A network construction method, system, medium and equipment for steel structure monitoring

Abstract

The present invention relates to the technical field of steel structure monitoring, and discloses a network construction method, system, medium and equipment for steel structure monitoring, including: establishing a finite element model of the steel structure to be tested, and performing modal analysis to obtain the inherent modal parameters of the steel structure; based on the inherent modal parameters, calculating the trace change rate of the Fisher information matrix to determine the modal order; taking the modal order as the lower limit of the number of sensor arrangements, optimizing the selection of sensor arrangement schemes with the goal of information effectiveness and optimal network performance; on the basis of the selected sensor arrangement schemes, optimizing the routing path with the goal of the node energy safety balance ratio being within the safety value range or the total path cost being minimum, and obtaining a steel structure monitoring network. The service life of the steel structure monitoring network is extended.

Description

Technical Field

[0001] The present invention relates to the technical field of steel structure monitoring, and in particular to a network construction method, system, medium and equipment for steel structure monitoring. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Large steel structures have a huge number of components and many degrees of freedom, and are limited by practical factors such as actual monitoring conditions, monitoring costs, and communication capabilities. The deployment of health monitoring systems faces practical engineering challenges such as limited number of sensor arrangements, limited compatibility of different types of sensors, and limited life of wireless data transmission networks.

[0004] Since large steel structures such as trusses and grids have a large degree of freedom, it is usually impossible to deploy sensors point by point to achieve full area coverage due to cost and stability considerations. Therefore, optimizing sensor layout and network construction is the basic link of intelligent operation and maintenance of large steel structure projects. Its main purpose is to achieve complete coverage perception of key points of steel structure status and real-time long-distance and large-capacity transmission of measurement data based on a limited number of sensors.

[0005] In current practice and research, the optimal layout of sensors and the construction of sensor networks are often carried out independently. However, in the monitoring of large steel structures, the two must be optimized and designed in a coordinated manner. Communication requirements should be considered when arranging sensors, and the sensor network should be built based on the layout plan. However, there is no relevant research at present. Summary of the invention

[0006] In order to solve the above problems, the present invention provides a network construction method, system, medium and equipment for steel structure monitoring, studies large-scale steel structures such as truss structures, grid structures and other massive degrees of freedom and complex engineering constraints, takes the modal order as the lower limit of the number of sensors to be arranged, and optimizes the dynamic optimization layout strategy of sensors with the goal of information effectiveness and optimal network performance. On this basis, the flooding direction of routing request packets is guided to avoid excessively long useless transmission paths, and finally realizes real-time online dynamic routing path multi-objective optimization, thereby extending the service life of the steel structure monitoring network.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A first aspect of the present invention provides a network construction method for steel structure monitoring, comprising:

[0009] Establish a finite element model of the steel structure to be tested, and perform modal analysis to obtain the inherent modal parameters of the steel structure; based on the inherent modal parameters, calculate the trace change rate of the Fisher information matrix and determine the modal order; use the modal order as the lower limit of the number of sensors to be arranged, and optimize the selection of sensor arrangement scheme with the goal of information effectiveness and optimal network performance;

[0010] On the basis of the selected sensor arrangement scheme, the routing path is optimized with the goal of ensuring that the node energy safety balance ratio is within the safety value range and / or the total path cost is minimized to obtain the steel structure monitoring network.

[0011] Furthermore, the trace change rate of the Fisher information matrix is:

[0012] ;

[0013] in, is the trace rate of change of the Fisher information matrix, For the front of the steel structure The dynamic response sensitivity matrix of the order structure, , For the front of the steel structure The mode shape, For the front of the steel structure Damage sensitivity matrix of order, For the front of the steel structure The dynamic response sensitivity matrix of the order structure, tr ( ) represents the trace of the matrix, T Indicates transpose.

[0014] Furthermore, the optimization selection of the sensor layout scheme adopts a multi-objective solution algorithm that integrates a genetic algorithm and a particle swarm optimization algorithm, and if there are multiple sensor layout schemes in the Pareto front of the multi-objective solution algorithm, then for each sensor layout scheme, the values ​​of multiple evaluation indicators are calculated respectively, and a comprehensive score is obtained through weighted summation, and the sensor layout scheme with the highest comprehensive score is selected.

[0015] Furthermore, the multiple evaluation indicators include: maximum modal recognition capability, parameter recognition error, energy consumption and coverage.

[0016] Furthermore, the weight of the indicator in the weighted sum is: a comprehensive weight obtained by linearly combining the subjective weight and the objective weight.

[0017] Furthermore, the total path cost is:

[0018] ;

[0019] ;

[0020] in, is the total path cost, n is the number of sensors in the selected sensor layout scheme, and is the regulating factor, and is the weight factor, Indicates sensor The remaining energy, Indicates sensor The average residual energy of neighbor sensors, Represents the average residual energy of sensors in the entire steel structure monitoring network.

[0021] Furthermore, the network performance is:

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] in, For network performance, It is a sensor As the energy consumption coefficient of the transmitter, is the energy consumed to transmit 1 byte per unit distance. is the loss index, is the size of the data packet, To send the sensor The distance to the receiving sensor, It is a sensor As the energy consumption factor of the receiver, is the largest total energy consumption among all sensors, Indicates the number of sensor arrangements, Monitor network connectivity for steel structures, std ( ) represents standard deviation.

[0029] A second aspect of the present invention provides a network construction system for steel structure monitoring, comprising:

[0030] The sensor layout optimization module is configured to: establish a finite element model of the steel structure to be tested, and perform modal analysis to obtain the inherent modal parameters of the steel structure; based on the inherent modal parameters, calculate the trace change rate of the Fisher information matrix and determine the modal order; use the modal order as the lower limit of the number of sensors to be arranged, and optimize the selection of sensor layout schemes with the goal of information effectiveness and optimal network performance;

[0031] The routing path optimization module is configured to: based on the selected sensor layout scheme, optimize the routing path with the goal of keeping the node energy safety balance ratio within the safety value range and / or minimizing the total path cost to obtain a steel structure monitoring network.

[0032] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor, and the program, when executed by the processor, implementing the steps in a network construction method for steel structure monitoring as described above.

[0033] A fourth aspect of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the program, the steps in a network construction method for steel structure monitoring as described above are implemented.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention studies large steel structures with massive degrees of freedom and complex engineering constraints, such as truss structures and grid structures, and takes the modal order as the lower limit of the number of sensors to be arranged. With the goal of optimizing information effectiveness and network performance, the present invention optimizes the dynamic optimization arrangement strategy of sensors, thereby achieving complete steel structure status monitoring and effective data transmission through a limited number of sensors.

[0036] The present invention proposes multiple evaluation indicators for network load balancing, including total path cost and node energy safety balance ratio, to guide the flooding direction of routing request packets, avoid excessively long useless transmission paths, and ultimately achieve real-time online dynamic routing path multi-objective optimization to extend the network service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which constitute a part of the specification of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but do not constitute limitations of the present invention.

[0038] Figure 1 This is an architecture diagram of a network construction method for steel structure monitoring according to the first embodiment of the present invention;

[0039] Figure 2This is a flow chart of routing path optimization according to the first embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0042] In the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention is further described below with reference to the accompanying drawings and embodiments.

[0043] Terminology explanation:

[0044] RREQ (Routing Request): Routing request message, used to find a path.

[0045] RREP (Route Reply): is a message used to respond to a route request (RREQ) to help the source node establish a routing path to the destination node.

[0046] Zigbee: is a protocol for low-power, wireless network communications, widely used in the field of Internet of Things.

[0047] Embodiment 1

[0048] The purpose of this first embodiment is to provide a network construction method for steel structure monitoring.

[0049] This embodiment provides a network construction method for steel structure monitoring, an optimization layout method for massive degree-of-freedom sensors considering communication constraints, constructs a layout scheme evaluation framework including communication capabilities, and establishes an integrated perception network of multiple numbers and types of sensors based on this.

[0050] This embodiment provides a network construction method for steel structure monitoring, such as Figure 1 As shown, including:

[0051] Step 1: Multi-objective optimization arrangement of large steel structure monitoring sensors to solve the problem of how to select measurement locations.

[0052] The difficulties in optimizing the arrangement of sensors are mainly due to the huge degrees of freedom of large steel structures, the heterogeneous types of sensors, the many actual engineering constraints, and the non-unique optimization indicators, which theoretically constitute a large-scale nonlinear multi-objective optimization problem.

[0053] Step 101: Conduct finite element modal analysis.

[0054] In order to determine the reasonable modal order and avoid the problems of drastic increase in computational complexity and decreased numerical stability caused by too high an order, a modal order selection method based on the damage sensitivity matrix is ​​introduced, and the matrix trace criterion is used to select the appropriate target modal order.

[0055] Among them, the modal order selection method based on the damage sensitivity matrix includes the following specific steps:

[0056] Step (1), establishing a finite element model: establishing a finite element model of the large steel structure to be tested, and performing modal analysis to obtain the natural modal parameters (n-order modal frequency and n-order modal vibration shape) of the large steel structure.

[0057] Step (2), calculate the damage sensitivity matrix: Assuming that the large steel structure is a low-damping system, only considering the damage caused by stiffness changes, calculate the damage sensitivity matrix of each mode , Indicates Damage sensitivity matrix of the order mode:

[0058] ;

[0059] in, For the The damage factor of each unit, For the The stiffness matrix of each element is and Respectively The vibration shape and frequency of the first mode, is the total number of units, is the total number of structural degrees of freedom (i.e., the total number of modal orders), Representative The natural frequency of the order mode, Indicates The modal vibration shape vector of the first order. A unit refers to a specific component in the steel structure. Here, "unit" is the basic building block in finite element analysis and is used for steel discretization structure.

[0060] Step (3): Calculate the Fisher information matrix based on damage sensitivity and vibration mode .

[0061] Based on the damage sensitivity matrix , construct the overall damage sensitivity matrix S; is a matrix composed of damage sensitivity matrices of various orders, and The relationship between them is:

[0062] ;

[0063] in, is the initially selected modal order, For the Damage sensitivity matrix.

[0064] make , and define is the structural dynamic response sensitivity matrix; where For large steel structures Mode shape:

[0065] .

[0066] Then, before Fisher information matrix .

[0067] Step (4), calculate the rate of change of the trace: before calculation Step and front The rate of change of the trace of the order Fisher information matrix :

[0068] ;

[0069] in, Represents the trace of the matrix.

[0070] Step (5), determine the modal order: when the trace change rate of the Fisher information matrix When it approaches zero, it is considered The modal information contains most of the information of large steel structures, and m is taken as the target modal order.

[0071] In this embodiment, select The first-order modes are used as the basis for optimizing the arrangement of sensors.

[0072] Step 102, establish a multi-objective mathematical optimization problem for sensor layout. The lower limit of the number of sensors to be arranged is determined by the number of modal orders of the structure (i.e., the target modal order m selected by the matrix trace criterion in step 101), and the optimal number is usually determined by the inflection point of the sensor number-layout performance curve. Commonly used methods include the effective independence method, the minimum modal confidence criterion method, the modal matrix QR decomposition method, etc. In view of the different advantages and disadvantages of different schemes given by a single criterion layout method, this embodiment proposes a multi-objective optimization scheme, and mines a suitable objective function from the classical sensor optimization layout theory according to actual needs to form a multi-objective optimization problem. The details are as follows:

[0073] Assuming that a sensor deployment scheme including k sensors is known, the information validity index and the network performance index are used as the deployment performance of the sensor deployment scheme.

[0074] (1) Define information effectiveness indicators.

[0075] In this embodiment, a modal assurance criterion (MAC) matrix is ​​used to measure the information validity. MAC matrix: used to ensure the identifiability between modal vectors, and its largest non-diagonal element can help to better optimize the arrangement of sensors.

[0076] Use the QR (orthogonal triangular) decomposition method to obtain the initial measurement points and perform QR decomposition on the transpose of the modal vibration matrix: .in, : Modal vibration matrix, which represents the modal vibration information of the steel structure. : The unit transformation matrix. : orthogonal matrix. : upper triangular matrix.

[0077] set up is the modal matrix obtained after modal analysis, where represents the number of measurable degrees of freedom, is the number of selected modes. The MAC matrix can be expressed as:

[0078] ;

[0079] In the formula, and are the first and second sensors in the modal matrix under the sensor arrangement scheme containing k sensors. Column and It should be noted that the sensors are arranged at specific positions of the steel structure to collect dynamic response data of the steel structure, such as displacement, acceleration, etc. These response data reflect the vibration form of the steel structure at different modal frequencies, that is, the modal vibration vector and ; Therefore, different sensor layout schemes will cause the modal vibration vector and There are differences, and then the most appropriate solution is determined based on the MAC matrix, etc.

[0080] Information Validity It is represented by the largest off-diagonal element of the MAC matrix:

[0081] .

[0082] Add a sensor one by one from the degrees of freedom other than the initial measuring point, and calculate the maximum value of the non-diagonal elements of the MAC matrix; obtain the variation relationship diagram of the maximum value of the non-diagonal elements of the MAC with the increase of the number of measuring points; according to the variation relationship of the maximum value of the non-diagonal elements of the MAC matrix, combined with the structural conditions and the requirements of modal identification accuracy, determine the optimal number of sensors.

[0083] This embodiment uses Choose from among the candidate degrees of freedom As nodes for placing sensors, the maximum off-diagonal element of the MAC matrix is ​​minimized. The objective function is constructed as:

[0084] ;

[0085] In the formula, represents the sensor layout scheme, Represents the largest off-diagonal element of the MAC matrix. , “0” means that the two modal vectors are easier to distinguish from each other, and “1” means that there is a high linear correlation between the two modal vectors, that is, they are indistinguishable.

[0086] (2) Define network performance indicators: Network performance includes two aspects: network connectivity and energy efficiency.

[0087] (201) Network connectivity:

[0088] Using the adjacency matrix Represents the connectivity of the network, the adjacency matrix is a symmetric matrix whose elements Representation Node and nodes The connection relationship between:

[0089] ;

[0090] in, Is a node (i.e., sensor i, a node in the network is a sensor) and node The distance between is the maximum communication distance of the node. and nodes The distance between them is less than or equal to the maximum communication distance ( ),but , indicating that the two nodes can communicate directly; otherwise , indicating that the two nodes cannot communicate directly.

[0091] Introducing the Judgment Matrix :

[0092] ;

[0093] if If there is no element 0 in , the network is connected. Represents the adjacency matrix Powers of, ranging from 1 to ; Adjacency matrix is a size of A square matrix is ​​used to represent the connection relationship between nodes; Represents the number of nodes in the network, i.e., k.

[0094] The mathematical expression of network connectivity is:

[0095] .

[0096] if =0, indicating that the judgment matrix There are no zero elements in , indicating that the entire network is connected; otherwise, it is not connected.

[0097] (202) Energy efficiency:

[0098] node Send data packets to a distance of The energy consumed by the node The amount is: ;in, Is a node As the energy consumption coefficient of the transmitter, is the energy consumed to transmit 1 byte per unit distance. is the path loss exponent; The size of the data packet, usually in bits; For sending node The distance to the receiving node (the sensor that receives the data packet).

[0099] node Energy consumed to receive a packet for:

[0100] ;

[0101] in, Is a node As the energy consumption factor of the receiver.

[0102] node Total energy consumption for:

[0103] .

[0104] Normalize the energy consumption of all nodes:

[0105] ;

[0106] in, The node with the largest total energy consumption among all nodes.

[0107] The standard deviation (SD) of the energy consumption of all nodes after normalization is used to characterize the uniformity of energy consumption between nodes: .

[0108] (203) Network performance It is expressed as: .

[0109] (3) Model the wireless sensor placement optimization problem as a multi-objective optimization problem: .

[0110] Step 103: Conduct research on solving optimization problems. In view of the challenges of multiple objectives, multiple variables, strong nonlinearity, and huge search space for this optimization problem, an intelligent optimization algorithm based on large-scale parallel computing is constructed. The genetic algorithm and the particle swarm optimization algorithm are integrated to give full play to their respective advantages in processing integer variables and continuous variables, effectively improving the algorithm's efficient global exploration ability in a huge answer space.

[0111] The improved multi-objective solution algorithm combines the advantages of genetic algorithm (GA) and particle swarm optimization algorithm (PSO). The specific steps are as follows:

[0112] Step (1), initializing the particle swarm: randomly initializing the position and velocity of the particles, wherein the position of a particle encodes a sensor layout position, that is, a sensor layout scheme; the particle velocity is the changing trend of the sensor position during the search process.

[0113] Step (2), calculate the fitness value: according to the fitness function Calculate the fitness value of each particle, and determine whether the number of sensors q encoded by each particle is greater than m, and delete particles that are greater than m.

[0114] Step (3), update individual extreme values ​​and group extreme values: update the optimal position of each particle and the optimal position of the population according to the fitness value.

[0115] Step (4), particle position and velocity update.

[0116] Particle velocity update formula:

[0117] ;

[0118] in, : No. The particle in The speed at the iteration; : Inertia weight, which controls the particle's dependence on its previous velocity; and : learning factors, indicating that the particle is biased towards the individual optimal value and the global optimum the extent of; and : A random number that obeys a uniform distribution on [0,1]; : No. The best position that a particle has searched so far; : The optimal position searched by the entire particle swarm, that is, the optimal position of the population; : No. The particle in The position at the iteration.

[0119] Particle position update formula:

[0120] ;

[0121] in, : No. The particle in The position at the iteration; : No. The particle in The position at the iteration; : No. The particle in The speed at which the iterations are performed.

[0122] Step (5), selection and crossover operation: According to the fitness value, select the two best individuals for crossover operation to generate new individuals.

[0123] Step (6), mutation operation: perform mutation operation on new individuals to increase population diversity.

[0124] Step (7), determine the termination condition: determine whether the maximum number of iterations has been reached or the accuracy requirement has been met. If so, output multiple optimal solutions in the Pareto front; otherwise, return to step (2) and continue iterating.

[0125] Step 104: Make a decision on the optimal layout plan. The same sensor layout plan often does not perform well under all evaluation criteria. Therefore, the Pareto front given by multi-objective optimization often contains multiple optimal solutions that do not dominate each other, and certain trade-offs must be made in actual engineering applications. A multi-attribute decision-making evaluation scheme is adopted to create an indicator system for sensor layout plans, and then a subjective and objective weighting scheme is adopted for comprehensive weighting, a new model of heterogeneous indicator fusion decision-making is established, and multiple candidate layout plans in the Pareto front are ranked for decision-making, and the final sensor optimization layout plan is recommended.

[0126] In this embodiment, for each sensor arrangement scheme in the Pareto front x a , calculate the index values ​​of the four evaluation indicators respectively p ab ( b =1,2,3,4), and based on the comprehensive weight, a comprehensive score is obtained by weighted summation; the sensor layout scheme with the highest comprehensive score is selected.

[0127] Among them, the four evaluation indicators of the sensor layout plan include:

[0128] (1) Maximum Modal Identification: This is mainly used to ensure the accuracy and completeness of modal identification and is usually a priority indicator in structural health monitoring. Indicator Description: The sensor layout scheme is evaluated through the modal assurance criterion (MAC) to ensure the independence between modal vectors. Formula:

[0129] ;

[0130] .

[0131] (2) Minimize Parameter Identification Error. Improving the accuracy of parameter estimation by minimizing the error is crucial for accurately assessing the health of the structure. Indicator Description: Optimize sensor layout through Fisher Information Matrix (FIM) to reduce the uncertainty of structural parameter identification. Evaluate the effect of sensor layout scheme in parameter identification based on FIM trace. Formula:

[0132] ;

[0133] ;

[0134] Where FIM represents the Fisher information matrix; represents the measurement vector matrix, which contains the response data of all sensors at different positions.

[0135] (3) Minimum Energy Consumption: In wireless sensor networks, energy consumption is a key indicator that determines the continuous working time and maintenance cost of the monitoring system. Indicator Description: Particularly suitable for wireless sensor networks, it reduces energy consumption and extends sensor life through optimal layout. Formula: ECE stands for structural performance energy consumption evaluation, which is based on the calculation of structural dynamic response; ECP stands for wireless sensor network performance energy consumption evaluation, which is determined based on the communication and energy consumption model of the wireless sensor network.

[0136] (4) Coverage: Ensure that the coverage of the sensor is maximized. Indicator description: Consider the coverage of the sensor installation, especially in large-scale structures, formula:

[0137] .

[0138] The calculation method of comprehensive weight is:

[0139] (1) Subjective weighting method: analytic hierarchy process (AHP).

[0140] Constructing a judgment matrix , according to expert opinion, for each pair of indicators and Compare the relative importance and get the matrix elements :like Compare More importantly, ;like Compare Not important, then ; , indicating equal importance.

[0141] Calculate the judgment matrix The normalized feature vector of is used as the subjective weight:

[0142] Calculate the maximum eigenvalue of the judgment matrix and its corresponding eigenvector , after normalization, we get the weight vector. The formula is as follows:

[0143] , calculated using the eigenvalue decomposition method and , the normalized weight vector is:

[0144] .

[0145] Then the consistency check is performed: the consistency ratio (CR) is calculated: ,in, is the consistency indicator, is a random consistency indicator. Formula:

[0146] ;

[0147] For a 4-order matrix, The value of is 0.89. , then the judgment matrix is ​​considered to have satisfactory consistency.

[0148] (2) Objective weighting method: entropy weight method.

[0149] Standardize indicator data to eliminate dimension effects:

[0150] ;

[0151] Calculate the entropy value of each indicator :

[0152] ;

[0153] in, represents the jth evaluation index value of the i-th sample (one sample is one sensor layout scheme);

[0154] ;

[0155] in, is a constant, is the sample size.

[0156] Calculating objective weights :

[0157] .

[0158] (3) Comprehensive weighting method:

[0159] Combining subjective weight and objective weight , use the linear combination method to calculate the comprehensive weight of the i-th indicator:

[0160] ;

[0161] in, It is the adjustment coefficient of subjective and objective weights.

[0162] Step 2: Design standardized interfaces for multi-source heterogeneous sensors to solve the problem of how to support data compatibility.

[0163] Health monitoring of large steel structures involves multiple sensors of different types, brands and functions. The output data interfaces are not unified, compatible and interoperable, which makes maintenance, function expansion and notification application development very difficult. Therefore, it is necessary to unify and abstractly model the characteristic parameters, data formats and interaction protocols of sensor signal interfaces, establish standard data interfaces to facilitate the transmission, analysis and integration of multi-sensor data, and then use them to build a "plug-and-play" self-identifying and self-organizing intelligent perception network, and promote data fusion at the perception layer.

[0164] With the health monitoring of large steel structures as the main goal, a networked intelligent sensor standard interface based on the IEEE1451 (intelligent sensor interface standard) series of standards is developed. Among them, the main function of WTIM (wireless transmitter module) is to transmit the collected and monitored data to NCAP (network capable application processor). The fusion integration processing algorithm of multi-sensor data in the local area is deployed on the NCAP microcontroller, and a multi-modal one-dimensional convolutional neural network signal processing model is designed to self-diagnose the health status of each sensor, autonomously filter out environmental interference, and automatically correct measurement deviations to achieve self-compensation and self-correction of multiple sensors. After NCAP collects and intelligently processes the data in real time, it encodes the data according to the given transmission protocol and transmits the data of each edge node to the cloud through wired or 5G communication.

[0165] In this embodiment, the processing flow of the multimodal one-dimensional convolutional neural network signal processing model includes:

[0166] (1) Input signal preprocessing. For heterogeneous data from various sensors such as stress, acceleration, vibration, and temperature, normalization is first performed to eliminate the influence of different dimensions. Preprocessing methods such as low-pass filtering are used to remove high-frequency noise to ensure the stability of the input signal.

[0167] (2) Multimodal data fusion. In order to take advantage of the advantages of different sensors, a multi-input convolutional layer is designed to process the data channels of different sensors separately. The multi-input convolutional layer uses four parallel one-dimensional convolutional layers to process the input data of stress, acceleration, temperature, and vibration sensors respectively.

[0168] (3) Feature extraction and fusion. A shared convolutional layer (number of filters = 16, kernel size = 3) is designed to concatenate and fuse the features extracted from each channel to learn the association information between different sensors.

[0169] (4) Self-diagnosis and self-correction module. A health status prediction and self-correction module is added to the CNN model to monitor the sensor status, detect anomalies and perform automatic compensation, including baseline drift compensation, standard reference correction and environmental interference adaptive filtering.

[0170] Specifically, after the shared convolutional layer, a health status prediction module is designed to predict the health status of the sensor (normal, abnormal, drift, etc.) through a simple fully connected network. If an abnormal state is detected, the self-correction module is automatically triggered to adjust the output to compensate for the measurement error caused by sensor drift or failure.

[0171] (5) Edge computing optimization. Quantization and pruning techniques are used to further reduce computing overhead and make it suitable for real-time operation on edge devices.

[0172] Step 3: Research on large-capacity and low-energy consumption networking technology to achieve autonomous networking of sensor capacity expansion and energy reduction, and solve the problem of how to transmit data.

[0173] Large-scale steel structure monitoring involves many sensors, and the field environment such as bridges is relatively harsh, and manual maintenance is cumbersome, complicated, time-consuming and labor-intensive. Therefore, in the design of large-scale steel structure monitoring systems, the capacity, reliability and service life of the perception network must be considered. It is planned to use the ZigBee protocol for wireless networking, which has the characteristics of low power consumption, low cost, low complexity and high reliability, and a single network supports up to 65,000 sensor nodes. Depending on the area where the large steel structure is located, most sensors are suitable for battery power supply. This embodiment analyzes the network energy consumption and formulates an energy consumption optimization strategy for the perception network, which is crucial to ensuring network stability and extending network life.

[0174] First, we conduct an energy consumption analysis of the sensor nodes in the network. The energy consumption of each node is mainly composed of two parts: computing consumption and communication consumption. Since the large-scale steel structure monitoring network has a large range, the communication consumption of the RF chip constitutes most of the energy consumption. The RF chip has multiple working modes, among which the sleep mode has the lowest energy consumption, which is only a few tens of times that of the sending / receiving mode. We design energy consumption measurement experiments in various environments, form energy consumption reports for various types of monitoring sensors, and provide specific and effective parameter values ​​for subsequent energy consumption optimization.

[0175] Improve the sleep strategy and data fusion algorithm of wireless sensing networks to reduce energy consumption. ZigBee networks have a coordinator responsible for network configuration and management, which is usually deployed in the NCAP module in the standard interface. Based on this, a centralized network-wide sleep and wake-up strategy is established. First, define the request sleep instruction mechanism for each terminal device. When the coordinator receives the request sleep instruction from all other devices, it broadcasts the sleep command to the entire network, so that the entire network enters a low-power mode. In large-scale steel structure health monitoring applications, a periodic measurement mode is usually used. Explore and evaluate synchronous or asynchronous time-driven wake-up mechanisms, and switch the network from sleep state to normal working state in a timely manner through timing or event triggering to ensure real-time data transmission.

[0176] Multiple evaluation indicators of network load balancing are proposed, including the remaining energy of the node, the node energy safety balance ratio, and the energy balance cost of the routing path, and a multi-objective optimization problem with priority is constructed; decision rules are formulated to focus on evaluating the node energy safety balance ratio, guide the flooding direction of routing request packets, avoid excessively long useless transmission paths, and ultimately achieve real-time online dynamic routing path multi-objective optimization to extend the network service life.

[0177] In this embodiment, the node energy balance cost and the path energy balance cost are defined as follows:

[0178] Node energy balance cost The definitions are as follows:

[0179] ;

[0180] According to the node energy balance function, the path energy balance cost is defined as:

[0181] ;

[0182] Path energy balance cost It is used to measure the balance of node energy in the path. The greater the cost, the greater the difference in node energy and the lower the priority of path selection. n is the number of sensors deployed in the selected sensor deployment scheme.

[0183] in, and is a weight factor used to balance the influence of node residual energy and neighbor node energy. Representation Node Current remaining energy. Average energy of neighboring nodes Representation Node The average residual energy of the neighbor nodes. The average network energy Represents the average residual energy of nodes in the entire network.

[0184] Considering the energy balance cost and energy consumption cost comprehensively, the total path cost is defined :

[0185] ;

[0186] Including: Energy consumption cost It represents the remaining energy of nodes on the path. The lower the energy, the greater the cost. and It is an adjustment factor used to balance the weights of the two costs.

[0187] When broadcast networking is performed, nodes are layered in order of networking, and the number of layers is set to ; nodes that join the network at the same time are in the same layer; nodes that join the network later are sublayers. Incremental; the first one to enter the network is the parent layer, decrement; add a neighbor to the neighbor table value, and add the node that sends the RREQ packet to the route discovery request message RREQ Value, according to Perform directional selective forwarding.

[0188] Among them, dynamic routing path multi-objective optimization, such as Figure 2 As shown, the specific steps include:

[0189] Step 301: construct several routing paths. The steps of constructing the routing paths include:

[0190] (1) When the source node initiates a routing request, it directly broadcasts the RREQ packet and then proceeds to step (2).

[0191] (2) When other nodes receive the RREQ packet, if they are the destination node, they send a RREP packet in reply and discard the received RREQ, send a RREP to confirm the routing path, and establish the routing path; otherwise, go to step (3).

[0192] (3) For a node, if the RREQ packet contains and the node If the values ​​are the same, it means that the sending node is from the same layer, then the RREQ packet is broadcast and forwarded, and the process returns to step (2); otherwise, the process goes to step (4).

[0193] (4) For a node, if the RREQ group contains Than the node If the value is small, the RREQ packet is forwarded directly to If the sum of the nodes is equal and larger, update L in RREQ and return to step (2); otherwise, go to step (5).

[0194] (5) For a node, if Than the node If the value is large, the RREQ packet is forwarded directly to For nodes with equal or smaller values, update L in RREQ and return to step (2).

[0195] Step 302: For a certain routing path, calculate the node energy security balance ratio , if all nodes on the path satisfy , then select the path; otherwise, go to step 304.

[0196] Among them, the node energy security balance ratio is:

[0197] ;

[0198] in, is the safety value of node energy relative to the entire network. When , the energy of the node is at a lower level relative to the entire network and should be protected.

[0199] Step 303: For a certain routing path, determine whether there is a path that satisfies If there is a path with more nodes, go to step 302; otherwise, go to step 304.

[0200] Step 304: For each routing path, calculate the node energy balance cost and the path energy balance cost, and select the path that meets the conditions. The path to send data.

[0201] The present invention studies large steel structures with massive degrees of freedom and complex engineering constraints, such as truss structures and grid structures, and takes the modal order as the lower limit of the number of sensors to be arranged. With the goal of optimizing information effectiveness and network performance, the present invention optimizes the dynamic optimization arrangement strategy of sensors, thereby achieving complete steel structure status monitoring and effective data transmission through a limited number of sensors.

[0202] The present invention proposes multiple evaluation indicators for network load balancing, including total path cost and node energy safety balance ratio, to guide the flooding direction of routing request packets, avoid excessively long useless transmission paths, and ultimately achieve real-time online dynamic routing path multi-objective optimization to extend the network service life.

[0203] Embodiment 2

[0204] The purpose of this second embodiment is to provide a network construction system for steel structure monitoring, including:

[0205] The sensor layout optimization module is configured to: establish a finite element model of the steel structure to be tested, and perform modal analysis to obtain the inherent modal parameters of the steel structure; based on the inherent modal parameters, calculate the trace change rate of the Fisher information matrix and determine the modal order; use the modal order as the lower limit of the number of sensors to be arranged, and optimize the selection of sensor layout schemes with the goal of information effectiveness and optimal network performance;

[0206] The routing path optimization module is configured to: based on the selected sensor layout scheme, optimize the routing path with the goal of keeping the node energy safety balance ratio within the safety value range and / or minimizing the total path cost to obtain a steel structure monitoring network.

[0207] It should be noted here that each module in this embodiment corresponds to each step in Example 1 one by one, and the specific implementation process is the same, which will not be repeated here.

[0208] Embodiment 3

[0209] This embodiment provides a computer-readable storage medium on which a computer program is stored. The program is executed by a processor. When the program is executed by the processor, the steps in the network construction method for steel structure monitoring described in the above-mentioned embodiment 1 are implemented.

[0210] Embodiment 4

[0211] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, the steps in the network construction method for steel structure monitoring described in the above embodiment 1 are implemented.

[0212] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0213] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A network construction method for steel structure monitoring, characterized in that: include: Establish a finite element model of the steel structure to be tested, and perform modal analysis to obtain the inherent modal parameters of the steel structure; Based on the intrinsic modal parameters, the trace change rate of the Fisher information matrix is ​​calculated to determine the modal order; the modal order is used as the lower limit of the number of sensors to be arranged, with the information effectiveness and network performance Taking the best as the goal, optimize the sensor layout scheme; Based on the selected sensor layout scheme, the routing path is optimized with the goal of ensuring that the node energy safety balance ratio is within the safety value range and / or the total path cost is minimized to obtain the steel structure monitoring network; The optimization selection of the sensor layout scheme adopts a multi-objective solution algorithm that integrates genetic algorithm and particle swarm optimization algorithm; the multi-objective solution algorithm has the following specific steps: step (1), randomly initialize the position and speed of the particle, and the position of a particle encodes a sensor layout scheme; step (2), according to the fitness function Calculate the fitness value of each particle; step (3), update the optimal position of each particle and the optimal position of the population according to the fitness value; step (4), update the particle position and speed; step (5), select two optimal individuals according to the fitness value for crossover operation to generate new individuals; step (6), perform mutation operation on the new individuals; step (7), determine whether the maximum number of iterations has been reached or the accuracy requirements have been met. If so, output multiple optimal solutions in the Pareto front; otherwise, return to step (2) and continue iteration; The multi-objective optimization of routing paths includes the following specific steps: step 301, constructing several routing paths; step 302, for a certain routing path, calculating the node energy safety balance ratio, if all nodes of the path meet , then select the path where is the node energy safety balance ratio, is the safety value of the node energy relative to the entire network; otherwise, proceed to step 304; step 303, for a certain routing path, determine whether there is a satisfying If there is a path with more nodes, go to step 302; otherwise, go to step 304; Step 304, for each routing path, calculate the node energy balance cost and the path energy balance cost, and select the path that meets the conditions The path to send data, where is the total path cost.

2. A network construction method for steel structure monitoring according to claim 1, characterized in that: The trace rate of change of the Fisher information matrix is: ； in, is the rate of change of the trace of the Fisher information matrix, For the front of the steel structure The dynamic response sensitivity matrix of the order structure, , For the front of the steel structure The first mode shape, For the front of the steel structure Damage sensitivity matrix of order, For the front of the steel structure The dynamic response sensitivity matrix of the order structure, tr ( ) represents the trace of the matrix, T Indicates transpose.

3. A network construction method for steel structure monitoring according to claim 1, characterized in that: If there are multiple sensor layout schemes in the Pareto front of the multi-objective solution algorithm, the values ​​of multiple evaluation indicators are calculated for each sensor layout scheme, and a comprehensive score is obtained by weighted summation, and the sensor layout scheme with the highest comprehensive score is selected.

4. A network construction method for steel structure monitoring according to claim 3, characterized in that: The multiple evaluation indicators include: maximum modal recognition capability, parameter recognition error, energy consumption and coverage.

5. A network construction method for steel structure monitoring according to claim 3, characterized in that: The weight of the indicator in the weighted sum is: a comprehensive weight obtained by linearly combining the subjective weight and the objective weight.

6. A network construction method for steel structure monitoring according to claim 1, characterized in that: The total path cost is: ； ； in, is the total path cost, n is the number of sensors in the selected sensor layout scheme, and is the regulating factor, and is the weight factor, Indicates sensor The remaining energy, Indicates sensor The average residual energy of neighbor sensors, Represents the average residual energy of sensors in the entire steel structure monitoring network.

7. A network construction method for steel structure monitoring according to claim 1, characterized in that: The network performance is: ； ； ； ； ； ； in, For network performance, It is a sensor As the energy consumption coefficient of the transmitter, is the energy consumed to transmit 1 byte per unit distance. is the loss index, is the size of the data packet, To send the sensor The distance to the receiving sensor, It is a sensor As the energy consumption factor of the receiver, is the largest total energy consumption among all sensors, Indicates the number of sensor arrangements, is the connectivity of the steel structure monitoring network, and std( ) represents the standard deviation.

8. A network construction system for steel structure monitoring, characterized in that: include: A sensor arrangement optimization module is configured to: establish a finite element model of the steel structure to be tested, and perform modal analysis to obtain the inherent modal parameters of the steel structure; Based on the intrinsic modal parameters, the trace change rate of the Fisher information matrix is ​​calculated to determine the modal order. The modal order is used as the lower limit of the number of sensors to be deployed, and the sensor deployment scheme is optimized with the goal of information effectiveness and optimal network performance. A routing path optimization module is configured to: based on the selected sensor arrangement scheme, optimize the routing path with the goal of ensuring that the node energy safety balance ratio is within a safety value range and / or the total path cost is minimized, so as to obtain a steel structure monitoring network; The optimization selection of the sensor layout scheme adopts a multi-objective solution algorithm that integrates genetic algorithm and particle swarm optimization algorithm; the multi-objective solution algorithm has the following specific steps: step (1), randomly initialize the position and speed of the particle, and the position of a particle encodes a sensor layout scheme; step (2), according to the fitness function Calculate the fitness value of each particle; step (3), update the optimal position of each particle and the optimal position of the population according to the fitness value; step (4), update the particle position and speed; step (5), select two optimal individuals according to the fitness value for crossover operation to generate new individuals; step (6), perform mutation operation on the new individuals; step (7), determine whether the maximum number of iterations has been reached or the accuracy requirements have been met. If so, output multiple optimal solutions in the Pareto front; otherwise, return to step (2) and continue iteration; The multi-objective optimization of routing paths includes the following specific steps: step 301, constructing several routing paths; step 302, for a certain routing path, calculating the node energy safety balance ratio, if all nodes of the path meet , then select the path where is the node energy safety balance ratio, is the safety value of the node energy relative to the entire network; otherwise, proceed to step 304; step 303, for a certain routing path, determine whether there is a satisfying If there is a path with more nodes, go to step 302; otherwise, go to step 304; Step 304, for each routing path, calculate the node energy balance cost and the path energy balance cost, and select the path that meets the conditions The path to send data, where is the total path cost.

9. A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor, characterized in that: When the program is executed by a processor, the steps in a network construction method for steel structure monitoring as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the network construction method for steel structure monitoring as described in any one of claims 1 to 7 are implemented.

Images