CN113607325A - Intelligent monitoring method and system for looseness positioning of steel structure bolt group - Google Patents
Intelligent monitoring method and system for looseness positioning of steel structure bolt group Download PDFInfo
- Publication number
- CN113607325A CN113607325A CN202111176679.6A CN202111176679A CN113607325A CN 113607325 A CN113607325 A CN 113607325A CN 202111176679 A CN202111176679 A CN 202111176679A CN 113607325 A CN113607325 A CN 113607325A
- Authority
- CN
- China
- Prior art keywords
- multivariate
- layer
- stress wave
- steel structure
- improved
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01L—MEASURING FORCE, STRESS, TORQUE, WORK, MECHANICAL POWER, MECHANICAL EFFICIENCY, OR FLUID PRESSURE
- G01L5/00—Apparatus for, or methods of, measuring force, work, mechanical power, or torque, specially adapted for specific purposes
- G01L5/24—Apparatus for, or methods of, measuring force, work, mechanical power, or torque, specially adapted for specific purposes for determining value of torque or twisting moment for tightening a nut or other member which is similarly stressed
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- 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
Abstract
The invention discloses an intelligent monitoring method and system for steel structure bolt group loosening positioning, wherein the method comprises the following steps: collecting multichannel stress wave signals under different working conditions through a piezoelectric sensing monitoring device; preprocessing a stress wave signal; performing phase space reconstruction on the preprocessed stress wave signals according to an improved multivariate recursive graph algorithm to obtain a multivariate recursive graph of corresponding working conditions; training a multi-head attention mechanism improved convolutional neural network model through a multivariable recursive graph of corresponding working conditions; and performing loosening positioning on the steel structure bolt group through a convolution neural network model improved by a multi-head attention mechanism. The invention introduces a recursion analysis method into the field of piezoelectric active sensing, realizes multi-sensor information fusion, combines a multivariable recursion graph with an improved convolution neural network to realize the loosening positioning of a steel structure bolt group, and improves the diagnosis precision of a model.
Description
Technical Field
The invention belongs to the field of steel structure health monitoring, and particularly relates to an intelligent monitoring method and system for positioning supervision of looseness of a steel structure bolt group.
Background
The bolt connection has the advantages of easy construction, detachability, fatigue resistance and the like, and is a main connection mode of a steel structure at present. However, under the action of factors such as cyclic load and complex environment, the bolt connection part of the steel structure is very easy to loosen, and if the bolt loosening part is not found in time and repaired, the structure of the joint is very adversely affected, and even engineering accidents are possibly caused.
The method based on manual detection is time-consuming and labor-consuming and cannot realize real-time monitoring, so that a related research on bolt loosening positioning with high accuracy and good real-time performance needs to be developed by an innovative and breakthrough method. The piezoelectric active sensing monitoring method has the advantages of large monitoring range, quick response and the like, and is widely applied to structural damage identification. At present, most methods for realizing bolt loosening positioning based on a piezoelectric active sensing method and a convolutional neural network need manual feature extraction, the process is generally complicated, and has high uncertainty, so that accurate identification can be achieved by means of abundant feature extraction experience. Therefore, there is a need to find an efficient signal processing method that does not require feature value extraction.
Disclosure of Invention
In view of the above, the invention provides a bolt loosening positioning monitoring method, a bolt loosening positioning monitoring system, equipment and a storage medium of a convolutional neural network improved based on a multi-head attention mechanism, and is used for solving the problem that the loosening condition of a bolt group is difficult to position through conventional indexes such as a signal peak value, wavelet packet energy and the like by a piezoelectric active sensing method.
The invention discloses a steel structure bolt group loosening positioning intelligent monitoring method in a first aspect, which comprises the following steps:
collecting multichannel stress wave signals under different working conditions through a piezoelectric sensing monitoring device;
preprocessing a stress wave signal;
performing phase space reconstruction on the preprocessed stress wave signals according to an improved multivariate recurrence plot algorithm to construct a multivariate recurrence plot of corresponding working conditions;
training a multi-head attention mechanism improved convolutional neural network model through a multivariable recursive graph of corresponding working conditions;
and performing loosening positioning on the steel structure bolt group through a convolution neural network model improved by a multi-head attention mechanism.
Preferably, the preprocessing the original stress wave signal specifically includes:
only intercepting the head wave signal of the stress wave signal, carrying out wavelet denoising and normalization processing on the head wave signal, and discarding the rest stress wave signals.
Preferably, the phase space reconstruction of the preprocessed stress wave signal according to the improved multivariate recursive graph algorithm, and the construction of the multivariate recursive graph of the corresponding working condition specifically includes:
performing multivariate phase space reconstruction on the preprocessed stress wave signals, wherein the expression is as follows:
whereinX i Is the phase point at phase space time point i,Nandnrespectively representing the length of the sample and the number of sensors,x i (k) Is shown askAt the time point of the sensoriTo the corresponding data of the position,k =1,2,…,n；
in the phase spaceXIn (1), the distance threshold is set to a section form [ epsilon ]1,ε2]Constructing a multivariate recurrence plot for the respective operating conditions, elements of the multivariate recurrence plot
The expression of (a) is:
in the formula (I), the compound is shown in the specification,i,j=1,2,…,N，ε1and ε2The lower limit and the upper limit of the threshold value interval are obtained by the optimization of an intelligent optimization algorithm,
the norm is | | · |, the operator Θ isHeavisideA function.
Preferably, the intelligent optimization algorithm is an arithmetic optimization algorithm improved by a golden section operator, and the principle is as follows:
initializing algorithm related parameters and populations;
selecting a search stage by a mathematical optimizer acceleration function MOA (t) to generate a random number r between 0 and 11When r is1>When MOA (t), enter the exploration phase when r1Entering a development stage when the MOA is less than or equal to (t);
introducing a golden section coefficient in an exploration stage, realizing global search through multiplication and division, wherein a position updating formula is as follows:
introducing a golden section operator in a development stage, realizing local development through addition operation and subtraction operation, wherein a position updating formula is as follows:
wherein T is iteration number, T =1,2, …, T, X (T) is individual position at the T iteration, X (T +1) is individual position at the T +1 iteration, Xb(t) is the optimal position at the tth iteration,
and alpha is a sensitive parameter,ξis a minimum value;r 2、r 3、r 4、r 5are all random numbers, and
UB and LB are respectively the upper limit and the lower limit of the preset search range, u is a control parameter;x 1、x 2is a golden section parameter.
Preferably, the fitness function of the intelligent optimization algorithm is as follows:
wherein K is a preset recursion rate index value,Nthe number of phase space state vectors.
Preferably, the convolutional neural network model improved by the multi-head attention mechanism is formed by adding the multi-head attention mechanism on the basis of an AleNet network, and mainly comprises an input layer, an AleNet network layer, a multi-head attention layer, a full connection layer and an output layer; the AleNet network comprises a convolution layer, a local response normalization layer, a maximum pooling layer, a convolution layer, a maximum pooling layer and a convolution layer which are connected in sequence.
Preferably, the processing procedure of the multi-head attention layer is as follows:
converting the output of the AleNet network layer into a form of inquiring Q, a key K and a value V to be used as the input of the multi-head attention layer;
the query Q, the key K and the value V are respectively subjected to linear transformation, and the corresponding transformation matrixes are respectively
、
、
；
Respectively calculating the zooming point product attention based on the linear transformation result;
the output of all scaled dot product attention is combined to obtain the final result.
The invention discloses a steel structure bolt group loosening positioning intelligent monitoring system in a second aspect, which comprises:
a data acquisition module: the piezoelectric sensing monitoring device is used for acquiring multichannel original stress wave signals under different working conditions;
a data processing module: the device is used for preprocessing the stress wave signal; performing phase space reconstruction on the preprocessed stress wave signals according to an improved multivariate recurrence plot algorithm to construct a multivariate recurrence plot of corresponding working conditions;
a state evaluation module: the convolutional neural network model is used for training the multi-head attention mechanism improvement through the multivariable recursion graph of the corresponding working conditions; and performing loosening positioning on the steel structure bolt group through the convolution neural network model improved by the multi-head attention mechanism, and displaying a bolt loosening positioning result through a visual interface.
In a third aspect of the present invention, an electronic device is disclosed, comprising: at least one processor, at least one memory, a communication interface, and a bus;
the processor, the memory and the communication interface complete mutual communication through the bus;
the memory stores program instructions executable by the processor, which program instructions are invoked by the processor to implement the method according to the first aspect of the invention.
In a fourth aspect of the invention, a computer-readable storage medium is disclosed, which stores computer instructions for causing a computer to implement the method of the first aspect of the invention.
Compared with the prior art, the invention has the following beneficial effects:
1) the monitoring system organically combines a data acquisition module based on piezoelectric sensing, a data processing module embedded with an improved multivariable recursion graph algorithm and a state evaluation module embedded with an improved convolutional neural network algorithm to form a set of monitoring system for evaluating the loosening state of the steel structure bolt, can accurately evaluate information such as the loosening position of the bolt in real time, and can visually judge on a visual interface;
2) the invention introduces a recursion analysis method into the field of piezoelectric active sensing, sets a distance threshold interval, optimizes the critical value of the threshold interval by an arithmetic optimization algorithm improved by a golden section operator to enrich the recursion information of a multivariable recursion graph, fuses the nonlinear information of a plurality of piezoelectric signals by the multivariable recursion graph, can conveniently realize the information fusion of a plurality of sensors, extracts the implicit recursion information, better meets the actual requirement of arranging a sensor array in practical engineering application, and improves the accuracy of loosening positioning;
3) the invention provides a method for combining a multivariable recursion diagram with a convolutional neural network to realize the loosening positioning of a steel structure bolt group, and introduces a multi-head attention mechanism on the basis of an AleNet convolutional neural network to further improve the diagnosis precision of the model.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of a monitoring system for positioning bolt group loosening of a convolutional neural network based on improvement of a multi-head attention mechanism according to the present invention;
FIG. 2 is a schematic structural diagram of a piezoelectric sensing monitoring device in the monitoring system of the present invention;
FIG. 3 is a data processing flow diagram of the improved multivariate recurrence plot algorithm proposed by the present invention;
FIG. 4 is a schematic diagram of a convolutional neural network model with multi-head attention fusion according to the present invention;
FIG. 5(A) is an original stress wave signal of a loosening condition according to an embodiment of the present invention;
FIG. 5(B) is a multivariate recursive graph of the original stress wave signal obtained using a conventional recursive graph algorithm;
FIG. 5(C) is a multivariate recurrence plot of the raw stress wave signal improved using the multivariate recurrence plot algorithm of the present invention;
FIG. 6(A) is a stress wave head wave signal captured under a certain loosening condition in an embodiment of the present invention;
FIG. 6(B) is a multivariate recurrence plot of the head wave signal obtained using a conventional recurrence plot algorithm;
FIG. 6(C) is a multivariate recurrence plot of the improved head wave signal using the multivariate recurrence plot algorithm of the present invention;
FIG. 7 is a schematic diagram of a convolutional neural network model training and verification process with multi-head attention fusion according to the present invention;
FIG. 8 is a graph showing the results of a bolt cluster loosening localization test based on a convolutional neural network modified by a multi-head attention mechanism in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Referring to fig. 1, the present invention provides an intelligent monitoring system for steel structure bolt group loosening positioning, which includes: the system comprises a data acquisition module 100, a data processing module 200 and a state evaluation module 300 which are connected in sequence.
The data acquisition module 100 is used for acquiring multichannel original stress wave signals under different working conditions through the piezoelectric sensing monitoring device;
the structure of the piezoelectric sensing monitoring device is shown in fig. 2, and the piezoelectric sensing monitoring device comprises steel plates 1 and 2 connected by a bolt 5, a piezoelectric driver 3, a piezoelectric sensor 4, a power amplifier 8, a multifunctional NI signal acquisition card 9 and a computer terminal 11 provided with labview. The piezoelectric driver 3 and the piezoelectric sensor 4 are adhered to a steel plate connected with a bolt in advance through epoxy resin 6, the piezoelectric sensor 4 is connected to two input ends of a multifunctional NI signal acquisition card 9 through a BNC cable 7 respectively, the piezoelectric driver 3 is connected with a power amplifier 8 through the BNC cable 7, the power amplifier 8 is connected with the output end of the multifunctional NI signal acquisition card 9 through the BNC cable 7, and the multifunctional NI signal acquisition card 9 is connected with a computer analysis terminal 11 through a USB cable 10.
The process of acquiring signals by the piezoelectric active sensing method comprises the following steps: the computer terminal 11 generates a sine sweep frequency excitation electrical signal of a certain frequency band through a labview program, the sine sweep frequency excitation electrical signal is transmitted to the multifunctional NI signal acquisition card 9 through the USB cable 10, the sine sweep frequency excitation electrical signal is amplified by 50 times through the power amplifier 8 and then output to the piezoelectric driver 3, the piezoelectric driver 3 generates stress waves under the action of the excitation signal due to piezoelectric effect, the stress waves are received by an array formed by the piezoelectric sensors 4 after being transmitted in a steel plate and transmitted to the input end of the multifunctional NI signal acquisition card 9 through the BNC cable 7, the multifunctional NI signal acquisition card 9 converts the stress waves into digital signals and transmits the digital signals to the computer terminal 11 through the USB cable 10, and the computer terminal 11 can acquire the piezoelectric signals under different working conditions.
The piezoelectric ceramic sensor used in the piezoelectric sensing monitoring device adopts a compression type ceramic piece. The frequency of the excitation signal is 100Hz-250kHz, the amplitude is 1V, the sampling time is 0.01s, and the sampling frequency is 1 MHz.
In the embodiment, a bolt connecting member composed of four bolts is taken as an example, and a sensing scheme of 'one-sending-two-receiving' is constructed, that is, under each loosening working condition, two piezoelectric original signals can be acquired through the data acquisition module, so that data support is provided for subsequent data processing and information fusion.
The data processing module 200 is used for preprocessing the stress wave signal; and performing phase space reconstruction on the preprocessed stress wave signals according to an embedded improved multivariate recursive graph algorithm to construct a multivariate recursive graph of corresponding working conditions.
The data processing module 200 specifically includes a data preprocessing unit, a recursive graph construction unit, and an intelligent optimization unit.
And the data preprocessing unit is used for intercepting the head wave signals of the stress wave signals, discarding the rest stress wave signals, and respectively carrying out denoising, normalization and other processing on the head wave signals.
The stress wave head wave usually implies rich damage information, and the sequence of the head wave only occupies 1/6 of the complete stress wave model.
The recursive graph building unit is used for building a multivariable recursive graph of corresponding working conditions through the preprocessed stress wave signals;
compared with the conventional single variable, the multivariable recursive graph does not need to solve space construction parameters such as delay time and embedding dimensionality through a mutual information method and a false proximity point method, and can directly carry out phase space reconstruction through a plurality of sensing information.
The expression for phase space reconstruction of the preprocessed stress wave signal is as follows:
whereinX i As phase space time pointsiThe phase point of (a) is (b),Nandnrespectively representing the length of the sample and the number of sensors,x i (k) Is shown askAt the time point of the sensoriTo the corresponding data of the position,k =1,2,…,n；
and in the phase space X, constructing a multivariable recursion graph of the corresponding working condition based on an improved multivariable recursion graph algorithm. The conventional multivariable recursion graph can only set a single distance threshold value, the recursion characteristic of the recursion graph is single, and the distance threshold value is set to be in an interval form [ epsilon ] on the basis of the multivariable recursion graph1,ε2]And further extracting rich recursive information of the multivariate recursive graph. Elements of a constructed multivariate recursive graphR(i,j) The expression of (a) is:
in the formula (I), the compound is shown in the specification,i,j=1,2,…,N，ε1and ε2The lower limit and the upper limit of the threshold interval are obtained by optimizing an intelligent optimization algorithm，
The norm is | | · |, the operator Θ isHeavisideThe function of the function is that of the function,
. The multivariate recursive graph is symmetric about the principal diagonal and can be visualized directly with the distance of the two vectors.
An intelligent optimization unit for optimizing the lower limit epsilon of the threshold interval by an intelligent optimization algorithm1And an upper limit ε2. The intelligent optimization algorithm can be an arithmetic optimization algorithm improved by a golden section operator, and the principle comprises the following steps:
an initialization subunit: the method is used for initializing algorithm related parameters and population, and comprises population scale, search range [ LB, UB ], total iteration times T, minimum Min and maximum Max of an acceleration function and golden sine related parameters, wherein the golden sine related parameters comprise golden section ratio search initial values a and b, a = pi is generally set, b = pi, and then the population position is initialized.
A phase selection subunit: for selecting the search phase by means of a mathematical optimizer acceleration function moa (t),
. Generating a random number r between 0 and 11When r is1>When MOA (t), enter the exploration phase when r1Entering a development stage when the MOA is less than or equal to (t);
an exploration phase subunit: introducing a golden section coefficient in an exploration stage, realizing global search through multiplication and division, wherein a position updating formula is as follows:
development phase subunit: introducing a golden section operator in a development stage, realizing local development through addition operation and subtraction operation, wherein a position updating formula is as follows:
wherein T is iteration number, T =1,2, …, T, X (T) is individual position at the T iteration, X (T +1) is individual position at the T +1 iteration, Xb(t) is the optimal position at the tth iteration,
alpha is a sensitive parameter, MOP (t) gradually decreases with the increase of the iteration number,ξis a minimum value;r 2、r 3、r 4、r 5are all random numbers, and
UB and LB are respectively the upper limit and the lower limit of the preset search range, u is a control parameter;x 1、x 2which is a parameter of the golden section,
golden section ratio
，。
A fitness calculation unit: the fitness function of the intelligent optimization algorithm is as follows:
wherein K is a preset recursion rate index value,Ncalculating the fitness of each individual for the number of the phase space state vectors, and reserving the optimal individual with the minimum fitness.
A termination judgment unit: used for judging whether the end condition is met or not, and if so, outputting the optimal solution as the optimal epsilon1、ε2Otherwise, repeating the stage selection subunit, the exploration stage subunit, the development stage subunit and the fitness calculation unit until the end of the requirementAnd the ending condition is that the maximum iteration times are reached or the fitness value is smaller than the preset precision.
The method combines an arithmetic optimization operator with a golden sine optimization algorithm, and in the process of realizing global search through multiplication and division in the exploration stage of the arithmetic optimization operator, introduces a golden section coefficient by using the thought that the arithmetic optimization operator realizes global position update through high distribution of multiplication and division, further expands the search range near the current optimal position through multiplication, and improves the global optimization capability; meanwhile, in the development stage of the arithmetic optimization operator, in the process of realizing local development through addition operation and subtraction operation, a golden section operator is introduced, the search space is gradually reduced through the golden section ratio so as to lead the individual to approach the optimal solution of the algorithm, the convergence speed of local optimization is accelerated, the balance of exploration and development is realized, and the rapid and accurate optimization is realized.
Referring to fig. 3, the data processing flow of the improved multivariate recurrence plot algorithm embedded in the data processing module 200 of the present invention is to read the collected multichannel stress wave signals, intercept only the first wave signals for use on the stress wave signals, respectively preprocess the first wave signals, then perform the phase space reconstruction by the above formula (1) of the present invention, and form the improved multivariate recurrence plot by the formula (2).
The state evaluation module 300 is used for training a multi-head attention mechanism improved convolutional neural network model through a multivariable recursive graph of corresponding working conditions; and performing loosening positioning on the steel structure bolt group through the convolution neural network model improved by the multi-head attention mechanism, and displaying a bolt loosening positioning result through a visual interface.
The state evaluation module 300 mainly comprises two parts, one is an embedded convolutional neural network pre-training model, and the other is a state evaluation visualization interface. The embedded convolutional neural network model can judge the bolt loosening position, and the visual interface visually shows the structure loosening position in a chart form. The embedded convolutional neural network pre-training model is a convolutional neural network model which is built based on a machine learning platform pytorch and improved based on a multi-head attention mechanism, and is obtained by training based on a pre-collected data set. The state evaluation visualization interface mainly comprises a picture reading unit and a state visualization evaluation unit. A new untrained multivariable recursion graph is read through the picture reading unit, the embedded convolutional neural network pre-training model automatically judges the loosening working condition and determines information such as the loosening position, and the state visualization evaluation unit visually displays the information such as the loosening position through the graph, the table and the like.
The improved convolutional neural network model of the multi-head attention mechanism is a new network added with the multi-head attention mechanism improvement on the basis of AleNet, and mainly comprises an input layer, an AleNet network layer, a multi-head attention layer, a full connection layer and an output layer as shown in figure 4. The AleNet network comprises a convolution layer, a local response normalization layer, a maximum pooling layer, a convolution layer, a maximum pooling layer and a convolution layer which are connected in sequence. Each convolutional layer uses the Same coding, and the activation function uses the ReLU.
The specific structure of each layer is as follows:
1) an Input layer (Input) for unifying the inputted multivariable recursive graph and automatically converting the graph into 227 × 227 × 3;
2) stage conv 1:
convolutional layer (Conv 11 × 11): the input picture size is 227 multiplied by 3, the convolution kernel size is 11 multiplied by 11, the step size is 4, the number is 96, and the activation function uses the ReLU;
local response normalization layer (Local Contrast Norm): the local size is 5, and the final output size is 27 × 27 × 96;
maximum Pooling layer (Max Pooling 2 × 2): the size of an input picture is 55 multiplied by 96, the size of a pooling layer is 2 multiplied by 2, and the step length is 2;
3) stage conv 2:
convolutional layer (Conv 11 × 11): input picture size is 27 × 27 × 96, convolution kernel size: 11 × 11, step size 1, number 256, the activation function uses ReLU;
local response normalization layer (Local Contrast Norm): the local size is 5, and the final output size is 13 × 13 × 256;
maximum Pooling layer (Max Pooling 2 × 2): the input picture size is 7 × 27 × 256, the pooling layer size is 2 × 2, and the step size is 2;
4) stage conv 3:
convolutional layer (Conv 3 × 3): the size of an input picture is 13 multiplied by 384, the size of a convolution kernel is 3 multiplied by 3, the step size is 1, the number is 256, and the ReLU is used as an activation function;
5) stage conv 4:
convolutional layer (Conv 3 × 3): the input picture size is 13 × 13 × 384, the convolution kernel size is 3 × 3, the step size is 1, the number is 256, and the activation function uses ReLU;
6) stage conv 5:
maximum Pooling layer (Max Pooling 2 × 2): the input picture size is 13 × 13 × 256, the pooling layer size is 2 × 2, and the step size is 2; the final output size is 6 × 6 × 256;
convolutional layer (Conv 3 × 3): the input picture size is 13 × 13 × 384, the convolution kernel size is 3 × 3, the step size is 1, the number is 256, and the ReLU is used as the activation function;
7) Multi-Head Attention Layer (Multi-Head Self-Attention Layer)
Multi-headed attention is drawn to using multiple queries Q = [ Q1, …, qM]To select multiple pieces of information from the input information in parallel. Each focus is on a different part of the input information and then spliced. Converting the output of the AleNet network layer into a form of inquiring Q, a key K and a value V to be used as the input of the multi-head attention layer; the query Q, the key K and the value V are respectively subjected to linear transformation, and the corresponding transformation matrixes are respectively
、
、
(ii) a Respectively calculating the zooming point product attention based on the linear transformation result, wherein the expression is as follows:
combining all the scaled dot product attention outputs to obtain a final result, wherein the expression is as follows:
8) full Connection layer (Full Connection): and predicting by adopting a softmax algorithm and outputting an expected result.
The reliability of the monitoring system provided by the invention is further verified by taking the four bolted steel plate structures as objects in combination with a specific experimental process.
As shown in figure 2, a piezoelectric driver and two piezoelectric sensors are pasted on the upper surface of a steel plate, and stress wave signals are collected to carry out bolt loosening positioning intelligent monitoring by taking a certain loosening working condition as an example. Fig. 5(a) is an original stress wave signal under a certain loosening condition, fig. 5(B) and fig. 5(C) are corresponding recursive graphs respectively, wherein fig. 5(B) is a multivariate recursive graph obtained by using a conventional multivariate recursive graph algorithm, and fig. 5(C) is a multivariate recursive graph improved by using the method of the present invention. As shown in fig. 5(a), the acquired original stress wave signal has a data length of 10000, the time for performing phase space reconstruction based on a conventional multivariate recursive graph algorithm is about 30s, and the calculation is time-consuming; as shown in fig. 5(B), the recursion points of the recursion graph extracted by using the conventional recursion graph algorithm based on the original stress wave signal are fewer, and the recursion characteristic is not obvious; the multivariate recursive graph obtained by the improvement of the method of the invention is shown in FIG. 5(C), and the implicit recursive information is rich. Therefore, the improved multivariate recursive graph algorithm can extract implicit recursive information, enrich the recursive information of the multivariate recursive graph, better meet the actual requirement of arranging a sensor array in practical engineering application and improve the accuracy of loosening positioning.
Aiming at the problem that the phase space reconstruction of the original signal and the conventional multivariate recursion graph method are time-consuming in calculation, the invention provides that the head wave is extracted on the basis of the original signal, and then the multivariate recursion graph is solved by using an improved multivariate recursion graph algorithm. Fig. 6(a) is a stress wave head wave signal intercepted under a certain loosening condition in an embodiment of the present invention, and fig. 6(B) and 6(C) are corresponding multivariate recurrence plots, where fig. 6(B) is a multivariate recurrence plot of the head wave signal obtained by using a conventional recurrence plot algorithm, and fig. 6(C) is a multivariate recurrence plot improved by using the method of the present invention. As shown in fig. 6(a), the signal length after the first extraction is only 1500, the time length for performing the phase space reconstruction based on the improved multivariate recursive graph algorithm is about 14s, and the calculation time is greatly reduced. The information of the multivariable recursive graph improved by the method of the invention shown in FIG. 6(C) is obviously richer than that of the conventional recursive graph shown in FIG. 6(B), and a rich basis is provided for information mining of a subsequent convolutional neural network.
In order to train the convolutional neural network provided by the invention better, a connecting component composed of four bolts is taken as an example, single bolt looseness and a plurality of bolt looseness are considered, a permutation and combination mode is adopted, 16 working conditions are set in total, and 100 groups of signals are collected in each working condition. These signals are automatically computed by the data processing module 200 into its multivariate recursion map, thereby constructing a pre-training data set. Of which 80 were used for the training set and 20 were used for the validation set. As shown in fig. 7, when the model training is stable, i.e. the loss function value is close to 0 and does not change, the model training is completed, the model parameters are fixed, and the model is embedded into the state evaluation module, the accuracy of the model training set verification process and the loss function value are shown in fig. 7, and when the model converges, the accuracy of the model training set is as high as about 99%. As shown in fig. 8, based on the model stored after training, the effect of bolt loosening positioning is further measured by using test set data, and the result is presented in the form of a confusion matrix, each row in the confusion matrix represents an actually classified sample, each column represents a sample for prediction classification, the value on the main diagonal represents the number of correctly predicted samples, and since the number of samples in the test set is 20, the correctly predicted samples on the main diagonal show that the trained model of the present invention has higher accuracy. In addition, table 1 further compares the superiority of the method for extracting a head wave + improved multivariate recurrence plot proposed by the present invention from the aspects of both average recognition accuracy and recognition efficiency. The method for extracting the head wave + the improved multivariate recursive graph has the best identification effect, and the identification time of a single recursive graph is only 9s, so that the method has better real-time performance.
TABLE 1 comparison of average recognition accuracy and recognition efficiency for different methods
| Method | Average classification accuracy | Identifying single recursive graph time |
| Extraction of head wave signal + improved multivariate recursion map | 97.31% | 9s |
| Original stress wave signal + improved variable recurrence plot | 96.87% | 22s |
| Original stress wave signal + conventional multivariate recurrence plot | 95.05% | 25s |
The invention discloses a method for predicting bolt looseness by an intelligent monitoring system for positioning the looseness of a steel structure bolt group, which comprises the following steps:
(1) acquiring data through a data acquisition module 100, and calculating through a data processing module to obtain a pre-trained multivariate recursion graph data set (a training set and a verification set);
(2) using the data set to train the improved convolutional neural network model based on the multi-head attention mechanism in advance, and implanting the improved convolutional neural network model into the state evaluation module 300;
(3) when a certain bolt of the bolt connection structure is loosened, the data acquisition module 100 is used for acquiring the dual-channel piezoelectric original signal under the working condition;
(4) after the sensor and the data acquisition module acquire data, the data processing module 200 is activated, two piezoelectric original signals acquired by the data acquisition module 100 are automatically read through the matlab platform, and a multivariable recursion graph of the data acquisition module is calculated based on an embedded multivariable recursion graph algorithm;
(5) after the calculation of the multi-variable recursion map is completed, the state evaluation module 300 is activated, and the multi-variable recursion map is automatically read and basic information determined by the model is displayed through a visual interface, so that bolt loosening positioning prediction is realized.
Corresponding to the embodiment of the system, the invention also provides an intelligent monitoring method for the looseness positioning of the steel structure bolt group, which comprises the following steps:
and S1, acquiring multichannel stress wave signals under different working conditions through the piezoelectric sensing monitoring device.
S2, preprocessing the stress wave signal; the method comprises the steps of carrying out wavelet denoising and normalization processing on collected stress wave original signals, only intercepting the head wave of the stress wave signals, and discarding the rest stress wave signals.
And S3, performing phase space reconstruction on the preprocessed stress wave signals according to a multivariate recursive graph algorithm, and constructing a multivariate recursive graph of corresponding working conditions.
Specifically, multivariate phase space reconstruction is performed on the preprocessed stress wave signals, and the expression is as follows:
whereinX i As phase space time pointsiThe phase point of (a) is (b),Nandnrespectively representing the sampling length and the sensorThe number of the components is equal to or less than the total number of the components,x i (k) Is shown askAt the time point of the sensoriTo the corresponding data of the position,k =1,2,…,n；
in the phase spaceXIn (1), the threshold is set to a section form [ epsilon ]1,ε2]Constructing a multivariate recurrence plot for the respective operating conditions, elements of the multivariate recurrence plot
The expression of (a) is:
in the formula (I), the compound is shown in the specification,i,j=1,2,…,N，ε1and ε2The lower limit and the upper limit of the threshold value interval are obtained by the optimization of an intelligent optimization algorithm,
the norm is | | · |, the operator Θ isHeavisideA function.
And S4, training the improved convolutional neural network model of the multi-head attention mechanism through the multivariate recursive graph of the corresponding working conditions.
Specifically, the convolutional neural network model improved by the multi-head attention mechanism is formed by adding the multi-head attention mechanism on the basis of an AleNet network, and mainly comprises an input layer, an AleNet network layer, a multi-head attention layer, a full connection layer and an output layer; the AleNet network comprises a convolution layer, a local response normalization layer, a maximum pooling layer, a convolution layer, a maximum pooling layer and a convolution layer which are connected in sequence.
And S5, performing loosening positioning on the steel structure bolt group through the convolution neural network model improved by the multi-head attention mechanism.
Specifically, a multichannel stress wave signal to be detected is obtained and preprocessed; constructing a multivariable recursive graph to be tested through the preprocessed stress wave signals; and inputting the multivariable recursive graph to be tested into a convolutional neural network model improved by a multi-head attention mechanism, and outputting a judgment result of the bolt loosening position.
The above method embodiment corresponds to the system embodiment, and the brief description of the method real-time embodiment is only required to refer to the system embodiment.
The present invention also discloses an electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus; the processor, the memory and the communication interface complete mutual communication through the bus; the memory stores program instructions executable by the processor, which invokes the program instructions to implement the methods of the invention described above.
The invention also discloses a computer readable storage medium which stores computer instructions for causing the computer to implement all or part of the steps of the method of the embodiment of the invention. The storage medium includes: u disk, removable hard disk, ROM, RAM, magnetic disk or optical disk, etc.
The above-described system embodiments are merely illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts shown as units may or may not be physical units, i.e. may be distributed over a plurality of network units. Without creative labor, a person skilled in the art can select some or all of the modules according to actual needs to achieve the purpose of the solution of the embodiment.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. The intelligent monitoring method for the looseness positioning of the steel structure bolt group is characterized by comprising the following steps of:
collecting multichannel stress wave signals under different working conditions through a piezoelectric sensing monitoring device;
preprocessing a stress wave signal;
performing phase space reconstruction on the preprocessed stress wave signals according to an improved multivariate recurrence plot algorithm to construct a multivariate recurrence plot of corresponding working conditions;
training a multi-head attention mechanism improved convolutional neural network model through a multivariable recursive graph of corresponding working conditions;
and performing loosening positioning on the steel structure bolt group through a convolution neural network model improved by a multi-head attention mechanism.
2. The intelligent monitoring method for the loosening positioning of the steel structure bolt group according to claim 1, wherein the preprocessing of the original stress wave signal specifically comprises:
and only intercepting the head wave signal of the stress wave signal, and performing wavelet denoising and normalization processing on the head wave signal.
3. The intelligent monitoring method for the looseness positioning of the steel structure bolt group according to claim 2, wherein the phase space reconstruction is performed on the preprocessed stress wave signals according to an improved multivariate recurrence plot algorithm, and the construction of the multivariate recurrence plot of the corresponding working conditions specifically comprises the following steps:
performing multivariate phase space reconstruction on the preprocessed stress wave signals, wherein the expression is as follows:
whereinX i As phase space time pointsiThe phase point of (a) is (b),Nandnrespectively representing the length of the sample and the number of sensors,x i (k) Is shown askAt the time point of the sensoriTo the corresponding data of the position,k =1,2,…,n；
in the phase spaceXIn (1), the distance threshold is set to a section form [ epsilon ]1,ε2]Constructing a multivariate recurrence plot for the respective operating conditions, elements of the multivariate recurrence plotR(i,j) The expression of (a) is:
in the formula (I), the compound is shown in the specification,i,j=1,2,…,N，ε1and ε2Respectively the lower limit and the upper limit of the threshold interval, are obtained by the optimization of an intelligent optimization algorithm,
the norm is | | · |, the operator Θ isHeavisideA function.
4. The intelligent monitoring method for the looseness positioning of the steel structure bolt group according to claim 3, wherein the intelligent optimization algorithm is an arithmetic optimization algorithm improved through a golden section operator, and the principle is as follows:
initializing algorithm related parameters and populations;
selecting a search stage by a mathematical optimizer acceleration function MOA (t) to generate a random number r between 0 and 11When r is1>When MOA (t), enter the exploration phase when r1Entering a development stage when the MOA is less than or equal to (t);
introducing a golden section coefficient in an exploration stage, realizing global search through multiplication and division, wherein a position updating formula is as follows:
introducing a golden section operator in a development stage, realizing local development through addition operation and subtraction operation, wherein a position updating formula is as follows:
wherein T is iteration number, T =1,2, …, T, X (T) is individual position at the T iteration, X (T +1) is individual position at the T +1 iteration, Xb(t) is the optimal position at the tth iteration,
and alpha is a sensitive parameter,ξis a minimum value;r 2、r 3、r 4、r 5are all random numbers, and
UB and LB are respectively the upper limit and the lower limit of the preset search range, u is a control parameter;x 1、x 2is a golden section parameter.
5. The intelligent monitoring method for the loosening positioning of the steel structure bolt group according to claim 4, wherein the fitness function of the intelligent optimization algorithm is as follows:
wherein K is a preset recursion rate index value,Nthe number of phase space state vectors.
6. The intelligent monitoring method for the looseness positioning of the steel structure bolt group according to claim 1, wherein the convolution neural network model improved by the multi-head attention mechanism is a convolution neural network model formed by adding the multi-head attention mechanism on the basis of an AleNet network, and mainly comprises an input layer, an AleNet network layer, a multi-head attention layer, a full connection layer and an output layer; the AleNet network comprises a convolution layer, a local response normalization layer, a maximum pooling layer, a convolution layer, a maximum pooling layer and a convolution layer which are connected in sequence.
7. The intelligent monitoring method for the loosening positioning of the steel structure bolt group according to claim 6, wherein the processing process of the multi-head attention layer is as follows:
converting the output of the AleNet network layer into a form of inquiring Q, a key K and a value V to be used as the input of the multi-head attention layer;
the query Q, the key K and the value V are respectively subjected to linear transformation, and the corresponding transformation matrixes are respectively
、
、
；
Respectively calculating the zooming point product attention based on the linear transformation result;
the output of all scaled dot product attention is combined to obtain the final result.
8. The utility model provides a not hard up location intelligent monitoring system of steel construction bolt crowd, its characterized in that, the system includes:
a data acquisition module: the device is used for collecting multi-channel stress wave signals under different working conditions through the piezoelectric sensing monitoring device;
a data processing module: the device is used for preprocessing the stress wave signal; performing phase space reconstruction on the preprocessed stress wave signals according to an improved multivariate recurrence plot algorithm to construct a multivariate recurrence plot of corresponding working conditions;
a state evaluation module: the convolutional neural network model is used for training the multi-head attention mechanism improvement through the multivariable recursion graph of the corresponding working conditions; and performing loosening positioning on the steel structure bolt group through the convolution neural network model improved by the multi-head attention mechanism, and displaying a bolt loosening positioning result through a visual interface.
9. An electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus;
the processor, the memory and the communication interface complete mutual communication through the bus;
the memory stores program instructions executable by the processor, the processor invoking the program instructions to implement the method of any one of claims 1-7.
10. A computer-readable storage medium storing computer instructions for causing a computer to implement the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111176679.6A CN113607325B (en) | 2021-10-09 | 2021-10-09 | Intelligent monitoring method and system for looseness positioning of steel structure bolt group |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111176679.6A CN113607325B (en) | 2021-10-09 | 2021-10-09 | Intelligent monitoring method and system for looseness positioning of steel structure bolt group |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113607325A true CN113607325A (en) | 2021-11-05 |
| CN113607325B CN113607325B (en) | 2021-12-14 |
Family
ID=78343445
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111176679.6A Active CN113607325B (en) | 2021-10-09 | 2021-10-09 | Intelligent monitoring method and system for looseness positioning of steel structure bolt group |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113607325B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114266280A (en) * | 2021-12-29 | 2022-04-01 | 西安交通大学 | Bolt loosening state identification method based on improved convolutional neural network |
| CN114358091A (en) * | 2022-03-03 | 2022-04-15 | 中山大学 | Pile damage identification method, equipment and medium based on convolutional neural network |
| CN114492146A (en) * | 2022-04-02 | 2022-05-13 | 武汉地震工程研究院有限公司 | Bolt group loosening positioning and quantitative analysis method and system based on transfer learning |
| CN115060184A (en) * | 2022-05-18 | 2022-09-16 | 武汉迪信达科技有限公司 | Optical fiber perimeter intrusion detection method and system based on recursive graph |
| CN116659826A (en) * | 2022-08-23 | 2023-08-29 | 国家电投集团科学技术研究院有限公司 | Method and device for detecting state of wind power bolt |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503799A (en) * | 2016-10-11 | 2017-03-15 | 天津大学 | Deep learning model and the application in brain status monitoring based on multiple dimensioned network |
| US20200034436A1 (en) * | 2018-07-26 | 2020-01-30 | Google Llc | Machine translation using neural network models |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN111709233A (en) * | 2020-05-27 | 2020-09-25 | 西安交通大学 | Intelligent diagnosis guiding method and system based on multi-attention convolutional neural network |
| CN111755019A (en) * | 2019-03-28 | 2020-10-09 | 三星电子株式会社 | System and method for acoustic echo cancellation using deep multitask recurrent neural networks |
| US20200364576A1 (en) * | 2019-05-14 | 2020-11-19 | Adobe Inc. | Utilizing deep recurrent neural networks with layer-wise attention for punctuation restoration |
| CN112001110A (en) * | 2019-05-27 | 2020-11-27 | 浙江大学 | Structural damage identification monitoring method based on vibration signal space real-time recursive graph convolutional neural network |
| CN112396109A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Motor bearing fault diagnosis method based on recursion graph and multilayer convolution neural network |
| CN112668526A (en) * | 2020-12-31 | 2021-04-16 | 武汉地震工程研究院有限公司 | Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing |
| CN112816122A (en) * | 2020-12-31 | 2021-05-18 | 武汉地震工程研究院有限公司 | Bolt tightness degree monitoring method based on deep learning and piezoelectric active sensing |
| CN113358993A (en) * | 2021-05-13 | 2021-09-07 | 武汉大学 | Online fault diagnosis method and system for multi-level converter IGBT |
| CN113380346A (en) * | 2021-06-08 | 2021-09-10 | 河南大学 | Coupling reaction yield intelligent prediction method based on attention convolution neural network |
| CN113378243A (en) * | 2021-07-14 | 2021-09-10 | 南京信息工程大学 | Personalized federal learning method based on multi-head attention mechanism |
-
2021
- 2021-10-09 CN CN202111176679.6A patent/CN113607325B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503799A (en) * | 2016-10-11 | 2017-03-15 | 天津大学 | Deep learning model and the application in brain status monitoring based on multiple dimensioned network |
| US20200034436A1 (en) * | 2018-07-26 | 2020-01-30 | Google Llc | Machine translation using neural network models |
| CN111755019A (en) * | 2019-03-28 | 2020-10-09 | 三星电子株式会社 | System and method for acoustic echo cancellation using deep multitask recurrent neural networks |
| US20200364576A1 (en) * | 2019-05-14 | 2020-11-19 | Adobe Inc. | Utilizing deep recurrent neural networks with layer-wise attention for punctuation restoration |
| CN112001110A (en) * | 2019-05-27 | 2020-11-27 | 浙江大学 | Structural damage identification monitoring method based on vibration signal space real-time recursive graph convolutional neural network |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN111709233A (en) * | 2020-05-27 | 2020-09-25 | 西安交通大学 | Intelligent diagnosis guiding method and system based on multi-attention convolutional neural network |
| CN112396109A (en) * | 2020-11-19 | 2021-02-23 | 天津大学 | Motor bearing fault diagnosis method based on recursion graph and multilayer convolution neural network |
| CN112668526A (en) * | 2020-12-31 | 2021-04-16 | 武汉地震工程研究院有限公司 | Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing |
| CN112816122A (en) * | 2020-12-31 | 2021-05-18 | 武汉地震工程研究院有限公司 | Bolt tightness degree monitoring method based on deep learning and piezoelectric active sensing |
| CN113358993A (en) * | 2021-05-13 | 2021-09-07 | 武汉大学 | Online fault diagnosis method and system for multi-level converter IGBT |
| CN113380346A (en) * | 2021-06-08 | 2021-09-10 | 河南大学 | Coupling reaction yield intelligent prediction method based on attention convolution neural network |
| CN113378243A (en) * | 2021-07-14 | 2021-09-10 | 南京信息工程大学 | Personalized federal learning method based on multi-head attention mechanism |
Non-Patent Citations (3)
| Title |
|---|
| RASSIL, A 等: "The importance of local labels distribution and dominance for node classification in graph neural networks", 《2020 19TH IEEE INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND APPLICATIONS (ICMLA)》 * |
| 李宗源 等: "基于压电主动传感法的钢管螺纹接头松动监测", 《中国测试》 * |
| 梁振隆: "基于递归图和卷积神经网络的结构损伤识别研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114266280A (en) * | 2021-12-29 | 2022-04-01 | 西安交通大学 | Bolt loosening state identification method based on improved convolutional neural network |
| CN114266280B (en) * | 2021-12-29 | 2023-10-13 | 西安交通大学 | Bolt loosening state identification method based on improved convolutional neural network |
| CN114358091A (en) * | 2022-03-03 | 2022-04-15 | 中山大学 | Pile damage identification method, equipment and medium based on convolutional neural network |
| CN114358091B (en) * | 2022-03-03 | 2022-06-10 | 中山大学 | Pile damage identification method, equipment and medium based on convolutional neural network |
| CN114492146A (en) * | 2022-04-02 | 2022-05-13 | 武汉地震工程研究院有限公司 | Bolt group loosening positioning and quantitative analysis method and system based on transfer learning |
| CN115060184A (en) * | 2022-05-18 | 2022-09-16 | 武汉迪信达科技有限公司 | Optical fiber perimeter intrusion detection method and system based on recursive graph |
| CN116659826A (en) * | 2022-08-23 | 2023-08-29 | 国家电投集团科学技术研究院有限公司 | Method and device for detecting state of wind power bolt |
| CN116659826B (en) * | 2022-08-23 | 2024-02-06 | 国家电投集团科学技术研究院有限公司 | Method and device for detecting state of wind power bolt |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113607325B (en) | 2021-12-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113607325B (en) | Intelligent monitoring method and system for looseness positioning of steel structure bolt group | |
| CN103544392B (en) | Medical science Gas Distinguishing Method based on degree of depth study | |
| CN103648106B (en) | WiFi indoor positioning method of semi-supervised manifold learning based on category matching | |
| CN111562612B (en) | Deep learning microseismic event identification method and system based on attention mechanism | |
| CN111580151B (en) | SSNet model-based earthquake event time-of-arrival identification method | |
| CN112507915A (en) | Method for identifying loosening state of bolt connection structure based on vibration response information | |
| CN110440148B (en) | Method, device and system for classifying and identifying leakage acoustic signals | |
| CN111638249B (en) | Water content measuring method based on deep learning and application of water content measuring method in oil well exploitation | |
| CN102262211A (en) | Analog circuit failure diagnosis method based on dynamic classification of echo state network | |
| CN112668526A (en) | Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing | |
| CN115343676B (en) | Feature optimization method for positioning technology of redundant substances in sealed electronic equipment | |
| CN111721450B (en) | Impact positioning recognition device and method | |
| CN111898644A (en) | Intelligent identification method for health state of aerospace liquid engine under fault-free sample | |
| CN111753776B (en) | Structural damage identification method based on echo state and multi-scale convolution combined model | |
| CN114974306A (en) | Transformer abnormal voiceprint detection and identification method and device based on deep learning | |
| CN113533511B (en) | Steel rail weld joint monitoring method based on deep learning network model | |
| CN114492146B (en) | Bolt group loosening positioning and quantitative analysis method and system based on transfer learning | |
| CN101814146B (en) | Method for extracting action potential feature based on multiple wavelet transformation | |
| CN105426923A (en) | Semi-supervised classification method and system | |
| Alazzawi et al. | Deep convolution neural network for damage identifications based on time-domain PZT impedance technique | |
| CN111626324A (en) | Seabed observation network data heterogeneous analysis integration method based on edge calculation | |
| CN116068520A (en) | Cognitive radar joint modulation recognition and parameter estimation method based on transducer | |
| CN113973403B (en) | Temperature-induced strain field redistribution intelligent sensing method based on structure discrete measurement point topology | |
| CN112816122B (en) | Bolt tightness degree monitoring method based on deep learning and piezoelectric active sensing | |
| CN114266280A (en) | Bolt loosening state identification method based on improved convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |