CN111582634B - Multi-factor safety grading method and system for underground large-space construction - Google Patents
Multi-factor safety grading method and system for underground large-space construction Download PDFInfo
- Publication number
- CN111582634B CN111582634B CN202010225494.9A CN202010225494A CN111582634B CN 111582634 B CN111582634 B CN 111582634B CN 202010225494 A CN202010225494 A CN 202010225494A CN 111582634 B CN111582634 B CN 111582634B
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- security
- construction
- safety
- 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.)
- Active
Links
- 238000010276 construction Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013528 artificial neural network Methods 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 26
- 238000012544 monitoring process Methods 0.000 claims description 20
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000002474 experimental method Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 10
- 230000007246 mechanism Effects 0.000 abstract description 7
- 239000002689 soil Substances 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 description 10
- 230000005284 excitation Effects 0.000 description 7
- 230000015572 biosynthetic process Effects 0.000 description 6
- 238000003786 synthesis reaction Methods 0.000 description 6
- 238000003062 neural network model Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 210000002364 input neuron Anatomy 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a multi-factor security grading method and system for underground large space construction, comprising the following steps: establishing a multi-factor security grading comparison table for underground large-space construction; collecting data samples associated with each security level according to the comparison table; constructing a security classification BP neural network, training the security classification BP neural network by using the data samples associated with each security level, and enabling the security classification BP neural network to meet the following conditions: when the input layer has parameter input, the output layer automatically outputs a security level judging result; and inputting the data samples acquired in real time in construction into the safety grading BP neural network to predict the construction safety grade in real time. According to the invention, the construction safety grading model is established by establishing the corresponding safety grading comparison table and adopting the BP neural network, so that the safety grading evaluation can be carried out on the underground large-space construction before the construction without depending on the internal working mechanism of the rock-soil system, and the real-time prediction alarm can be carried out in the construction.
Description
Technical Field
The invention relates to the technical field of civil engineering construction, in particular to a multi-factor safety grading method and system for underground large-space construction.
Background
The existing monitoring systems aiming at the complex environment of urban underground large space construction are independent, the false alarm rate of the monitoring data is high, the analysis method is single-index evaluation, the consideration is incomplete, and a complete safety state grading early warning system cannot be formed, so that a standard aiming at multi-index grading evaluation of the safety state of urban underground large space construction is urgently needed, and powerful guarantee is provided for the safety of underground large space construction in China.
In addition, geotechnical engineering practice shows that engineering technicians can only measure and grasp the appearance of geotechnical, such as the pressure of a bearing plate, ground subsidence and the like, and the working mechanism of the geotechnical engineering practice is not completely known to a certain extent. Thus, in describing these external objective representations, there must be great difficulty in establishing the corresponding analytical expressions due to the unclear underlying working mechanisms of the geotechnical system. The representation of this phenomenon can be expressed in terms of a "black box" system, where one knows only the inputs and outputs of the system, and not the internal mechanisms of operation of the system, and the key to the problem is how to reproduce the macroscopic appearance of the system. In addition, as the relationship between the input and the interference factors of the geotechnical engineering system and the output is quite complex, the nonlinear characteristics of the geotechnical engineering system are very complex, the nonlinear description is a problem of suspension and no solution, and the difficulty of geotechnical engineering characteristic research is seen. Because of lack of cognition on the intrinsic working mechanism of the geotechnical system, the existing main adoption of artificial safety assessment aiming at multiple indexes of the safety state in construction has great errors and contingencies, and also has the problems of artificial false report data and great potential safety hazards in construction.
The neural network is particularly suitable for processing various nonlinear problems due to the characteristics of strong adaptivity, nonlinearity, fault tolerance and the like. It can extract the causal relationship implicit in the sample through the learning of a large number of samples. Therefore, the neural network provides a research thought completely different from mathematical modeling for the field of underground engineering, avoids a complex constitutive model, and becomes an effective way for solving the problem of underground engineering.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a multi-factor safety grading method and a system for underground large space construction, which are characterized in that a corresponding safety grading comparison table is established by researching multi-index grading evaluation of the safety state of urban underground large space construction, and a construction safety grading model is established by adopting a neural network with self-adaptability, nonlinearity and strong fault tolerance, so that the safety grading evaluation can be carried out on the underground large space construction in the construction process without depending on the internal working mechanism of a rock-soil system, thereby carrying out real-time prediction alarm in the construction.
In order to achieve the above object, the present invention adopts a technical scheme including the following aspects.
A multi-factor safety grading method for underground large space construction comprises the following steps:
step 101, determining a construction monitoring index and a threshold value thereof according to the actual conditions of engineering; establishing a multi-factor safety grading comparison table for underground large-space construction according to the construction monitoring index and the threshold value thereof;
102, randomly generating data samples associated with each security level according to the underground large space construction multi-factor security grading comparison table;
step 103, constructing a security classification neural network, training the security classification neural network by using the data samples associated with each security level to establish a nonlinear mapping relationship from an input layer to an output layer of the security classification neural network, and enabling the security classification neural network to satisfy the following conditions: when the input layer has parameter input, the output layer automatically outputs a security level judging result;
and 104, inputting the data samples acquired in real time in construction into the safety grading neural network to predict the construction safety level in real time.
Preferably, the determining the construction monitoring index and the threshold value thereof according to the actual engineering situation includes: existing indicators are selected, and proportional threshold reduction is performed on the selected existing indicators.
Preferably, the underground large space construction multi-factor safety grading comparison table comprises the safety level of a single monitoring index factor and the safety level of a comprehensive multi-monitoring index factor.
Preferably, the input parameters of the security hierarchical neural network input layer include: the support pile or the support wall top is vertically displaced, the support pile or the support wall top is horizontally displaced, the support pile or the support wall body is horizontally displaced, the upright post structure is vertically displaced, the earth surface subsides, the support wall structure stress, the upright post structure stress, the support shaft force and the anchor rod tension force.
Preferably, the total number of adjustable connection weights of the security hierarchical neural network is 56.
Preferably, a Sigmoid function is used as the error function of the security hierarchical neural network.
Preferably, the maximum iteration number of the safety grading neural network is 5000, and the learning rate eta is 0.5.
Preferably, when the error rate of the safety classification neural network is smaller than a preset value, the performance of the safety classification neural network is judged to be stable, and training is stopped.
In a further embodiment of the present invention, there is provided a multi-factor security prediction system for underground large space construction, comprising at least one processor, and a memory communicatively coupled to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the above-described multi-factor safety prediction method for underground large space construction.
In summary, due to the adoption of the technical scheme, the invention has at least the following beneficial effects:
by researching multi-index grading evaluation of the safety state of urban underground large space construction, establishing a corresponding safety grading comparison table, and establishing a construction safety grading model based on a neural network with strong adaptivity, nonlinearity and fault tolerance, the safety grading evaluation of underground large space construction can be carried out in the construction process without depending on the internal working mechanism of a rock-soil system, so that real-time prediction alarm is carried out in the construction. And through the automatic safety level identification of the neural network model, the construction safety prediction efficiency is improved, human intervention and false report data are avoided, and the potential safety hazard existing in construction is greatly reduced.
Drawings
Fig. 1 is a multi-factor security level hierarchy model diagram according to an exemplary embodiment of the present invention.
Fig. 2 is a network topology structure diagram of a neural network model according to an exemplary embodiment of the present invention.
Fig. 3 is a schematic structural view of a multi-factor safety prediction system for underground large space construction according to an exemplary embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, so that the objects, technical solutions and advantages of the present invention will become more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Example 1
FIG. 1 illustrates an exemplary multi-factor security grading method for underground large space construction of the present invention, comprising:
step 1011, determining a construction monitoring index and a threshold thereof, including two parts:
1. existing indicators are selected. Reasonable and available control indexes can be selected by integrating national standards and local standards, and when the indexes are determined, all risk factor control indexes conforming to the engineering are required to be proposed according to the actual engineering profile.
2. The existing indexes are reduced (aiming at special conditions such as large spans, and the like). Numerical experiments and model experiments are carried out, and the reduction coefficient is determined through a certain field experiment.
The specific monitoring index and the threshold value need to be comprehensively considered according to actual engineering.
Step 1012, the security level determining method is divided into two steps to finally determine:
1. and establishing single-factor security level division according to each monitoring index. The monitoring accumulated value and the change rate value are important indexes for measuring the safety state of the foundation pit by the monitoring index, and the standard specified accumulated alarm value and the standard specified change rate alarm value (special conditions such as large spans and the like need numerical experiments and model experiments, and a reduction coefficient is determined through a certain field experiment) are used as grading control standards of the safety state. The security level was classified into five levels according to the severity of dangerous occurrence as shown in table 1.
TABLE 1 grading Standard for Foundation pit Single factor evaluation index
2. And carrying out multi-factor security level division based on the single-factor security level division. The multi-factor security level is determined by an analytic hierarchy process whose idea is to decompose a complex problem into constituent elements and group the factors in a dominant relationship, thereby forming an ordered hierarchical structure. And comparing the elements of each layer in pairs according to a certain rule in each layer to form a matrix, calculating the weight of the relative importance sequence of each element of the layer to the criterion and the combined weight of each element of the layer to the overall target by using a mathematical method, sequencing, and analyzing and deciding the problem by using the sequencing result. The multi-index underground engineering construction safety level division is described below by taking the control values of foundation pit supporting structures and surrounding rock-soil body monitoring projects of the open cut method and the cover cut method as an example in combination with the technical Specification for monitoring urban rail transit engineering.
Firstly, establishing a hierarchical structure, classifying factors related to a problem, constructing a hierarchical structure model of interconnection among elements, wherein the uppermost hierarchy is a preset target of the problem, and generally only elements in an element middle layer are generally a criterion layer and the bottommost layer of a sub-criterion layer is generally a decision scheme. On the basis of the deep analysis specification, analyzing factors and interrelationships thereof contained in the problem, and decomposing each related factor into a plurality of layers from top to bottom according to different attributes. The multi-factor security level hierarchy model is shown in fig. 1.
Secondly, constructing a pairwise comparison judgment matrix (according to specific construction conditions, selecting (1) a fuzzy statistics method (2) an expert investigation method (3) a distribution method (4) a binary comparison ordering method (5) to ask experienced experts or engineering technicians to directly mark and the like for determination), taking a target layer A and n indexes B1, B2 and … … Bn of the next layer related to the target layer A as examples, and establishing a corresponding judgment matrix form as shown in table 2.
TABLE 2 determination matrix form
Wherein b ij Representation of B for A i Pair B j The numerical value of the relative importance is represented. General b ij 1,2, …,9 and their reciprocal can be taken as the scale [ i ]][ii]The meanings are given in Table 3.
Again, the hierarchical single ranking is the ranking of the priority order of the indexes related to the present hierarchy for the index of the previous hierarchy according to the importance matrix. The method can be summarized as the problem of calculating the characteristic root and the characteristic vector of the judgment matrix, wherein the judgment matrix is the evaluation data for comparing the index related to the index of the previous layer with each other. I.e. for the decision matrix B, its maximum eigenvalue amax, and the corresponding eigenvector W, are calculated such that bw=λmaxw. λmax and W typically use a method of approximation calculation.
Table 3 importance scale
Then, the total rank order of the layers uses the results of all the single ranks of the layers in the same layer to calculate the importance weights of all the indexes of the layer for the previous layer, which is the total rank order of the layers. The total order of the layers is needed to be carried out layer by layer from top to bottom, and for the second layer below the highest layer, the single order of the layers is the total order. The total ordering of the layers needs to be performed layer by layer from top to bottom. For the highest layer, its hierarchical single order is the hierarchical total order.
Finally, judging the consistency of the matrix, in practical problems, due to the complexity of things and the limitation of people in judging the problem, when comparing every two, the judgment matrix is difficult to have strict consistency, but the consistency is required to be approximate. Therefore, after λmax is calculated, the consistency of the judgment matrix is also checked. Reference is made to the calculation of the consistency index c.i. equation (1) and the calculation of the average random consistency index correction value r.i. determination method (table 4).
TABLE 4 average random uniformity index correction values
Through the above analytic hierarchy process calculation, the sequential weight of each decision scheme relative to the total target can be finally obtained, and decisions can be made according to the sequential weights.
102, randomly generating data samples associated with each security level according to the underground large space construction multi-factor security grading comparison table; after the security level determining method is determined, n groups of data samples are randomly generated, namely a multi-factor security grading comparison table (specific grading samples are affected by expert evaluation in advance) for the construction of the large space under security can be established as shown in table 5, and the level of the randomly generated data is determined by the grading method.
TABLE 5 Multi-factor safety grading comparison Table for safety underground large space construction
Step 103, constructing a security classification neural network, training the security classification neural network by using the data samples associated with the security classes to establish a nonlinear mapping relationship from an input layer to an output layer of the security classification neural network, and enabling the security classification neural network to satisfy the following conditions: when the input layer has parameter input, the output layer automatically outputs a security level judging result;
specifically, n level samples in table 5 are used as learning samples of the neural network, and a security grading neural network model is constructed and trained. The safety grading neural network model network identification model comprises a hidden layer, wherein the input layer, the hidden layer and the output layer respectively consist of 9 input neurons, 4 hidden units and 5 output neurons. The model structure is shown in fig. 2. When the system performs safe hierarchical network learning, related hierarchical model data which is determined as a learning sample in a database is automatically invoked.
The security hierarchical neural network determines the level according to the randomly generated data through an analytic hierarchy process and then takes the level as a sample, takes the randomly generated data of the monitoring index as an input parameter, and outputs the corresponding level after network calculation. The input parameters of the security grading network mainly comprise nine (the number of which can be corrected later), i.e. the vertical displacement of the top of the support pile (wall),the support pile (wall) is horizontally displaced, the upright post structure is vertically displaced, the earth surface subsides, the support wall structure stress, the upright post structure stress, the support axial force and the anchor rod pulling force are supported. Correspondingly, the output parameters of the security hierarchical neural network are 5, namely output levels. Thus, in this embodiment, the number of units of the actual input layer of the constructed security hierarchical neural network is 9, and the number of units of the output layer is 5. Three layers of neural networks with hidden layers are adopted between the input layer and the output layer, and the number of hidden layer units is as follows: n is n 1 ＝log 2 9, take n 1 ＝4。
As shown in fig. 2, the security hierarchical neural network has a total of 3 layers: the input layer is the 1 st layer, the hidden layer is the 2 nd layer, and the output layer is the 3 rd layer. The total number of the connection weights w is 56, and the connection weights w are expressed as the following by symbolsConnection weight of the ith element of the k-1 layer to the jth element of the k layer, and an array can be used for programming):
layer 1-2 (36 connection weights altogether):
element 1:
element 2:
element 3:
element 4:
element 5:
element 6:
element 7:
element 8:
element 9:
expressed by an array w1[9] [4 ]: dimension 1 is the number of elements (9) of the first layer (input layer), each element having 3 connection weights with the elements (4) of the second layer; dimension 2 is the second layer element (4); the number of connection weights is 36.
Layer 2-3 (total 20 connection weights):
element 1:
element 2:
element 3:
element 4:
expressed by an array w2[4] [5 ]: dimension 1 is the number of elements of the second layer (4), each element having 5 connection weights with the third layer (output layer) element (5); dimension 2 is the third layer element (5); the number of the connection weights is 20. The total number of adjustable connection weights W of the neural network adopted in the example is 56. The error preset minimum value is set, training is terminated when the actual output error is smaller than the given error, and if the network model meets the learning sample requirement or does not meet the learning sample requirement, the network model is required to continue learning.
Specifically, the specific parameter settings of the security hierarchical neural network in this example mainly include: the total learning times of the network are the total number of samples, namely the total circulation times; setting the maximum iteration number of the network to 5000; each weight w1[ i ] in the network][j]，w2[i][j]Assigning small non-zero random values [ -0.1,0.1)]The learning rate η is set to 0.5 (the error ε can be set in the interface). Wherein,the connection weight value from the ith element of the k-1 layer to the jth element of the k layer is used; />The adjustment value of the connection weight value from the ith element of the k-1 layer to the jth element of the k layer is used; />An input sum of the ith element of the kth layer; />The output of the ith element of the kth layer;
further, 36 connection weights of layers 1-2 in the network are specifically set:
element 1:(w1[0][0]，w1[0][1]，w1[0][2],w1[0][3])；
element 2:(w1[1][0]，w1[1][1]，w1[1][2]，w1[1][3])；
element 3:(w1[2][0]，w1[2][1]，w1[2][2]，w1[2][3])；
element 4:(w1[3][0]，w1[3][1]，w1[3][2]，w1[3][3])；
element 5:(w1[4][0]，w1[4][1]，w1[4][2]，w1[4][3])；
element 6:(w1[5][0]，w1[5][1]，w1[5][2]，w1[5][3])；
element 7:(w1[6][0]，w1[6][1]，w1[6][2]，w1[6][3])；
element 8:(w1[7][0]，w1[7][1]，w1[7][2]，w1[7][3])；
element 9:(w1[8][0]，w1[8][1]，w1[8][2]，w1[8][3])。
represented by array w1[9] [4] (see above): dimension 1 is the number of elements (9) of the first layer (input layer), each element having 4 connection weights with the elements (4) of the second layer; dimension 2 is the second layer element (4); the number of connection weights is 36.
Setting 20 connection weights of the 2 nd layer to the 3 rd layer as follows:
element 1:(w2[0][0]，w2[0][1]，w2[0][2]，w2[0][3]，w2[0][4])
element 2:(w2[1][0]，w2[1][1]，w2[1][2]，w2[1][3]，w2[1][4])
element 3:(w2[2][0]，w2[2][1]，w2[2][2]，w2[2][3]，w2[2][4])
element 4:(w2[3][0]，w2[3][1]，w2[3][2]，w2[3][3]，w2[3][4])
expressed by an array w2[3] [1 ]: dimension 1 is the number of elements of the second layer (4), each element having 5 connection weights with the third layer (output layer) element (5); dimension 2 is the third layer element (5); the number of the connection weights is 20. According to the input data of the input layer, calculating the output of the layer 1 (input layer) according to the excitation function, then calculating the input and output of each element of the layer 2, and then calculating the input and output of each element of the layer 2.
The error reverse propagation learning algorithm adopts a gradient descent method to obtain the minimum value of the error function in the weight vector space, and can combine the weight value with the minimum error function into the solution of the learning problem. Since the gradient of the error function is calculated at each step in the learning iterative calculation, the continuous scalability of the error function must be ensured.
In the example, a Sigmoid function is adopted as a network excitation function; sigmoid function (or Sigmoid function) is the most commonly used excitation function in network, and its expression is
Where the constant c can be arbitrarily chosen, and its reciprocal 1/c is called the temperature parameter in the random neural network.
When c=1, S c (x) In a simple form, i.eThe derivative is as follows:
when-infinity<x<At infinity, there is 0<S c (x)<1. The derivative of this function can be represented by itself, greatly reducing the computational effort in algorithm iterations.
Specifically, layer 1 (input layer)
The input of this layer is the input parameter of each sample pair, so only the output of each element is calculated.
Element 1:wherein: f (x) -excitation function S c (x)；
Element 2:
element 3:
element 4:
element 5:
element 6:
element 7:
8 thElement (b):
element 9:
represented by array O1[9 ]: the output of the first layer 9 elements is represented, the upper 9 outputs being the 9 elements of this array.
Specifically, layer 2 (hidden layer)
The input of each element in the layer is the sum of the product of the output of each element in the layer 1 and the weight input, and the output of each element is obtained by calculation according to the excitation function and the input of each element.
(1) Input of elements (4)
Element 1:
element 2:
element 3:
element 4:
represented by array I2[4 ]: representing the input of the second layer of 4 elements, the upper 4 outputs are the 4 elements of this array.
Element 1:
I2[0]＝O1[0]w1[0][0]+O1[1]w1[1][0]+O1[2]w1[2][0]+O1[3]w1[3][0]+O1[4]w1[4][0]+O1[5]w1[5][0]+O1[6]w1[6][0]+O1[7]w1[7][0]+O1[8]w1[8][0]
element 2:
I2[1]＝O1[0]w1[0][1]+O1[1]w1[1][1]+O1[2]w1[2][1]+O1[3]w1[3][1]+O1[4]w1[4][1]+O1[5]w1[5][1]+O1[6]w1[6][1]+O1[7]w1[7][1]+O1[8]w1[8][1]
element 3:
I2[2]＝O1[0]w1[0][2]+O1[1]w1[1][2]+O1[2]w1[2][2]+O1[3]w1[3][2]+O1[4]w1[4][2]+O1[5]w1[5][2]+O1[6]w1[6][2]+O1[7]w1[7][2]+O1[8]w1[8][2]
element 4:
I2[3]＝O1[0]w1[0][3]+O1[1]w1[1][3]+O1[2]w1[2][3]+O1[3]w1[3][3]+O1[4]w1[4][4]+O1[5]w1[5][4]+O1[6]w1[6][4]+O1[7]w1[7][4]+O1[8]w1[8][4]
the synthesis is as follows: (i (0 < = i < 4) represents the i-th element of layer 2, j (0 < = j < 9) represents the j-th element of layer 1)
I2[i]＝O1[0]w1[0][i]+O1[1]w1[1][i]+O1[2]w1[2][i]+O1[3]w1[3][i]+O1[4]w1[4][i]+O1[5]w1[5][i]+O1[6]w1[6][i]+O1[7]w1[7][i]+O1[8]w1[8][i]
I2[0] =i2 [1] =i2 [2] =i2 [3] =0; (having an initial value of 0)
for(i＝0；i<4；i++)
(2) Output of each element (3)
Element 1:wherein: f (x) -excitation function S c (x)；
Element 2:
element 3:
element 4:
represented by array O2[4 ]: representing the output of the 4 elements of the second layer.
The synthesis is as follows: i (0 < = i < 4) represents the i-th element of layer 2
O2[i]＝f(I2[i]) Wherein: f (x) -excitation function S c (x)
Last layer 3 (output layer)
The input of each element in the layer is the sum of the product of the output of each element in the layer 2 and the weight input, and the output of each element is obtained by calculation according to the excitation function and the input of each element.
(1) Input of each element (5)
Element 1:
element 2:
element 3:
element 4:
element 5:
represented by array I3[5 ]: representing the input of the third layer of 5 elements, the upper 5 outputs are the 5 elements of this array.
Element 1: i3[0] =O2 [0] w20 ] +O2[1] w21 ] +O2[2] w22 ] [0] +O2[3] w23 ] [0];
element 2: i3[0] =O2 [0] w20 ] [1] +O2[1] w21 ] +O2[2] w22 ] [1] +O2[3] w23 ] [1];
element 3: i3[0] =O2 [0] w20 ] [2] +O2[1] w21 ] +O2[2] w22 ] [2] +O2[3] w23 ] [2];
element 4: i3[0] =O2 [0] w20 ] [3] +O2[1] w21 ] +O2[2] w22 ] [3] +O2[3] w23 ];
element 5: i3[0] =O2 [0] w20 ] [4] +O2[1] w21 ] [4] +O2[2] w22 ] [4] +O2[3] w23 ] [4];
the synthesis is as follows: i3[0] =i3 [1] =i3 [2] =i3 [3] =i3 [4] =0; (having an initial value of 0)
(2) Output of each element (1)
Element 1:wherein: f (x) -excitation function S c (x)
Element 2:
element 3:
element 4:
element 5:
represented by array O3[5 ]: the output of the third layer 5 elements is represented, the upper 5 outputs being the 5 elements of this array.
The synthesis is as follows: o3[ i ]]＝f(I3[i]) Wherein: f (x) -excitation function S c (x)
Then calculate the error
After all the outputs of the 3 rd layer, i.e. the output layer, are calculated, the sum of squares r of the errors of the elements of the output layer can be calculated, and then the calculated error value is compared with the set error value epsilon. If r > epsilon, the error is too large to adjust each connection weight; otherwise, the error requirement is met, all weights are not required to be adjusted, and all current connection weights are reserved so as to carry out risk classification by using the network determined by the weights.
The error calculation formula:
if r > epsilon, entering the next step to adjust each connection weight; if r < = epsilon, the reserved connection weight exits.
Finally, the weight is adjusted by back propagation
If r > ε, which indicates that the actual output is far from the desired output (sample value), the individual connection weights must be adjusted. In the system, the number of the weights to be adjusted is 15 from layer 2 to layer 3, and 36 from layer 1 to layer 2 are corresponding to the connection weights.
1) Adjusting layer 2 to layer 3 weights (15 total)
(1) Computing output layer (layer 3)
The d value representing the j-th element of the output layer (layer 3).
Element 1:wherein: f' (x) -derivative S of the excitation function c (x)(1-S c (x))/>
Element 2:
element 3:
element 4:
element 5:
represented by array d3[5 ]: the d value of the third layer 5 elements is represented, the upper 5 values being the 5 elements of the array.
Element 1: d30= (O3 [0] -class (1)) f' (I3 [0 ])
Element 2: d3[1] = (O3 [1] -class (2)) f' (I3 [1 ])
Element 3: d3[2] = (O3 [2] -class (3)) f' (I3 [2 ])
Element 4: d33= (O3 [3] -class (4)) f' (I3 [3 ])
Element 5: d34= (O3 [4] -class (5)) f' (I3 [4 ])
Wherein: f' (x) -derivative S of the excitation function c (x)(1-S c (x))
(2) Calculating an adjustment weight
According to the formulaThe formula of the adjustment value of the layer 2 to layer 3 weights can be calculated as follows:
elements 1 to 3 of the second layer (5):
second layer 2 nd to third layer each element (5):
second layer 3 rd to third layer individual elements (5):
elements 4 to 3 of the second layer (5):
expressed by array dw2[5] [5 ]: the 1 st dimension is the number of elements (4) of the second layer, each element has 1 connection weight adjustment value with the elements (5) of the third layer (output layer); dimension 2 is the third layer element (5); the number of the connection weight adjustment values is 20.
Elements 1 to 3 of the second layer (5):
dw2[0][0]＝-ηd3[0]O2[0]，dw2[0][1]＝-ηd3[1]O2[0]，dw2[0][2]＝-ηd3[2]O2[0]，dw2[0][3]＝-ηd3[3]O2[0]，dw2[0][4]＝-ηd3[4]O2[0]。
second layer 2 nd to third layer each element (5):
dw2[1][0]＝-ηd3[0]O2[1]，dw2[1][1]＝-ηd3[1]O2[1]，dw2[1][2]＝-ηd3[2]O2[1]，dw2[1][3]＝-ηd3[3]O2[1]，dw2[1][4]＝-ηd3[4]O2[1]。
second layer 3 rd to third layer individual elements (5):
dw2[2][0]＝-ηd3[0]O2[2]，dw2[2][1]＝-ηd3[1]O2[2]，dw2[2][2]＝-ηd3[2]O2[2]，dw2[2][3]＝-ηd3[3]O2[2]，dw2[2][4]＝-ηd3[4]O2[2]。
elements 4 to 3 of the second layer (5):
dw2[3][0]＝-ηd3[0]O2[3]，dw2[3][1]＝-ηd3[1]O2[3]，dw2[3][2]＝-ηd3[2]O2[3]，dw2[3][3]＝-ηd3[3]O2[3]，dw2[3][4]＝-ηd3[4]O2[3]。
the synthesis is as follows: (i (0 < = i < 4) represents the i-th element of layer 2, j (0 < = j < 5) represents the j-th element of layer 3)
dw2[i][j]＝-η*d3[j]*O2[i]
for(i＝0；i<4；i++)
{
dw2[i][j]＝-η*d3[j]*O2[i]；
}
(3) Adjusting weights
After the adjustment value of each connection weight is calculated, the formula can be usedAnd adjusting each connection weight, wherein the formula is modified as follows:
elements 1 to 3 of the second layer (5):
second layer 2 nd to third layer each element (5):
second layer 3 rd to third layer individual elements (5):
elements 4 to 3 of the second layer (5):
/>
2) Adjusting layer 1 to layer 2 weights (18 total)
(1) Calculating hidden layer (layer 2)
The d value representing the j-th element of the hidden layer (layer 2).
Element 1:
wherein: f (x) -excitation function S c (x)
Element 2:
element 3:
element 4:
represented by array d2[4 ]: the d value representing the second layer of 4 elements, the upper 4 values being the 4 elements of the array.
Element 1:
d2[0]＝(d3[0]w2[0][0]+d3[1]w2[0][1]+d3[2]w2[0][2]+d3[3]w2[0][3]+d3[4]w2[0][4]+d3[5]w2[0][5]+d3[6]w2[0][6]+d3[7]w2[0][7]+d3[6]w2[0][8])f(I2[0])
wherein: f (x) -excitation function S c (x)
Element 2:
d2[1]＝(d3[0]w2[1][0]+d3[1]w2[1][1]+d3[2]w2[1][2]+d3[3]w2[1][3]+d3[4]w2[1][4]+d3[5]w2[1][5]+d3[6]w2[1][6]+d3[7]w2[1][7]+d3[8]w2[1][8])f(I2[1])
element 3:
d2[2]＝(d3[0]w2[2][0]+d3[1]w2[2][1]+d3[2]w2[2][2]+d3[3]w2[2][3]+d3[4]w2[2][4]+d3[5]w2[2][5]+d3[6]w2[2][6]+d3[7]w2[2][7]+d3[8]w2[2][8])f(I2[2])
element 4:
d2[3]＝(d3[0]w2[3][0]+d3[1]w2[3][1]+d3[2]w2[3][2]+d3[3]w2[3][3]+d3[4]w2[3][4]+d3[5]w2[3][5]+d3[6]w2[3][6]+d3[7]w2[3][7]+d3[8]w2[3][8])f(I2[3])
(2) calculating an adjustment weight
According to the formulaThe formula of the adjustment value of the layer 2 to layer 3 weights can be calculated as follows:
first layer 1 st element to second layer each element (4):
dw1[0][0]＝-ηd2[0]O1[0]，dw1[0][1]＝-ηd2[1]O1[0]，dw1[0][2]＝-ηd2[2]O1[0]，dw1[0][3]＝-ηd2[3]O1[0]。
first layer 2 nd element to second layer each element (4):
dw1[1][0]＝-ηd2[0]O1[1]，dw1[1][1]＝-ηd2[1]O1[1]，dw1[1][2]＝-ηd2[2]O1[1]，dw1[1][3]＝-ηd2[3]O1[1]。
first layer 3 rd element to second layer individual element (4):
dw1[2][0]＝-ηd2[0]O1[2]，dw1[2][1]＝-ηd2[1]O1[2]，dw1[2][2]＝-ηd2[2]O1[2]，dw1[2][3]＝-ηd2[3]O1[2]。
first layer 4 th element to second layer each element (4):
dw1[3][0]＝-ηd2[0]O1[3]，dw1[3][1]＝-ηd2[1]O1[3]，dw1[3][2]＝-ηd2[2]O1[3]，dw1[3][3]＝-ηd2[3]O1[3]。
first layer 5 th element to second layer individual element (4):
dw1[4][0]＝-ηd2[0]O1[4]，dw1[4][1]＝-ηd2[1]O1[4]，dw1[4][2]＝-ηd2[2]O1[4]，dw1[4][3]＝-ηd2[3]O1[4]。
first layer 6 th element to second layer individual element (4):
dw1[5][0]＝-ηd2[0]O1[5]，dw1[5][1]＝-ηd2[1]O1[5]，dw1[5][2]＝-ηd2[2]O1[5]，dw1[5][3]＝-ηd2[3]O1[5]。
first layer 7 th element to second layer individual element (4):
dw1[6][0]＝-ηd2[0]O1[6]，dw1[6][1]＝-ηd2[1]O1[6]，dw1[6][2]＝-ηd2[2]O1[6]，dw1[6][3]＝-ηd2[3]O1[6]。
first layer 8 th element to second layer individual element (4):
dw1[7][0]＝-ηd2[0]O1[7]，dw1[7][1]＝-ηd2[1]O1[7]，dw1[7][2]＝-ηd2[2]O1[7]，dw1[7][3]＝-ηd2[3]O1[7]。
first layer 9 th element to second layer individual element (4):
dw1[8][0]＝-ηd2[0]O1[8]，dw1[8][1]＝-ηd2[1]O1[8]，dw1[8][2]＝-ηd2[2]O1[8]，dw1[8][3]＝-ηd2[3]O1[8]。
expressed by an array dw1[9] [4 ]: dimension 1 is the number of elements (9) of the first layer, each element having 4 connection weight adjustment values with the second layer (hidden layer) elements (4); dimension 2 is the second layer element (4); the number of the connection weight adjustment values is 36.
The synthesis is as follows: (i (0 < = i < 9) represents the i-th element of layer 1, j (0 < = j < 4) represents the j-th element of layer 2)
dw1[i][j]＝-η*d2[j]*O1[i]
for(i＝0；i<9；i++)
(3) Adjusting weights
After the adjustment value of each connection weight is calculated, the formula can be usedAnd adjusting each connection weight, wherein the formula is modified as follows:
first layer 1 st element to second layer each element (4):
first layer 2 nd element to second layer each element (4):
first layer 3 rd element to second layer individual element (4):
first layer 4 th element to second layer each element (4):
first layer 5 th element to second layer individual element (4):
first layer 6 th element to second layer individual element (4):
first layer 7 th element to second layer individual element (4):
first layer 8 th element to second layer individual element (4):
first layer 9 th element to second layer individual element (4):
the number of times to be recorded after the adjustment is automatically increased by 1 to record the total adjustment number of times, and if the adjustment number of times is larger than the prescribed maximum number of times (5000) and still does not meet the error requirement, the adjustment is stopped so as not to enter a dead loop.
After the adjustment of each connection weight is completed, returning to the step 3, and recalculating the input, output, error and the like of each element, as in the previous step, until the error is satisfied or the adjustment has been completed more than 5000 times.
The above steps are that after one sample pair is input, the calculation and the connection weight adjustment are carried out, after the completion, the next sample pair is input, and the above steps are repeated until all samples are finished. After the input and adjustment of all sample pairs are completed, the network model learning is finished, and all the connection weights obtained after the adjustment are saved for use in risk classification.
And 104, inputting the data samples acquired in real time in construction into the safety grading neural network to predict the construction safety level in real time.
After the system model is built, data samples in construction are collected in real time, and each relevant parameter of monitoring measurement is input to conduct hierarchical prediction.
Example 2
FIG. 3 illustrates a multi-factor security prediction system for underground large space construction, i.e., an electronic device 310 (e.g., a computer server with program execution capabilities) comprising at least one processor 311, a power supply 314, and a memory 312 and an input-output interface 313 communicatively coupled to the at least one processor 311, according to an exemplary embodiment of the present invention; the memory 312 stores instructions executable by the at least one processor 311 to enable the at least one processor 311 to perform the method disclosed in any one of the preceding embodiments; the input/output interface 313 may include a display, a keyboard, a mouse, and a USB interface for inputting/outputting data; the power supply 314 is used to provide power to the electronic device 310.
Those skilled in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.
The above-described integrated units of the invention, when implemented in the form of software functional units and sold or used as stand-alone products, may also be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in essence or a part contributing to the prior art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.
The foregoing is a detailed description of specific embodiments of the invention and is not intended to be limiting of the invention. Various alternatives, modifications and improvements will readily occur to those skilled in the relevant art without departing from the spirit and scope of the invention.
Claims (7)
1. A multi-factor security grading method for underground large space construction, which is characterized by comprising the following steps:
step 101, determining a construction monitoring index and a threshold value thereof according to the actual conditions of engineering; establishing a multi-factor safety grading comparison table for underground large-space construction according to the construction monitoring index and the threshold value thereof;
102, randomly generating data samples associated with each security level according to the underground large space construction multi-factor security grading comparison table;
step 103, constructing a security classification neural network, training the security classification neural network by using the data samples associated with each security level to establish a nonlinear mapping relationship from an input layer to an output layer of the security classification neural network, and enabling the security classification neural network to satisfy the following conditions: when the input layer has parameter input, the output layer automatically outputs a security level judging result;
104, inputting a data sample acquired in real time in construction into the safety grading neural network to predict the construction safety level in real time;
the step 101 includes:
step 1011, determining a construction monitoring index and a threshold value thereof: selecting existing indexes according to standards, determining index alarm values by numerical experiments and model experiments aiming at a large-span structure, and determining reduction coefficients by certain field experiments; proportional reduction is carried out on the index alarm value based on the reduction coefficient;
step 1012, security level determination method: establishing single-factor security level division according to each monitoring index; and, based on the division of the single-factor security level, dividing the multi-factor security level by a hierarchical analysis method;
the multi-factor security grading by the analytic hierarchy process comprises the following steps: constructing an hierarchical structure model with the indexes mutually connected; constructing index pairwise comparison judgment matrixes, and respectively calculating the maximum eigenvalue and the corresponding eigenvector of each judgment matrix; and carrying out consistency test on the judgment matrix based on the maximum eigenvalue, and obtaining a multi-factor security level classification result based on the tested judgment matrix.
2. The method of claim 1, wherein the input parameters of the security hierarchical neural network input layer comprise: the support pile or the support wall top is vertically displaced, the support pile or the support wall top is horizontally displaced, the support pile or the support wall body is horizontally displaced, the upright post structure is vertically displaced, the earth surface subsides, the support wall structure stress, the upright post structure stress, the support shaft force and the anchor rod tension force.
3. The method of claim 1, wherein the total number of adjustable connections of the security hierarchical neural network is 56.
4. A method according to any of claims 1-3, characterized in that a Sigmoid function is used as the error function of the security hierarchical neural network.
5. A method according to any one of claims 1-3, wherein the safety classification neural network has a maximum number of iterations of 5000 and a learning rate η of 0.5.
6. A method according to any one of claims 1 to 3, wherein the performance of the safety classification neural network is judged to be stable and training is stopped when the error rate of the safety classification neural network is less than a preset value.
7. A multi-factor safety prediction system for underground large space construction, comprising at least one processor and a memory communicatively connected with the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010225494.9A CN111582634B (en) | 2020-03-26 | 2020-03-26 | Multi-factor safety grading method and system for underground large-space construction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010225494.9A CN111582634B (en) | 2020-03-26 | 2020-03-26 | Multi-factor safety grading method and system for underground large-space construction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111582634A CN111582634A (en) | 2020-08-25 |
| CN111582634B true CN111582634B (en) | 2024-02-23 |
Family
ID=72111479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010225494.9A Active CN111582634B (en) | 2020-03-26 | 2020-03-26 | Multi-factor safety grading method and system for underground large-space construction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111582634B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113256261A (en) * | 2021-06-03 | 2021-08-13 | 北京汽车集团越野车有限公司 | Method and device for determining automobile modeling scheme and storage medium |
| CN114089055A (en) * | 2021-09-30 | 2022-02-25 | 安徽继远软件有限公司 | Method and system for monitoring safety state of power grid limited space operating personnel |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937784A (en) * | 2012-10-30 | 2013-02-20 | 中冶南方工程技术有限公司 | Artificial neural network based method for controlling online prediction of casting billet quality |
| CN104217122A (en) * | 2014-09-15 | 2014-12-17 | 北京市市政工程研究院 | Tunnel construction process safety evaluation method based on multi-element information early warning system |
| CN105260575A (en) * | 2015-11-17 | 2016-01-20 | 中国矿业大学 | Roadway surrounding rock deformation predicting method based on neural network |
| CN106355540A (en) * | 2016-09-18 | 2017-01-25 | 三峡大学 | Small- and medium-sized reservoir dam safety evaluating method based on GRA-BP (grey relational analysis and back propagation) neural network |
| CN107782785A (en) * | 2017-09-08 | 2018-03-09 | 中冶建筑研究总院有限公司 | A kind of underground utilities security assessment method |
| CN108241964A (en) * | 2017-11-23 | 2018-07-03 | 贵州电网有限责任公司铜仁供电局 | Capital construction scene management and control mobile solution platform based on BP artificial nerve network model algorithms |
| CN109190872A (en) * | 2018-07-05 | 2019-01-11 | 珠海许继芝电网自动化有限公司 | A kind of power distribution network Supply Security integrated evaluating method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| BRPI0910573B1 (en) * | 2008-04-17 | 2019-09-03 | The Travelers Indemnity Company | system for processing a property insurance claim |
-
2020
- 2020-03-26 CN CN202010225494.9A patent/CN111582634B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937784A (en) * | 2012-10-30 | 2013-02-20 | 中冶南方工程技术有限公司 | Artificial neural network based method for controlling online prediction of casting billet quality |
| CN104217122A (en) * | 2014-09-15 | 2014-12-17 | 北京市市政工程研究院 | Tunnel construction process safety evaluation method based on multi-element information early warning system |
| CN105260575A (en) * | 2015-11-17 | 2016-01-20 | 中国矿业大学 | Roadway surrounding rock deformation predicting method based on neural network |
| CN106355540A (en) * | 2016-09-18 | 2017-01-25 | 三峡大学 | Small- and medium-sized reservoir dam safety evaluating method based on GRA-BP (grey relational analysis and back propagation) neural network |
| CN107782785A (en) * | 2017-09-08 | 2018-03-09 | 中冶建筑研究总院有限公司 | A kind of underground utilities security assessment method |
| CN108241964A (en) * | 2017-11-23 | 2018-07-03 | 贵州电网有限责任公司铜仁供电局 | Capital construction scene management and control mobile solution platform based on BP artificial nerve network model algorithms |
| CN109190872A (en) * | 2018-07-05 | 2019-01-11 | 珠海许继芝电网自动化有限公司 | A kind of power distribution network Supply Security integrated evaluating method |
Non-Patent Citations (3)
| Title |
|---|
| "Neural Network Model for the Prediction of Safe Work Behavior in Construction Projects";D.A.Patel 等;《Journal of Construction Engineering and Management》;第1-11 页 * |
| "神经网络范式下硐室群施工安全风险预警研究";江新等;《中国安全科学学报》;20180215;第183-184页 * |
| 江新等."神经网络范式下硐室群施工安全风险预警研究".《中国安全科学学报》.2018, * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111582634A (en) | 2020-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107480341B (en) | A kind of dam safety comprehensive method based on deep learning | |
| CN113837477B (en) | Method, device and equipment for predicting power grid faults under typhoon disasters driven by data | |
| CN109885907B (en) | Cloud model-based satellite attitude control system health state assessment and prediction method | |
| Perera et al. | Static–dynamic multi-scale structural damage identification in a multi-objective framework | |
| Ding et al. | Structural damage identification by sparse deep belief network using uncertain and limited data | |
| CN111582634B (en) | Multi-factor safety grading method and system for underground large-space construction | |
| CN111797364B (en) | Landslide multilayer safety evaluation method based on composite cloud model | |
| CN116150897A (en) | Machine tool spindle performance evaluation method and system based on digital twin | |
| CN110472363B (en) | Surrounding rock deformation grade prediction method and system suitable for high-speed railway tunnel | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| CN115270266B (en) | Slope monitoring state risk judging method based on improved analytic hierarchy process | |
| CN115270635A (en) | Bayes-neural network high-rise building earthquake demand and vulnerability prediction method | |
| CN116562331B (en) | Method for optimizing SVM by improving reptile search algorithm and application thereof | |
| CN114580545A (en) | Wind turbine generator gearbox fault early warning method based on fusion model | |
| CN114741758B (en) | Building earthquake-resistant toughness preliminary design method and system based on machine learning | |
| CN115982141A (en) | Characteristic optimization method for time series data prediction | |
| CN112347670B (en) | Rockfill material creep parameter prediction method based on neural network response surface | |
| Jensen et al. | Compromise design of stochastic dynamical systems: A reliability-based approach | |
| CN113610268A (en) | Carbon emission prediction method based on residential area space form | |
| CN117251926A (en) | Earthquake motion intensity index optimization method for earthquake response prediction | |
| CN117010691A (en) | Transmission tower safety evaluation method and device, computer equipment and storage medium | |
| CN111026790A (en) | Structure safety assessment and forecasting method based on data mining | |
| CN116307352A (en) | Engineering quantity index estimation method and system based on machine learning | |
| CN115392349A (en) | Fault diagnosis method and device for cutting part of heading machine and heading machine | |
| CN114118688A (en) | Power grid engineering cost risk early warning method based on sequence relation analysis |
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 |