2017-12-13 2018-05-29 上海海事大学 A kind of gantry crane trend prediction method based on BP neural network
CN109242147A (en) * 2018-08-07 2019-01-18 重庆大学 Signal fused fan condition prediction technique based on Bp neural network
CN110502991A (en) * 2019-07-18 2019-11-26 武汉理工大学 Internal combustion engine health monitor method and system based on random convolutional neural networks structure
CN110702418A (en) * 2019-10-10 2020-01-17 山东超越数控电子股份有限公司 Aircraft engine fault prediction method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
吕明;: "基于萤火虫优化BP神经网络的数控机床故障诊断", 机械设计与研究, no. 06 *
陈琛;潘文超;: "岸桥金属材料与焊接结构寿命预测", 机械科学与技术, no. 10 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112965376A (en) * 2021-02-03 2021-06-15 清华大学 Intelligent control method and device for arch dam temperature stress and transverse joint working state
CN112965376B (en) * 2021-02-03 2022-09-09 清华大学 Intelligent control method and device for arch dam temperature stress and transverse joint working state
CN113511592A (en) * 2021-06-30 2021-10-19 三一海洋重工有限公司 Shore bridge monitoring method and system and shore bridge
CN113511592B (en) * 2021-06-30 2023-09-22 三一海洋重工有限公司 Shore bridge monitoring method and system and shore bridge
CN113723010A (en) * 2021-09-08 2021-11-30 福建中锐网络股份有限公司 Bridge damage early warning method based on LSTM temperature-displacement correlation model
CN113723010B (en) * 2021-09-08 2024-01-26 福建中锐网络股份有限公司 Bridge damage early warning method based on LSTM temperature-displacement correlation model
CN114357594A (en) * 2022-03-08 2022-04-15 深圳市城市交通规划设计研究中心股份有限公司 Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU
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
CN111860839A (en) Shore bridge fault monitoring method based on multi-signal fusion and Adam optimization algorithm
CN111914883B (en) Spindle bearing state evaluation method and device based on deep fusion network
CN104390657B (en) A kind of Generator Unit Operating Parameters measurement sensor fault diagnosis method and system
CN112131212A (en) Hybrid cloud scene-oriented time sequence data anomaly prediction method based on ensemble learning technology
CN113128561A (en) Machine tool bearing fault diagnosis method
CN110530650B (en) Method for monitoring performance state of heavy-duty gas turbine based on generalized regression neural network and box diagram analysis
CN112101431A (en) Electronic equipment fault diagnosis system
CN111191726B (en) Fault classification method based on weak supervision learning multilayer perceptron
CN110705812A (en) Industrial fault analysis system based on fuzzy neural network
CN108709744B (en) Motor bearings method for diagnosing faults under a kind of varying load operating condition
CN111337244A (en) Method and device for monitoring and diagnosing faults of input shaft of fan gearbox
CN111122811A (en) Sewage treatment process fault monitoring method of OICA and RNN fusion model
CN101539137B (en) Residual life gray prediction method of aerospace hydraulic pump based on delta filter
CN112395684A (en) Intelligent fault diagnosis method for high-speed train running part system
CN115481726A (en) Industrial robot complete machine health assessment method and system
CN110320802B (en) Complex system signal time sequence identification method based on data visualization
CN110705114B (en) Ventilation fault diagnosis method without training sample
CN116662925A (en) Industrial process soft measurement method based on weighted sparse neural network
TWI794907B (en) Prognostic and health management system for system management and method thereof
CN115310693A (en) Comprehensive energy data cleaning method, system, equipment and medium
CN105699043B (en) A kind of wind tunnel sensors that improve measure stability and the method for precision
CN114800036A (en) Equipment health state assessment method
CN112307918A (en) Diagnosis method for transformer direct-current magnetic biasing based on fuzzy neural network
CN110705186A (en) Real-time online instrument checking and diagnosing method through RBF particle swarm optimization algorithm
Wang et al. Complex equipment diagnostic reasoning based on neural network algorithm

Legal Events

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