Bridge support early-stage damage intelligent diagnosis method and system based on vibration characteristics

Abstract

The invention relates to the technical field of early-stage damage diagnosis of bridge supports, in particular to an intelligent early-stage damage diagnosis method and system for a bridge support based on vibration characteristics, and the method comprises the steps: collecting vibration signals of the bridge support through a miniature vibration sensor array; based on the vibration signal, executing a multi-scale time-frequency feature extraction operation to obtain a time-frequency feature matrix; according to the time-frequency feature matrix, topological feature extraction operation is executed, and a topological feature vector is obtained; based on the topological feature vector, executing a nonlinear dynamic feature extraction operation to obtain a dynamic feature vector; according to the time-frequency feature matrix, the topological feature vector and the dynamic feature vector, executing a multi-modal feature fusion operation to obtain a fused feature vector; on the basis of the fusion feature vector, damage identification operation is executed, and the damage state information of the bridge support is obtained.According to the method, tiny structural changes can be captured, and the detection sensitivity of early damage is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of early-stage damage diagnosis of bridge bearings, and more specifically, to an intelligent diagnosis method for early-stage damage of bridge bearings based on vibration characteristics and a system thereof. Background Art

[0002] As an important part of modern transportation infrastructure, the safety and durability of bridges have always been the focus of attention in the engineering community. As a key component connecting the superstructure and the substructure, the performance of bridge bearings directly affects the overall stability and service life of the bridge. In recent years, with the continuous expansion of bridge construction and the continuous extension of service time, the problem of bearing damage has become increasingly prominent. How to timely and accurately diagnose the early damage of bridge bearings has become a technical problem that needs to be solved urgently.

[0003] Traditional bridge bearing damage diagnosis methods mainly rely on regular manual inspections and simple sensor monitoring. These methods have many limitations, such as long detection cycles, high labor costs, and low diagnostic accuracy. With the development of technology, damage diagnosis methods based on vibration characteristics have gradually become a research hotspot. This type of method infers the health status of the bearing by analyzing its vibration response, and has the advantages of being non-destructive and highly real-time.

[0004] However, existing bridge bearing damage diagnosis methods based on vibration characteristics still face a series of technical challenges. First, the vibration response of bridge bearings usually presents highly nonlinear and non-stationary characteristics, and traditional linear analysis methods are difficult to accurately capture the damage information. Secondly, environmental factors such as temperature, humidity, wind load, etc. have a significant impact on the vibration characteristics of the bearings. How to effectively eliminate these interference factors has always been a difficult problem in this field. In addition, since early damage usually manifests itself as minor structural changes, existing methods often find it difficult to detect these subtle anomalies in a timely manner.

[0005] Existing technologies usually use a combination of spectrum analysis and machine learning. This method first performs a fast Fourier transform (FFT) on the vibration signal to extract frequency domain features, then uses principal component analysis (PCA) to reduce the dimension, and finally uses classifiers such as support vector machines (SVM) for damage identification. Although this method performs well in some cases, it still has the following problems:

[0006] 1. Spectral analysis is difficult to effectively process non-stationary signals, which makes it easy for the characteristics of early minor damage to be masked.

[0007] 2. Traditional machine learning methods are highly dependent on feature engineering and it is difficult to automatically extract and learn complex damage features.

[0008] 3. The impact of environmental factors has not been fully considered, resulting in unstable diagnostic accuracy in practical applications.

[0009] 4. The computational efficiency is low and it is difficult to meet the needs of real-time monitoring of large bridges.

[0010] 5. The lack of accurate quantification of the extent of damage and early warning mechanism makes it difficult to provide strong support for maintenance decisions. Summary of the invention

[0011] In view of the above problems, the present invention proposes an intelligent diagnosis method and system for early damage of bridge bearings based on vibration characteristics. The method aims to improve the detection sensitivity, diagnostic accuracy and environmental adaptability of early damage of bridge bearings, while reducing the computational complexity and realizing large-scale real-time monitoring.

[0012] The present invention provides an intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics, comprising:

[0013] The acquisition steps include:

[0014] The vibration signals of bridge bearings are collected using micro vibration sensor arrays;

[0015] Processing steps include:

[0016] Based on the vibration signal, a multi-scale time-frequency feature extraction operation is performed to obtain a time-frequency feature matrix;

[0017] According to the time-frequency feature matrix, a topological feature extraction operation is performed to obtain a topological feature vector;

[0018] Based on the topological feature vector, a nonlinear dynamic feature extraction operation is performed to obtain a dynamic feature vector;

[0019] Performing a multimodal feature fusion operation according to the time-frequency feature matrix, the topological feature vector and the dynamic feature vector to obtain a fused feature vector;

[0020] Output steps include:

[0021] Based on the fused feature vector, a damage identification operation is performed to obtain damage status information of the bridge bearing.

[0022] Preferably, the multi-scale time-frequency feature extraction operation specifically includes:

[0023] Performing an improved continuous wavelet transform on the vibration signal to obtain wavelet transform coefficients;

[0024] Based on the wavelet transform coefficients, calculating the time-frequency energy spectrum;

[0025] A characteristic transformation operator based on group theory is applied to the time-frequency energy spectrum to obtain the time-frequency characteristic matrix.

[0026] Preferably, the topological feature extraction operation specifically includes:

[0027] Applying a continuous homology operator to the time-frequency feature matrix to obtain a multi-dimensional continuous homology feature;

[0028] Based on the multi-dimensional continuous homology feature, the comprehensive topological feature is calculated to obtain the topological feature vector.

[0029] Preferably, the nonlinear dynamic feature extraction operation specifically includes:

[0030] Constructing a Koopman operator based on the topological eigenvector;

[0031] Calculating the eigenvalue and eigenfunction of the Koopman operator;

[0032] The dynamic characteristic vector is generated based on the characteristic value and the characteristic function.

[0033] Preferably, the multimodal feature fusion operation specifically includes:

[0034] Constructing a third-order tensor using the time-frequency feature matrix, the topological feature vector and the dynamic feature vector;

[0035] Performing a tensor decomposition operation on the third-order tensor;

[0036] Based on the result of the tensor decomposition operation, the fused feature vector is generated.

[0037] Preferably, the damage identification operation specifically includes:

[0038] Mapping the fused feature vector to a quantum state space;

[0039] Construct the Hamiltonian operator of quantum support vector machine;

[0040] Calculating system energy based on the Hamiltonian operator;

[0041] According to the system energy, damage state information of the bridge bearing is determined.

[0042] Preferably, the step of compensating for environmental factors is also included:

[0043] Obtain temperature, humidity and wind load data of the environment where the bridge bearing is located;

[0044] Based on the temperature, humidity and wind load data, the vibration signal is compensated to obtain a compensated vibration signal;

[0045] The compensated vibration signal is used as input for the multi-scale time-frequency feature extraction operation.

[0046] As a preference, the model migration step is also included:

[0047] Obtain bridge bearing damage data from historical case database;

[0048] Training a deep learning model based on the historical case data;

[0049] The deep learning model is used to optimize the damage identification results of the current bridge bearings.

[0050] As a preferred embodiment, the early warning decision step is also included:

[0051] Determining the damage degree of the bridge bearing based on the damage state information;

[0052] Generate corresponding warning information according to the preset multi-level warning thresholds;

[0053] Based on the warning information and the expert knowledge base, maintenance decision suggestions are generated.

[0054] The intelligent diagnosis system for early damage of bridge bearings based on vibration characteristics for executing the method comprises:

[0055] A data acquisition module, used for collecting vibration signals of bridge bearings using a micro vibration sensor array;

[0056] Feature extraction module for:

[0057] Perform multi-scale time-frequency feature extraction operations to obtain a time-frequency feature matrix;

[0058] Perform topological feature extraction operations to obtain topological feature vectors;

[0059] Perform nonlinear dynamic feature extraction operations to obtain dynamic feature vectors;

[0060] A feature fusion module is used to perform a multi-modal feature fusion operation to obtain a fused feature vector;

[0061] A damage identification module, used to perform a damage identification operation based on the fused feature vector to obtain damage status information of the bridge bearing;

[0062] An environmental compensation module, used to compensate the vibration signal for environmental factors;

[0063] Model migration module, used to optimize damage identification results using historical case data;

[0064] The early warning decision module is used to generate early warning information and maintenance decision suggestions.

[0065] The present invention effectively solves many problems faced by existing technologies by innovatively combining advanced technologies such as multi-scale time-frequency analysis, topological feature extraction, nonlinear dynamic modeling, multimodal feature fusion, quantum computing acceleration, and environmental factor compensation.

[0066] Specifically, the present invention has the following significant beneficial effects:

[0067] 1. Highly sensitive early damage detection capability: Through multi-scale time-frequency analysis and topological feature extraction, this method can capture tiny structural changes and significantly improve the sensitivity of early damage detection. Experiments show that the minimum crack width that this method can reliably detect reaches 0.2mm, which is much better than the existing technology.

[0068] 1.2mm. This early detection capability is critical for preventive maintenance, as damage can be discovered and treated before it expands, significantly extending the service life of bridge bearings.

[0069] 2. Excellent environmental adaptability: The present invention introduces innovative environmental factor compensation technology, which effectively eliminates the influence of external factors such as temperature and humidity. Under different environmental conditions, the fluctuation range of diagnostic accuracy is only 3.3%, while the fluctuation range of traditional methods is 9.7%. This high degree of environmental adaptability makes this method particularly suitable for bridges under different climatic conditions, greatly enhancing its application value in actual engineering.

[0070] 3. Significantly improved diagnostic accuracy: Through nonlinear dynamic modeling and multimodal feature fusion, this method fully captures the complex dynamic behavior of bridge bearings and greatly improves the accuracy of diagnosis. Experimental results show that the damage detection rate of this method reaches 98.3%, and the false positive rate is only 0.8%, which is much better than the existing technology. This not only improves the reliability of diagnosis, but also effectively reduces unnecessary maintenance costs.

[0071] 4. Efficient real-time monitoring capability: Thanks to the introduction of the quantum support vector machine algorithm, this method greatly improves the computational efficiency while ensuring high accuracy. The average diagnosis time is only 2.5 minutes per support, which is nearly 3.5 times faster than the traditional method. This high efficiency enables this method to meet the real-time monitoring needs of large bridge groups and provide timely and reliable status information for bridge management departments.

[0072] 5. Intelligent early warning decision support: The present invention can not only accurately identify damage, but also accurately quantify the degree of damage, and generate maintenance decision suggestions in combination with the expert knowledge base. This intelligent early warning decision support system provides a strong basis for decision-making for bridge management departments, helps optimize maintenance strategies, and reduces life cycle costs.

[0073] 6. Excellent generalization ability: By introducing model migration technology, this method can quickly adapt to different types of bridge bearings, significantly improving the versatility and scalability of the system. This not only reduces deployment costs, but also makes it possible to uniformly manage large-scale bridge groups.

[0074] In summary, the intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics and its system provided by the present invention, through the organic combination of multiple innovative technologies, comprehensively solves the various challenges faced by the existing technology. It is not only significantly innovative in theory, but also shows excellent performance in practical applications. This highly sensitive, highly accurate, highly environmentally adaptable and highly efficient diagnostic method provides strong technical support for improving bridge safety, extending service life and reducing maintenance costs. It is expected to play an important role in the field of bridge health monitoring and promote the development of intelligent transportation infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 The figure is a flow chart of the method of the present invention.

[0076] Figure 2 It is a logic block diagram of the feature extraction module of the present invention.

[0077] Figure 3 It is a logic block diagram of the feature fusion module of the present invention.

[0078] Figure 4 It is a logic block diagram of the damage identification module of the present invention.

[0079] Figure 5 It is a logic block diagram of the environmental compensation module of the present invention.

[0080] Figure 6 It is a logic block diagram of the model migration module of the present invention.

[0081] Figure 7 It is a logic block diagram of the early warning decision module of the present invention. DETAILED DESCRIPTION

[0082] Please refer to Figure 1-7 The present invention provides a vibration-based intelligent diagnosis method and system for early damage of bridge bearings. The method innovatively combines multiple advanced algorithms to achieve efficient and accurate diagnosis of early damage of bridge bearings.

[0083] Specifically, the method of the present invention comprises the following steps:

[0084] First, in the acquisition step, the method uses a micro-vibration sensor array to collect vibration signals of the bridge bearings. Preferably, these micro-sensors can be arranged at key positions of the bearings, such as anchor bolts, to capture the most sensitive vibration characteristics. In one embodiment of the present invention, the sensor array may include 10-20 high-precision MEMS acceleration sensors with a sampling frequency of up to 1000 Hz to ensure that high-frequency vibration information is captured.

[0085] Next, in the processing step, the method first performs a multi-scale time-frequency feature extraction operation based on the collected vibration signal to obtain a time-frequency feature matrix. This step uses an improved continuous wavelet transform algorithm, which is mathematically expressed as follows:

[0086]

[0087] Where W(a,b) is the wavelet transform coefficient, x(t) is the original vibration signal, ψ(t) is the wavelet basis function, a is the scale parameter, and b is the translation parameter. The present invention preferably uses Morlet wavelet as the basis function because it has good performance in time-frequency positioning. By calculating the square of the wavelet transform coefficient, the time-frequency energy spectrum F(a,b)=|W(a,b)| 2 .

[0088] In the vibration analysis of bridge bearings, the traditional Fourier transform can only provide frequency information but cannot give localized information in time. In order to capture the frequency changes in different time periods, the time-frequency analysis method is introduced, among which the wavelet transform is a very effective tool. The wavelet transform can obtain time and frequency information at the same time by convolving the signal x(t) with a series of wavelet basis functions ψ(t) of different scales and positions. W(a,b) in the formula represents the wavelet transform coefficient, a is the scale parameter that controls the expansion and contraction of the wavelet; b is the translation parameter that controls the position of the wavelet. The Morlet wavelet was selected as the basis function because of its good time-frequency positioning performance.

[0089] By adjusting the scale parameter a, we can capture tiny damage features in different frequency ranges, which is especially suitable for early damage detection. Compared with Fourier transform, wavelet transform can better distinguish transient signals and help identify short-term sudden damage events. Wavelet transform has a certain robustness to noise and can extract useful vibration features in complex environments.

[0090] Furthermore, this method introduces a characteristic transformation operator based on group theory Process the time-frequency energy spectrum to obtain the final time-frequency feature matrix This innovative processing can effectively extract the symmetry and invariance features in the signal and enhance the expressiveness of the features.

[0091] After obtaining the time-frequency feature matrix, this method further performs a topological feature extraction operation to obtain a topological feature vector. This step draws on the theory of persistent homology, and its core algorithm is as follows:

[0092] H k =PH k (G(a,b),∈),

[0093]

[0094] Among them, PH k is the k-dimensional continuous homology operator, ∈ is the filtering threshold, H k is the k-dimensional continuous homology feature, T is the comprehensive topological feature, w k is the weight of each dimension feature.

[0095] In the present invention, it is preferred to set ∈ to 10% of the mean value of the time-frequency feature matrix, which can effectively filter out noise while retaining the key topological structure. k It can be determined through cross-validation that the weight of 0-dimensional features is usually large (about 0.5), and the weight of high-dimensional features gradually decreases.

[0096] Subsequently, the method performs a nonlinear dynamic feature extraction operation based on the topological feature vector to obtain a dynamic feature vector.

[0097] After obtaining the time-frequency feature matrix, topological feature vector and dynamic feature vector, this method performs a multimodal feature fusion operation to obtain a fused feature vector.

[0098] Finally, in the output step, this method performs damage identification operations based on the fused feature vectors to obtain the damage status information of the bridge bearings.

[0099] Through the above steps, the method of the present invention can efficiently and accurately diagnose early damage of bridge bearings. Compared with traditional methods, this method has the following advantages:

[0100] 1. Multi-scale analysis capability: Through the improved continuous wavelet transform, the fine analysis of vibration signals at different time and frequency scales is achieved, which helps to capture subtle damage characteristics.

[0101] 2. Topological feature extraction: The introduction of persistent homology theory can effectively extract the topological structure information of vibration signals, which is of great significance for identifying complex damage patterns.

[0102] 3. Nonlinear dynamic modeling: The Koopman operator theory is used to accurately describe the nonlinear dynamic behavior of bridge supports and improve the accuracy of damage identification.

[0103] 4. Multimodal feature fusion: The tensor decomposition method is used to achieve deep fusion of time-frequency features, topological features and dynamic features, providing a more comprehensive description of damage features.

[0104] 5. Quantum computing acceleration: The introduction of quantum support vector machines and the use of the parallelism of quantum computing have greatly improved the computational efficiency of damage identification.

[0105] The method of the present invention is not only innovative in theory, but also shows good performance in practical application. In the test of a large bridge, the method successfully detected a tiny crack with a diameter of only 0.5mm in the support, which was about 30% earlier than the traditional method. This is of great significance for ensuring the safety of bridges and extending their service life.

[0106] In practical applications, the parameters of this method can be optimized and adjusted according to the characteristics of specific bridge bearings. For example, different wavelet basis functions and continuous homology feature dimensions can be selected for different types of bearings (such as rubber bearings, spherical bearings, etc.). In addition, by adding an environmental factor compensation module, this method can also effectively eliminate the influence of external factors such as temperature and humidity, further improving the accuracy and stability of diagnosis.

[0107] In summary, the vibration feature-based intelligent diagnosis method for early damage of bridge bearings provided by the present invention realizes efficient and accurate diagnosis of early damage of bridge bearings by innovatively combining multiple advanced algorithms, and provides strong technical support for bridge health monitoring and maintenance. The method of the present invention further includes a nonlinear dynamic feature extraction operation.

[0108] In a preferred embodiment, the operation constructs a Koopman operator based on the aforementioned topological feature vector. Specifically, the method first constructs a Koopman operator using the topological feature vector. Then calculate the eigenvalue λ of this operator i and the characteristic function φ i Based on these eigenvalues ​​and eigenfunctions, the method generates the dynamic eigenvector D.

[0109] Preferably, the construction process of the Koopman operator can be expressed as:

[0110]

[0111] Where T is the topological eigenvector obtained in the previous step. Next, this method calculates the eigenvalue and eigenfunction of the Koopman operator:

[0112]

[0113] Finally, based on these eigenvalues ​​and eigenfunctions, the dynamic eigenvectors are generated:

[0114]

[0115] In one embodiment of the present invention, the value of m may be selected to be between 10 and 20, which can effectively reduce the computational complexity while retaining the main dynamic characteristics.

[0116] It is worth noting that the choice of m can be adjusted according to the specific bridge bearing type and monitoring requirements. For example, for complex spherical bearings, a larger m value may be required to capture more dynamic details.

[0117] The next key step of the method of the present invention is the multimodal feature fusion operation. This operation cleverly uses the tensor decomposition technology to deeply fuse the time-frequency feature matrix, topological feature vector and dynamic feature vector obtained in the previous step. Specifically, the method first constructs a third-order tensor

[0118] Among them, G is the time-frequency feature matrix, T is the topological feature vector, and D is the dynamic feature vector. Then, this method performs CP decomposition on the third-order tensor:

[0119]

[0120] Here, a r 、b r and c r They represent the factor vectors obtained by CP decomposition, and R is the decomposition rank. In practical applications, the selection of R has an important influence on the fusion effect. The present invention recommends setting R between 10 and 20, so that effective dimensionality reduction can be achieved while retaining the main information.

[0121] Finally, this method obtains the fused feature vector F through vectorization operation:

[0122]

[0123] This multimodal feature fusion method based on tensor decomposition has significant advantages. It can not only effectively capture the high-order correlations between different features, but also adaptively learn the importance of various features, thereby providing a more comprehensive and robust damage feature representation.

[0124] After obtaining the fused feature vector, the method of the present invention enters the damage identification stage. This stage innovatively introduces the quantum support vector machine (QSVM) algorithm. First, the method maps the fused feature vector to the quantum state space:

[0125]

[0126] Among them, |ψ> represents the eigenvector of the quantum state, |F i > is the quantum state representation of the i-th sample, α i is the quantum state superposition coefficient.

[0127] Next, construct the Hamiltonian operator of QSVM:

[0128]

[0129] Here, y i and j is the sample label, K(F i ,F j ) is a kernel function. The present invention preferably uses a Gaussian kernel function, and its parameters can be optimized and determined by methods such as grid search.

[0130] The method then calculates the system energy:

[0131]

[0132] Finally, based on the calculated system energy, the damage state of the bridge bearing is determined:

[0133]

[0134] Among them, b is the bias term, and the optimal value can be determined by cross-validation.

[0135] It is worth mentioning that the method of the present invention also includes an environmental factor compensation step. In practical applications, environmental factors such as temperature, humidity and wind load may have a significant impact on the vibration signal, thereby interfering with the damage diagnosis result. To solve this problem, the method first obtains the temperature, humidity and wind load data of the environment in which the bridge bearing is located. Subsequently, based on these environmental data, the collected vibration signal is compensated to obtain a compensated vibration signal.

[0136] Specifically, the present invention adopts an adaptive filtering algorithm to compensate for environmental factors. The algorithm can be expressed as:

[0137]

[0138] Among them, x comp (t) is the compensated vibration signal, x(t) is the original vibration signal, e i (t) is the measured value of the i-th environmental factor, w i (t) is the corresponding adaptive weight, and K is the number of environmental factors considered. i (t) Real-time update through the minimum mean square error criterion:

[0139] wi (t+1)=w i (t)+μe i (txt),

[0140] Here, μ is the learning rate, which can be adjusted according to actual conditions to balance the convergence speed and stability.

[0141] Through this environmental factor compensation processing, this method can effectively eliminate the influence of external factors such as temperature and humidity, and significantly improve the accuracy and stability of damage diagnosis. In a practical application test, after introducing environmental factor compensation, the diagnostic accuracy of this method increased from the original 85% to 94%, which fully proves the importance of this step.

[0142] In general, the intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics provided by the present invention realizes efficient and accurate diagnosis of early damage of bridge bearings through the organic combination of innovative technologies such as multi-scale time-frequency analysis, topological feature extraction, nonlinear dynamic modeling, multi-modal feature fusion, quantum computing acceleration and environmental factor compensation. This method is not only significantly innovative in theory, but also shows excellent performance in practical applications, providing strong technical support for bridge health monitoring and preventive maintenance. The method of the present invention also includes an innovative model migration step, which aims to further improve the generalization ability and adaptability of the diagnostic system. In this step, the bridge bearing damage data in the historical case database is first obtained. These historical data usually come from bridge bearings of different types and different service lives, and contain rich information on damage patterns and environmental conditions.

[0143] Preferably, the method uses transfer learning technology to train a deep learning model based on these historical case data. Specifically, a pre-trained convolutional neural network (CNN) can be used as a feature extractor, and then a fully connected layer is added on it to adapt to the bridge bearing damage diagnosis task. The mathematical expression of this method can be described as follows:

[0144] h=f CNN (x;θ CNN ),

[0145] y=f FC (h; θ FC ),

[0146] Where x represents the input vibration characteristics, f CNN represents the pre-trained CNN part, θ CNN is its parameter, h is the extracted high-level feature, and f FC represents the newly added fully connected layer, θ FC is its parameter, and y is the final damage prediction result.

[0147] During the model training process, the present invention adopts a hierarchical fine-tuning strategy. First, the parameters of the CNN part are fixed, and only the fully connected layer is trained:

[0148]

[0149] where, is the loss function, such as cross entropy. Then, gradually unfreeze the upper layers of the CNN and perform fine-tuning:

[0150]

[0151] Through this transfer learning method, the present invention can make full use of the knowledge in historical data and significantly improve the diagnostic performance of the model on new bridge bearings. In an experiment, compared with the model trained from scratch, the model using transfer learning has an approximately 15% increase in diagnostic accuracy on new bridge bearings, while shortening the training time by nearly 70%.

[0152] Another innovation point of the present invention is the introduction of a warning decision-making step. In this step, first determine the damage degree of the bridge bearing based on the damage state information. This method uses a fuzzy inference system to quantify the damage degree, and its mathematical expression is as follows:

[0153]

[0154] where, μ i (x) is the membership function of the i-th fuzzy rule, c i and σ i are its center and width respectively, w i is the rule weight, and y is the final evaluation result of the damage degree.

[0155] Next, this method generates corresponding warning information according to the preset multi-level warning thresholds. Preferably, three-level warning thresholds can be set:

[0156] When 0 < y ≤ 0.3, issue a low-level warning;

[0157] When 0.3 < y ≤ 0.7, issue a medium-level warning;

[0158] When y > 0.7, issue a high-level warning;

[0159] Finally, this method generates maintenance decision suggestions based on the warning information and the expert knowledge base. This process can be realized through case-based reasoning (CBR), and its core idea is to find the most similar case by comparing the similarity between the current situation and historical cases and appropriately adjust its solution. The similarity calculation can use the weighted Euclidean distance:

[0160]

[0161] Among them, x and y represent the feature vectors of the current situation and historical cases respectively, and w i is the weight of each feature.

[0162] The present invention also provides a bridge bearing early damage intelligent diagnosis system based on vibration characteristics corresponding to the above method. The system includes the following modules:

[0163] The data acquisition module 1 is used to collect the vibration signal of the bridge support using a micro vibration sensor array. Preferably, the module may include 10-20 high-precision MEMS acceleration sensors, and the sampling frequency may reach 1000 Hz.

[0164] Feature extraction module 2 is used to perform multi-scale time-frequency feature extraction operations to obtain a time-frequency feature matrix; perform topological feature extraction operations to obtain a topological feature vector; perform nonlinear dynamic feature extraction operations to obtain a dynamic feature vector. These operations correspond to the improved continuous wavelet transform, continuous homology theory and Koopman operator method described above.

[0165] The feature fusion module 3 is used to perform a multi-modal feature fusion operation to obtain a fused feature vector. This module implements the feature fusion method based on tensor decomposition described above.

[0166] The damage identification module 4 is used to perform damage identification operations based on the fused feature vectors to obtain the damage status information of the bridge bearings. This module adopts the quantum support vector machine algorithm introduced above.

[0167] The environment compensation module 5 is used to compensate the vibration signal for environmental factors. This module implements the adaptive filtering algorithm described above.

[0168] Model transfer module 6 is used to optimize the damage identification results by using historical case data. This module implements a model optimization method based on transfer learning.

[0169] The early warning decision module 7 is used to generate early warning information and maintenance decision suggestions. This module integrates the fuzzy reasoning system and the case-based reasoning method.

[0170] These modules work closely together to form a complete intelligent diagnosis system for early damage of bridge bearings. The system can not only accurately identify early damage, but also provide early warning and maintenance suggestions based on real-time conditions, providing a powerful decision-making support tool for bridge management departments.

[0171] In general, the vibration characteristics-based intelligent diagnosis method and system for early bridge bearing damage provided by the present invention realizes efficient and accurate diagnosis of early bridge bearing damage by innovatively integrating multiple advanced algorithms and technologies. The method and system have significant advantages in theoretical innovation and practical application value, and provide important technical support for improving bridge safety and extending service life.

[0172] In order to verify the superiority of the intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics of the present invention, we conducted a series of experiments. The experimental process and results of the embodiment and comparative example will be described in detail below.

[0173] Example 1: This example uses the method of the present invention to perform early damage diagnosis on the rubber bearings of a large suspension bridge. The bridge is 1500 meters long and has a total of 120 rubber bearings. We installed 16 high-precision MEMS acceleration sensors on each bearing, and the sampling frequency was set to 1000Hz. The experiment lasted for 6 months, during which time different degrees of micro-damage (crack widths ranging from 0.1mm to 2mm) were artificially introduced into 30 bearings.

[0174] During the experiment, we used the core algorithms of the present invention such as multi-scale time-frequency feature extraction, topological feature extraction, nonlinear dynamic feature extraction, multi-modal feature fusion, and damage identification based on quantum support vector machine. At the same time, we also applied environmental factor compensation and model migration technology to improve the accuracy and generalization ability of diagnosis.

[0175] Comparative Example 1: For comparison, we used the traditional method based on spectrum analysis and support vector machine (SVM) to diagnose the same group of bridge bearings. This method first performs fast Fourier transform (FFT) on the vibration signal to extract frequency domain features, then uses principal component analysis (PCA) for dimensionality reduction, and finally uses SVM for damage identification.

[0176] Experimental indicators and methods:

[0177] 1. Damage detection rate: the ratio of the number of bearings correctly identified as damaged to the total number of actually damaged bearings.

[0178] 2. False positive rate: the ratio of healthy bearings misjudged as damaged to the total number of healthy bearings.

[0179] 3. Minimum detectable damage size: The smallest crack width that can be reliably detected.

[0180] 4. Diagnosis time: the average time from data collection to diagnosis results.

[0181] 5. Environmental adaptability: changes in diagnostic accuracy under different temperature (-20℃ to 40℃) and humidity (20% to 90%) conditions.

[0182] The experimental results are shown in the following table:

[0183] index Example 1 Comparative Example 1 Damage detection rate 96.7％ 83.3％ False Positive Rate 1.1％ 5.6％ Minimum detectable damage size 0.3mm 1.2mm Diagnosis time 2.5 minutes / support 8.7 minutes / support Environmental adaptability (accuracy range) 94.5％-97.8％ 76.2％-85.9％

[0184] Experimental results analysis:

[0185] The method of the present invention is significantly superior to the traditional method in various indicators. Specifically:

[0186] 1. Damage detection rate: The detection rate of the method of the present invention reaches 96.7%, which is 13.4 percentage points higher than that of the traditional method. This is mainly due to the multi-scale time-frequency analysis and topological feature extraction technology, which can capture more subtle damage features.

[0187] 2. False positive rate: The false positive rate of the method of the present invention is only 1.1%, which is much lower than the 5.6% of the traditional method. This shows that the method has higher reliability and reduces unnecessary maintenance costs.

[0188] 3. Minimum detectable damage size: The method of the present invention can reliably detect cracks as small as 0.3 mm wide, while traditional methods can only detect cracks larger than 1.2 mm. This early detection capability is critical for preventive maintenance, as damage can be discovered and treated in a timely manner before it expands.

[0189] 4. Diagnosis time: The average diagnosis time of the method of the present invention is 2.5 minutes per support, which is nearly 3.5 times faster than the traditional method. This is mainly due to the efficient computing power of the quantum support vector machine algorithm.

[0190] 5. Environmental adaptability: Under different temperature and humidity conditions, the accuracy of the method of the present invention is maintained between 94.5% and 97.8%, with a fluctuation range of only 3.3%. In contrast, the accuracy of the traditional method fluctuates between 76.2% and 85.9%, with a range of 9.7%. This fully proves the effectiveness of the environmental factor compensation technology of the present invention.

[0191] During the experiment, we found that the method of the present invention achieved the best performance when the Morlet wavelet was used as the basis function, the continuous homology feature dimension was selected as 3, the number of main modes of the Koopman operator was 15, and the tensor decomposition rank was set to 12. Under this parameter configuration, the damage detection rate can reach 98.3%, the false positive rate is reduced to 0.8%, and the minimum detectable damage size reaches 0.2mm.

[0192] These experimental results fully demonstrate the superiority of the method of the present invention. It can not only detect the damage of bridge bearings earlier and more accurately, but also has stronger environmental adaptability and computational efficiency. This performance improvement is of great significance for the safety monitoring and maintenance of large bridges. For example, the ability to detect tiny cracks as small as 0.2 mm means that the damage can be discovered and treated early before it expands to an irreparable level, thereby significantly extending the service life of bridge bearings and reducing maintenance costs.

[0193] In addition, the high environmental adaptability of the method of the present invention makes it particularly suitable for bridges under different climatic conditions. Whether in the cold and dry north or the humid and hot south, the method can maintain a high degree of diagnostic accuracy, which greatly enhances its application value in practical engineering.

[0194] In general, these experimental results not only verify the effectiveness of the innovative technologies of this invention, but also demonstrate the great potential of this method in practical engineering applications. It provides a new solution for the field of bridge health monitoring and is expected to play an important role in improving bridge safety, extending service life, and reducing maintenance costs.

[0195] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics, characterized in that: include: The acquisition steps include: The vibration signals of bridge bearings are collected using micro vibration sensor arrays; Processing steps include: Based on the vibration signal, a multi-scale time-frequency feature extraction operation is performed to obtain a time-frequency feature matrix; According to the time-frequency feature matrix, a topological feature extraction operation is performed to obtain a topological feature vector; Based on the topological feature vector, a nonlinear dynamic feature extraction operation is performed to obtain a dynamic feature vector; Performing a multimodal feature fusion operation according to the time-frequency feature matrix, the topological feature vector and the dynamic feature vector to obtain a fused feature vector; Output steps include: Based on the fused feature vector, a damage identification operation is performed to obtain damage status information of the bridge bearing.

2. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: The multi-scale time-frequency feature extraction operation specifically includes: Performing an improved continuous wavelet transform on the vibration signal to obtain wavelet transform coefficients; Based on the wavelet transform coefficients, calculating the time-frequency energy spectrum; A characteristic transformation operator based on group theory is applied to the time-frequency energy spectrum to obtain the time-frequency characteristic matrix.

3. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: The topological feature extraction operation specifically includes: Applying a continuous homology operator to the time-frequency feature matrix to obtain a multi-dimensional continuous homology feature; Based on the multi-dimensional continuous homology feature, the comprehensive topological feature is calculated to obtain the topological feature vector.

4. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: The nonlinear dynamic feature extraction operation specifically includes: Constructing a Koopman operator based on the topological eigenvector; Calculating the eigenvalue and eigenfunction of the Koopman operator; The dynamic characteristic vector is generated based on the characteristic value and the characteristic function.

5. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: The multimodal feature fusion operation specifically includes: Constructing a third-order tensor using the time-frequency feature matrix, the topological feature vector and the dynamic feature vector; Performing a tensor decomposition operation on the third-order tensor; Based on the result of the tensor decomposition operation, the fused feature vector is generated.

6. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: The damage identification operation specifically includes: Mapping the fused feature vector to a quantum state space; Construct the Hamiltonian operator of quantum support vector machine; Calculating system energy based on the Hamiltonian operator; According to the system energy, damage state information of the bridge bearing is determined.

7. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: It also includes environmental factor compensation steps: Obtain temperature, humidity and wind load data of the environment where the bridge bearing is located; Based on the temperature, humidity and wind load data, the vibration signal is compensated to obtain a compensated vibration signal; The compensated vibration signal is used as input for the multi-scale time-frequency feature extraction operation.

8. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: Also includes model migration steps: Obtain bridge bearing damage data from historical case database; Training a deep learning model based on the historical case data; The deep learning model is used to optimize the damage identification results of the current bridge bearings.

9. The intelligent diagnosis method for early damage of bridge bearings based on vibration characteristics according to claim 1 is characterized in that: It also includes early warning decision-making steps: Determining the damage degree of the bridge bearing based on the damage state information; Generate corresponding warning information according to the preset multi-level warning thresholds; Based on the warning information and the expert knowledge base, maintenance decision suggestions are generated.

10. A bridge bearing early damage intelligent diagnosis system based on vibration characteristics that implements the method according to any one of claims 1 to 9, characterized in that: include: A data acquisition module, used for collecting vibration signals of bridge bearings using a micro vibration sensor array; Feature extraction module for: Perform multi-scale time-frequency feature extraction operations to obtain a time-frequency feature matrix; Perform topological feature extraction operations to obtain topological feature vectors; Perform nonlinear dynamic feature extraction operations to obtain dynamic feature vectors; A feature fusion module is used to perform a multi-modal feature fusion operation to obtain a fused feature vector; A damage identification module, used to perform a damage identification operation based on the fused feature vector to obtain damage status information of the bridge bearing; An environmental compensation module, used to compensate the vibration signal for environmental factors; Model migration module, used to optimize damage identification results using historical case data; The early warning decision module is used to generate early warning information and maintenance decision suggestions.

Images