WO2023173531A1 - 一种基于NExT-递归图的结构损伤检测方法及系统 - Google Patents

一种基于NExT-递归图的结构损伤检测方法及系统 Download PDF

Info

Publication number
WO2023173531A1
WO2023173531A1 PCT/CN2022/088073 CN2022088073W WO2023173531A1 WO 2023173531 A1 WO2023173531 A1 WO 2023173531A1 CN 2022088073 W CN2022088073 W CN 2022088073W WO 2023173531 A1 WO2023173531 A1 WO 2023173531A1
Authority
WO
WIPO (PCT)
Prior art keywords
acceleration response
recursive
response time
time history
structural
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.)
Ceased
Application number
PCT/CN2022/088073
Other languages
English (en)
French (fr)
Inventor
段元锋
诸锜
章红梅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University
Original Assignee
Zhejiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University filed Critical Zhejiang University
Priority to GB2306117.9A priority Critical patent/GB2620668A/en
Priority to US18/270,718 priority patent/US20240386164A1/en
Publication of WO2023173531A1 publication Critical patent/WO2023173531A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
    • B65D—CONTAINERS FOR STORAGE OR TRANSPORT OF ARTICLES OR MATERIALS, e.g. BAGS, BARRELS, BOTTLES, BOXES, CANS, CARTONS, CRATES, DRUMS, JARS, TANKS, HOPPERS, FORWARDING CONTAINERS; ACCESSORIES, CLOSURES, OR FITTINGS THEREFOR; PACKAGING ELEMENTS; PACKAGES
    • B65D85/00—Containers, packaging elements or packages, specially adapted for particular articles or materials
    • B65D85/30—Containers, packaging elements or packages, specially adapted for particular articles or materials for articles particularly sensitive to damage by shock or pressure
    • B65D85/36—Containers, packaging elements or packages, specially adapted for particular articles or materials for articles particularly sensitive to damage by shock or pressure for bakery products, e.g. biscuits
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • G01M5/0033—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings by determining damage, crack or wear
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
    • B65D—CONTAINERS FOR STORAGE OR TRANSPORT OF ARTICLES OR MATERIALS, e.g. BAGS, BARRELS, BOTTLES, BOXES, CANS, CARTONS, CRATES, DRUMS, JARS, TANKS, HOPPERS, FORWARDING CONTAINERS; ACCESSORIES, CLOSURES, OR FITTINGS THEREFOR; PACKAGING ELEMENTS; PACKAGES
    • B65D2585/00—Containers, packaging elements or packages specially adapted for particular articles or materials
    • B65D2585/30—Containers, packaging elements or packages specially adapted for particular articles or materials for articles particularly sensitive to damage by shock or pressure
    • B65D2585/36—Containers, packaging elements or packages specially adapted for particular articles or materials for articles particularly sensitive to damage by shock or pressure for biscuits or other bakery products
    • B65D2585/363—Containers, packaging elements or packages specially adapted for particular articles or materials for articles particularly sensitive to damage by shock or pressure for biscuits or other bakery products specific products
    • B65D2585/366—Pizza
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Definitions

  • the present invention relates to the technical field of structural damage detection, and in particular to a structural damage detection method and system based on NExT-recursive graph.
  • the use of relatively original acceleration response time history data can better avoid the loss of information.
  • the time history data is not direct in displaying the dynamic characteristics and has flaws.
  • the generated recursive graph samples have the characteristics of non-consistency, that is, under the same damage state, the generated recursive graphs are also very different. , therefore, the accuracy of the damage identification results is not high.
  • the purpose of the present invention is to provide a structural damage detection method and system based on NExT-recursive graphs, which can effectively detect structural damage and maximize the accuracy and robustness of detection.
  • the present invention provides the following solutions:
  • a structural damage detection method based on NExT-recursive graph including:
  • the NExT method is used to process the acceleration response time history signal to obtain the cross-correlation function signal of the acceleration response of different structural points;
  • the convolutional neural network model is trained and verified respectively through the training set and the verification set;
  • Structural damage detection is performed through the trained convolutional neural network model.
  • obtain the acceleration response time history signals of each point of the structure under different damage conditions including:
  • the random wind field generated by Kaimal spectrum is used to generate wind load
  • the buffeting wind force is loaded onto the numerical model under different damage conditions, and the acceleration response time history signals of each point of the structure under different damage conditions are obtained.
  • the cross-correlation function signal before performing recursive graph processing on the cross-correlation function signal, it also includes:
  • the cross-correlation function signal is normalized.
  • the robustness of the trained convolutional neural network model is tested through the test set.
  • the present invention also provides a structural damage detection system based on NExT-recursive graph, including:
  • the acceleration response time history signal acquisition module is used to obtain the acceleration response time history signal of each point of the structure under different damage conditions
  • the cross-correlation function signal acquisition module is used to process the acceleration response time history signal using the NExT method to obtain the cross-correlation function signal of the acceleration response of different structural points;
  • a three-dimensional recursive graph determination module is used to perform recursive graph processing on the cross-correlation function signal, and stack the generated recursive graphs to obtain a three-dimensional recursive graph;
  • a dividing module used to divide the three-dimensional recursive graph into a training set and a verification set
  • Training and verification modules used to train and verify the convolutional neural network model through the training set and the verification set respectively;
  • the structural damage detection module is used to detect structural damage through the trained convolutional neural network model.
  • the acceleration response time history signal acquisition module specifically includes:
  • Wind load generation unit used to generate wind load using random wind field generated by Kaimal spectrum
  • a buffet wind force calculation unit configured to calculate the buffet wind force based on the wind load
  • Different damage condition simulation units are used to simulate different damage conditions of the structure through stiffness reduction
  • the acceleration response time history signal acquisition unit is used to load the buffeting wind force onto the numerical model under different damage conditions to obtain the acceleration response time history signals of each point of the structure under different damage conditions.
  • the present invention discloses the following technical effects:
  • This invention uses the data processing method of NExT-recursive graph to generate convolutional neural network training samples under different damage conditions, which can effectively realize structural damage detection under different wind loads and is highly robust to white noise present in the signal. sex.
  • convolutional neural networks Compared with traditional machine learning algorithms, convolutional neural networks have inherent advantages in feature extraction of high-dimensional data of two dimensions and above, and can effectively improve its training efficiency and generalization ability in structural damage identification, and have better accuracy and lower training cost.
  • Figure 1 is a flow chart of a structural damage detection method based on NExT-recursive graph according to an embodiment of the present invention
  • Figure 2 shows the structure diagram of the convolutional neural network.
  • the purpose of the present invention is to provide a structural damage detection method and system based on NExT-recursive graphs, which can effectively detect structural damage and maximize the accuracy and robustness of detection.
  • the structural damage detection method based on NExT-recursive graph includes the following steps:
  • Step 101 Obtain the acceleration response time history signals of each point of the structure under different damage conditions.
  • Step 102 Use the NExT method to process the acceleration response time history signal to obtain cross-correlation function signals of the acceleration response of different structural points.
  • Step 103 Perform recursive graph processing on the cross-correlation function signal, and stack the generated recursive graphs to obtain a three-dimensional recursive graph.
  • Step 104 Divide the three-dimensional recurrence graph into a training set and a verification set.
  • Step 105 Train and verify the convolutional neural network model through the training set and the verification set respectively.
  • Step 106 Perform structural damage detection through the trained convolutional neural network model.
  • step 101 specifically includes:
  • the wind load is generated based on the random wind field generated by the Kaimal spectrum, the pulsating wind speed at different locations on the structure is calculated, and the buffeting wind force of each particle on the structure is calculated. Since static wind force does not produce acceleration, it does not need to be considered; and since the average wind speed experienced by the actual structure is random, for the samples used for training, the average wind speed used during generation is different from that used for testing.
  • different damage conditions of the structure are simulated through stiffness reduction.
  • Different damage conditions include different stiffness reduction rates and different damage locations. Including single damage and multiple damage; then use the buffeting wind force to load the numerical structural model under different damage conditions, and obtain the acceleration response time history signals of each point of the structure under different damage conditions.
  • steps 102-103 specifically include:
  • S AB (k) is the discrete cross-spectral density function
  • k is the discrete frequency
  • R AB (n) is the cross-correlation function at discrete time n.
  • the cross-correlation function results generated by NExT technology are R 11 , R 12 , R 13 , R 14 ,..., R 1N .
  • R is the effective section of the signal
  • N is the effective section of the signal
  • Steps 104-105 specifically include:
  • 70% of the three-dimensional recurrence graph is used as a training set, and the remaining 30% is used as a validation set for verification.
  • the convolutional neural network architecture established in this example is shown in Figure 2.
  • the construction of the convolutional neural network consists of a convolutional layer, a pooling layer and a fully connected layer.
  • the convolutional layer is used to extract features from the matrix information, and then the convolutional neural network is used to extract features from the matrix information.
  • the pooling layer reduces the data dimension and reduces the occurrence of over-fitting, and then uses a fully connected layer similar to a traditional neural network to map the extracted features into the target space.
  • the last layer of the convolutional neural network selects the mean square error function as the cost function to calculate the error between the predicted damage and the actual damage.
  • the formula is: Among them, f ij represents the j-th value on the label corresponding to the i-th sample, and y ij represents the j-th value of the convolutional neural network on the damage prediction vector of the i-th sample.
  • the convolutional layer uses the ReLU activation function, the formula is Among them, ⁇ is a positive number close to 0.
  • the gradient of the cost function on each training parameter is calculated through backpropagation, and each parameter is updated using a small-batch stochastic gradient descent algorithm.
  • This invention is used for non-destructive damage identification of structures in the field of civil engineering. It proposes a method of pre-processing the acceleration response of multiple points on the structure with NExT to obtain a quasi-free attenuation signal, which is used for the generation of recursion diagrams, so that the acceleration response under the same damage can be Recursive graphs have significant similar features, while convolutional neural networks are used for feature extraction. NExT processing can obtain similar structural signals from the original structural signals. Recursive graphs can effectively represent the development trends and patterns of phase space trajectories over time. Compared with traditional machine learning algorithms, convolutional neural networks have better performance on two-dimensional and The above feature extraction of high-dimensional data has inherent advantages, which can effectively improve its training efficiency and generalization ability in structural damage identification, with better accuracy and lower training cost.
  • the present invention also provides a structural damage detection system based on NExT-recursive graph, including:
  • the acceleration response time history signal acquisition module is used to obtain the acceleration response time history signals of each point of the structure under different damage conditions.
  • the cross-correlation function signal acquisition module is used to process the acceleration response time history signal using the NExT method to obtain the cross-correlation function signal of the acceleration response of different structural points.
  • a three-dimensional recursive graph determination module is used to perform recursive graph processing on the cross-correlation function signal, and stack the generated recursive graphs to obtain a three-dimensional recursive graph.
  • a dividing module used to divide the three-dimensional recursive graph into a training set and a verification set.
  • a training and verification module configured to respectively train and verify the convolutional neural network model through the training set and the verification set.
  • the structural damage detection module is used to detect structural damage through the trained convolutional neural network model.
  • the acceleration response time history signal acquisition module specifically includes:
  • Numerical model building unit is used to build numerical models of structures.
  • the wind load generation unit is used to generate wind loads using random wind fields generated by the Kaimal spectrum.
  • a buffet wind force calculation unit is used to calculate the buffet wind force based on the wind load.
  • Different damage condition simulation units are used to simulate different damage conditions of the structure through stiffness reduction.
  • the acceleration response time history signal acquisition unit is used to load the buffeting wind force onto the numerical model under different damage conditions to obtain the acceleration response time history signals of each point of the structure under different damage conditions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Geometry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Medical Informatics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mechanical Engineering (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种基于NExT-递归图的结构损伤检测方法及系统。该方法包括：获取不同损伤状况下结构各点的加速度响应时程信号；采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号；对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图；将所述三维递归图划分为训练集和验证集；通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证；通过训练好的卷积神经网络模型进行结构损伤检测。本发明能够有效对结构的损伤进行检测，最大限度提高检测的准确性和鲁棒性。

Description

一种基于NExT-递归图的结构损伤检测方法及系统
本申请要求于2022年3月16日提交中国专利局、申请号为202210257028.8、发明名称为“一种基于NExT-递归图的结构损伤检测方法及系统”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及结构损伤检测技术领域，特别是涉及一种基于NExT-递归图的结构损伤检测方法及系统。
背景技术
传统的基于结构动力特性的损伤识别算法通常需要对结构损伤和结构动力的特性关系进行复杂的公式推导，对于不同的结构，动力特性与损伤的关系各不相同。对于复杂结构来说，相关的理论推导会大大提高该方法的使用门槛，并且结构的损伤往往同时影响其基频、模态以及模态曲率等，因此单一地根据某些动力特性进行损伤识别很难得到较好的准确度，且这些特性在提取过程中往往会受到干扰导致失真，损伤识别结果精确度有待提高。
在传统的损失识别算法的基础上，采用较为原始的加速度响应时程数据能更好地避免信息的丢失，但是时程数据对于动力特性的展现不太直接，也有缺陷。直接使用加速度响应所生成的递归图样本进行卷积神经网络的损伤检测，所生成的递归图样本具有非一致性的特点，即在相同的损伤状态下，所生成的递归图也有很大的差别，因此，损伤识别结果精度不高。
发明内容
本发明的目的是提供一种基于NExT-递归图的结构损伤检测方法及系统，有效对结构的损伤进行检测，最大限度提高检测的准确性和鲁棒性。
为实现上述目的，本发明提供了如下方案：
一种基于NExT-递归图的结构损伤检测方法，包括：
获取不同损伤状况下结构各点的加速度响应时程信号；
采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号；
对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图；
将所述三维递归图划分为训练集和验证集；
通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证；
通过训练好的卷积神经网络模型进行结构损伤检测。
可选地，获取不同损伤状况下结构各点的加速度响应时程信号，具体包括：
搭建结构的数值模型；
采用Kaimal谱生成的随机风场产生风荷载；
基于所述风荷载计算抖振风力；
通过刚度折减模拟结构的不同损伤状况；
将所述抖振风力加载到不同损伤状况下的所述数值模型上，得到不同损伤状况下结构各点的加速度响应时程信号。
可选地，在对所述互相关函数信号进行递归图处理之前，还包括：
对所述互相关函数信号进行归一化处理。
可选地，当验证集中代价函数的误差计算值小于设定目标时，确定所述卷积神经网络模型训练完成。
可选地，还包括：
生成与所述训练集和所述验证集不同平均风速、不同损伤状况以及在加速度响应中添加不同信噪比的白噪声的三维递归图，构建测试集；
通过所述测试集对训练好的卷积神经网络模型的鲁棒性进行测试。
本发明还提供了一种基于NExT-递归图的结构损伤检测系统，包括：
加速度响应时程信号获取模块，用于获取不同损伤状况下结构各点的加速度响应时程信号；
互相关函数信号获取模块，用于采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号；
三维递归图确定模块，用于对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图；
划分模块，用于将所述三维递归图划分为训练集和验证集；
训练和验证模块，用于通过所述训练集和所述验证集分别对卷 积神经网络模型进行训练和验证；
结构损伤检测模块，用于通过训练好的卷积神经网络模型进行结构损伤检测。
可选地，加速度响应时程信号获取模块具体包括：
数值模型搭建单元，用于搭建结构的数值模型；
风荷载产生单元，用于采用Kaimal谱生成的随机风场产生风荷载；
抖振风力计算单元，用于基于所述风荷载计算抖振风力；
不同损伤状况模拟单元，用于通过刚度折减模拟结构的不同损伤状况；
加速度响应时程信号获取单元，用于将所述抖振风力加载到不同损伤状况下的所述数值模型上，得到不同损伤状况下结构各点的加速度响应时程信号。
根据本发明提供的具体实施例，本发明公开了以下技术效果：
本发明利用NExT-递归图的数据处理方法生成不同损伤状况下的卷积神经网络训练样本，能有效实现对不同风荷载下的结构损伤检测，并且对信号中存在的白噪声具有较高鲁棒性。并且卷积神经网络相对于传统的机器学习算法，其对二维及以上的高维数据的特征提取具有先天优势，能有效提高其在结构损伤识别上的训练效率和泛化能力，具备较好的精度及较低的训练成本。
说明书附图
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对 实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本发明实施例基于NExT-递归图的结构损伤检测方法的流程图；
图2为卷积神经网络结构图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明的目的是提供一种基于NExT-递归图的结构损伤检测方法及系统，有效对结构的损伤进行检测，最大限度提高检测的准确性和鲁棒性。
为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本发明作进一步详细的说明。
如图1所示，本发明提供的基于NExT-递归图的结构损伤检测方法，包括以下步骤：
步骤101：获取不同损伤状况下结构各点的加速度响应时程信号。
步骤102：采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号。
步骤103：对所述互相关函数信号进行递归图处理，并将生成的递归 图进行堆叠得到三维递归图。
步骤104：将所述三维递归图划分为训练集和验证集。
步骤105：通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证。
步骤106：通过训练好的卷积神经网络模型进行结构损伤检测。
其中，步骤101具体包括：
搭建结构的数值模型，用来模拟特定损伤下结构在风荷载下的动力响应；因考虑到社会安全以及成本，真实的结构无法人为地引入多种损伤工况，因此需要在结构的数值模型上引入损伤，然后在数值结构上施加模拟的风荷载。
根据Kaimal谱生成的随机风场产生风荷载，计算结构上不同位置的脉动风速，计算出结构上各质点的抖振风力。由于静风力不产生加速度，则不用考虑；并且由于实际结构所受到的平均风速是随机的，因此对于用于训练的样本，生成时采用的平均风速与用于测试时采用的不相同。
考虑实际结构在出现损伤时，通常会引起刚度的折减，因此在本实例中通过刚度折减对结构的不同损伤状况进行模拟，不同损伤状况包括不同的刚度折减率以及不同的损伤位置，包括单处损伤以及多处损伤；然后使用抖振风力对不同损伤状况下的数值结构模型进行加载，得到不同损伤状况下结构各点的加速度响应时程信号。
其中，步骤102-103具体包括：
自然激励技术所采用的互相关函数计算公式为：
Figure PCTCN2022088073-appb-000001
其中S AB(k)为离散的交叉谱密度函数，k为离散的频率，R AB(n)为离散时间n处的互相关函数。如针对自由度为N的结构，利用NExT技术所产生的互相关函数结果为R 11,R 12,R 13,R 14,...,R 1N。
在生成递归图前，需要对NExT处理后的信号有效段进行归一化处理：
Figure PCTCN2022088073-appb-000002
其中R为信号有效段，
Figure PCTCN2022088073-appb-000003
为归一化后的信号，对归一化后的信号分别进行递归图处理，则可以得到N种对应的递归图(尺寸为M×M，M为信号有效段对应点数)，将N张递归图进行堆叠即可得到三维(M×M×N)递归图。
其中步骤104-105具体包括：
将所述三维递归图中的70％作为训练集，剩余30％作为验证集进行验证。
本实例所建立的卷积神经网络架构如图2所示，卷积神经网络的搭建由卷积层、池化层和全连接层组成，首先使用卷积层对矩阵信息进行特征提取，之后用池化层降低数据维度并降低过拟合现象发生，之后再使用与传统神经网络类似的全连接层，将提取出来的特征映射到目标空间中。
卷积神经网络最后一层选择均方差函数作为代价函数，来计算预测损伤与实际损伤之间的误差，公式为
Figure PCTCN2022088073-appb-000004
其中f ij代表第i个样本对应的标签上的第j个数值，y ij代表卷积神经网络在第i个样本的损伤预测向量上的第j个数值。
卷积层采用ReLU激活函数，公式为
Figure PCTCN2022088073-appb-000005
其中，α为一个接近0的正数，通过反向传播计算代价函数在各个训练参数上的梯度，采用小批量随机梯度下降算法对各参数进行更新。
利用上述步骤产生的训练集和训练方式对卷积神经网络进行训练，直至对验证集中代价函数的误差计算值小于设定目标，对各点损伤程度的预测精度达到要求。
按照上述步骤生成与训练集和验证集不同平均风速、不同损伤状况、在加速度响应中添加不同信噪比的白噪声的递归图测试样本集，对训练得到的卷积神经网络进行鲁棒性测试。
本发明用于对土木工程领域的结构进行无损损伤识别，提出了对结构上多点的加速度响应进行预先进行NExT处理的方法获得类自由衰减信号，用于递归图的生成，使得相同损伤下的递归图具有显著的类似特征，同时采用卷积神经网络进行特征提取。NExT处理能从原始结构信号中获取到类似的结构信号，递归图能够有效表征相空间轨迹随时间的发展变化趋势和规律，而卷积神经网络相对于传统的机器学习算法，其对二维及以上的高维数据的特征提取具有先天优势，能有效提高其在结构损伤识别上的训练效率和泛化能力，具备较好的精度及较低的训练成本。
本发明还提供了一种基于NExT-递归图的结构损伤检测系统，包括：
加速度响应时程信号获取模块，用于获取不同损伤状况下结构各点的加速度响应时程信号。
互相关函数信号获取模块，用于采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号。
三维递归图确定模块，用于对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图。
划分模块，用于将所述三维递归图划分为训练集和验证集。
训练和验证模块，用于通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证。
结构损伤检测模块，用于通过训练好的卷积神经网络模型进行结构损伤检测。
其中，加速度响应时程信号获取模块具体包括：
数值模型搭建单元，用于搭建结构的数值模型。
风荷载产生单元，用于采用Kaimal谱生成的随机风场产生风荷载。
抖振风力计算单元，用于基于所述风荷载计算抖振风力。
不同损伤状况模拟单元，用于通过刚度折减模拟结构的不同损伤状况。
加速度响应时程信号获取单元，用于将所述抖振风力加载到不同损伤状况下的所述数值模型上，得到不同损伤状况下结构各点的加速度响应时程信号。
本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似部分互相参见即可。对于实施例公开的系统而言，由于其与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。
本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围 上均会有改变之处。综上所述，本说明书内容不应理解为对本发明的限制。

Claims (7)

  1. 一种基于NExT-递归图的结构损伤检测方法，其特征在于，包括：
    获取不同损伤状况下结构各点的加速度响应时程信号；
    采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号；
    对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图；
    将所述三维递归图划分为训练集和验证集；
    通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证；
    通过训练好的卷积神经网络模型进行结构损伤检测。
  2. 根据权利要求1所述的基于NExT-递归图的结构损伤检测方法，其特征在于，获取不同损伤状况下结构各点的加速度响应时程信号，具体包括：
    搭建结构的数值模型；
    采用Kaimal谱生成的随机风场产生风荷载；
    基于所述风荷载计算抖振风力；
    通过刚度折减模拟结构的不同损伤状况；
    将所述抖振风力加载到不同损伤状况下的所述数值模型上，得到不同损伤状况下结构各点的加速度响应时程信号。
  3. 根据权利要求1所述的基于NExT-递归图的结构损伤检测方法，其特征在于，在对所述互相关函数信号进行递归图处理之前，还包括：
    对所述互相关函数信号进行归一化处理。
  4. 根据权利要求1所述的基于NExT-递归图的结构损伤检测方法，其特征在于，当验证集中代价函数的误差计算值小于设定目标时，确定所述卷积神经网络模型训练完成。
  5. 根据权利要求1所述的基于NExT-递归图的结构损伤检测方法，其特征在于，还包括：
    生成与所述训练集和所述验证集不同平均风速、不同损伤状况以及在加速度响应中添加不同信噪比的白噪声的三维递归图，构建测试集；
    通过所述测试集对训练好的卷积神经网络模型的鲁棒性进行测试。
  6. 一种基于NExT-递归图的结构损伤检测系统，其特征在于，包括：
    加速度响应时程信号获取模块，用于获取不同损伤状况下结构各点的加速度响应时程信号；
    互相关函数信号获取模块，用于采用NExT方法对所述加速度响应时程信号进行处理，获得不同结构点位加速度响应的互相关函数信号；
    三维递归图确定模块，用于对所述互相关函数信号进行递归图处理，并将生成的递归图进行堆叠得到三维递归图；
    划分模块，用于将所述三维递归图划分为训练集和验证集；
    训练和验证模块，用于通过所述训练集和所述验证集分别对卷积神经网络模型进行训练和验证；
    结构损伤检测模块，用于通过训练好的卷积神经网络模型进行结构损伤检测。
  7. 根据权利要求6所述的基于NExT-递归图的结构损伤检测系统，其特征在于，加速度响应时程信号获取模块具体包括：
    数值模型搭建单元，用于搭建结构的数值模型；
    风荷载产生单元，用于采用Kaimal谱生成的随机风场产生风荷载；
    抖振风力计算单元，用于基于所述风荷载计算抖振风力；
    不同损伤状况模拟单元，用于通过刚度折减模拟结构的不同损伤状况；
    加速度响应时程信号获取单元，用于将所述抖振风力加载到不同损伤状况下的所述数值模型上，得到不同损伤状况下结构各点的加速度响应时程信号。
PCT/CN2022/088073 2022-03-16 2022-04-21 一种基于NExT-递归图的结构损伤检测方法及系统 Ceased WO2023173531A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
GB2306117.9A GB2620668A (en) 2022-03-16 2022-04-21 Next-recurrence plot-based structural damage detection method and system
US18/270,718 US20240386164A1 (en) 2022-03-16 2022-04-21 METHOD AND SYSTEM FOR DETECTING STRUCTURAL DAMAGE BASED ON NExT-RECURRENCE PLOTS

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210257028.8A CN114595726A (zh) 2022-03-16 2022-03-16 一种基于NExT-递归图的结构损伤检测方法及系统
CN202210257028.8 2022-03-16

Publications (1)

Publication Number Publication Date
WO2023173531A1 true WO2023173531A1 (zh) 2023-09-21

Family

ID=81818023

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/088073 Ceased WO2023173531A1 (zh) 2022-03-16 2022-04-21 一种基于NExT-递归图的结构损伤检测方法及系统

Country Status (4)

Country Link
US (1) US20240386164A1 (zh)
CN (1) CN114595726A (zh)
GB (1) GB2620668A (zh)
WO (1) WO2023173531A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120046402A (zh) * 2024-12-31 2025-05-27 中建研科技股份有限公司 一种基于深度学习的塔式结构损伤识别方法及系统
CN119470644B (zh) * 2025-01-09 2025-03-25 西南交通大学 一种基于数据驱动的重载机车一系钢弹簧损伤检测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706355A (zh) * 2009-11-17 2010-05-12 福州大学 基于NExT/ARMA的结构响应分析方法
CN108318129A (zh) * 2018-02-01 2018-07-24 石家庄铁道大学 桥梁结构模态参数真假甄别方法及终端设备
US20190017911A1 (en) * 2017-07-11 2019-01-17 Qatar University Real-time structural damage detection by convolutional neural networks
CN112001110A (zh) * 2019-05-27 2020-11-27 浙江大学 一种基于振动信号空间时时递归图卷积神经网络的结构损伤识别监测方法
CN114021403A (zh) * 2021-10-28 2022-02-08 山东大学 基于应变模态的承力结构件损伤识别方法及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100420944C (zh) * 2005-12-29 2008-09-24 西北工业大学 基于互相关函数幅值向量的随机振动结构损伤的检测方法
US7363172B2 (en) * 2006-01-05 2008-04-22 United States Of America As Represented By The Secretary Of The Navy Method and apparatus for detecting damage in structures
US8577628B2 (en) * 2009-04-10 2013-11-05 University Of South Carolina System and method for modal identification using smart mobile sensors
CN102128788A (zh) * 2010-12-21 2011-07-20 东南大学 基于改进自然激励技术的钢框架损伤诊断方法
CN109682561B (zh) * 2019-02-19 2020-06-16 大连理工大学 一种自动检测高速铁路桥梁自由振动响应以识别模态的方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706355A (zh) * 2009-11-17 2010-05-12 福州大学 基于NExT/ARMA的结构响应分析方法
US20190017911A1 (en) * 2017-07-11 2019-01-17 Qatar University Real-time structural damage detection by convolutional neural networks
CN108318129A (zh) * 2018-02-01 2018-07-24 石家庄铁道大学 桥梁结构模态参数真假甄别方法及终端设备
CN112001110A (zh) * 2019-05-27 2020-11-27 浙江大学 一种基于振动信号空间时时递归图卷积神经网络的结构损伤识别监测方法
CN114021403A (zh) * 2021-10-28 2022-02-08 山东大学 基于应变模态的承力结构件损伤识别方法及系统

Also Published As

Publication number Publication date
CN114595726A (zh) 2022-06-07
GB202306117D0 (en) 2023-06-07
US20240386164A1 (en) 2024-11-21
GB2620668A (en) 2024-01-17

Similar Documents

Publication Publication Date Title
CN110046379B (zh) 一种基于空间-频率信息的结构整体损伤识别方法
CN112001110B (zh) 一种基于振动信号空间时时递归图卷积神经网络的结构损伤识别监测方法
CN108846835B (zh) 基于深度可分离卷积网络的图像变化检测方法
CN111178206B (zh) 一种基于改进yolo的建筑预埋件检测方法及系统
CN111325190A (zh) 一种表情识别方法、装置、计算机设备及可读存储介质
CN115273814B (zh) 伪语音检测方法、装置、计算机设备和存储介质
CN115457365A (zh) 一种模型的解释方法、装置、电子设备及存储介质
CN114580239A (zh) 一种考虑不确定性的桥梁损伤识别方法
CN111595541A (zh) 一种海量振动传递率数据卷积神经网络处理的多维结构损伤识别方法
CN110222607A (zh) 人脸关键点检测的方法、装置及系统
CN110442149A (zh) 一种基于改进差分进化算法的无人机气动参数辨识方法
CN112802011A (zh) 一种基于vgg-bls的风机叶片缺陷检测方法
CN107832789A (zh) 基于平均影响值数据变换的特征加权k近邻故障诊断方法
US20240386164A1 (en) METHOD AND SYSTEM FOR DETECTING STRUCTURAL DAMAGE BASED ON NExT-RECURRENCE PLOTS
CN112183269A (zh) 一种适用于智能视频监控的目标检测方法与系统
CN112766381A (zh) 有限样本下属性引导的sar图像生成方法
Kong et al. Dynamic multi-scale convolution for dialect identification
CN116680969A (zh) 一种pso-bp算法的充填体评估参数预测方法及装置
CN117195042A (zh) 一种基于多变量递归图和卷积神经网络的结构损伤识别方法
CN114358091B (zh) 一种基于卷积神经网络的桩损伤识别方法、设备及介质
CN113642468B (zh) 身份认证方法及装置
Hussain et al. An efficient incremental learning algorithm for sound classification
CN115588112A (zh) 一种基于rfef-yolo目标检测方法
JP2020091813A (ja) ニューラルネットワークの学習方法、コンピュータプログラム及びコンピュータ装置
CN119068912A (zh) 基于声纹识别的工业诊断方法及相关设备

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 202306117

Country of ref document: GB

Kind code of ref document: A

Free format text: PCT FILING DATE = 20220421

WWE Wipo information: entry into national phase

Ref document number: 2306117.9

Country of ref document: GB

WWE Wipo information: entry into national phase

Ref document number: 18270718

Country of ref document: US

WWP Wipo information: published in national office

Ref document number: 2306117.9

Country of ref document: GB

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22931558

Country of ref document: EP

Kind code of ref document: A1