Multivariate data acquisition and fusion analysis method and system for structural health monitoring

Abstract

The invention provides a multivariate data acquisition and fusion analysis method and system for structural health monitoring. The method comprises the following steps: acquiring target building or structure monitoring data by using an integrated multivariate sensor and transmitting the target building or structure monitoring data; processing the monitoring data; establishing an LSTM neural network modal parameter prediction model and an LSTM neural network correction dip angle prediction model, and performing training; and determining a modal parameter sequence and a corrected inclination angle sequence for prediction, respectively inputting the modal parameter prediction model and the corrected inclination angle prediction model of the LSTM neural network to obtain a modal parameter prediction sequence and an inclination angle prediction sequence, and carrying out trend analysis and early warning. According to the method, the influence of environmental parameters on the parameters of the measured structure and the prediction result is effectively eliminated, the measured data of buildings or structures under different conditions are trained and fitted by using the LSTM neural network, and the nonlinear evolution characteristics of the time series data of the measured structure can be expressed with high precision.

Description

Technical Field

[0001] The present invention relates to the field of building health and safety, and in particular to a multi-data collection and fusion analysis method and system for structural health monitoring. Background Art

[0002] With the increasing number and usage time of existing buildings and structures in cities, the probability of structural risks is increasing. Real-time perception and early prediction of potential structural risks are effective means to reduce the operation and maintenance costs of existing buildings and structures. Improving the prediction accuracy of structural monitoring data in different scenarios is of great practical significance for preventing and reducing the losses caused by structural risks.

[0003] At present, a large number of structural health and safety monitoring systems are often used to conduct risk assessment on a single structural indicator. However, a large number of studies have shown that environmental factors have a significant impact on monitoring data. In the original monitoring data of structural response, the structural response caused by environmental variables such as temperature and the changes in structural parameters such as inclination may be at the same order of magnitude, and the original monitoring data may not reflect the true change trend of structural parameters. If this environmental impact is ignored, the accuracy of structural safety assessment results may decrease. Therefore, it is necessary to propose an analysis method, system and equipment to perform fusion analysis on multivariate health monitoring data to achieve multi-indicator early warning of structural safety. Summary of the invention

[0004] In view of the defects in the prior art, the object of the present invention is to provide a multi-data collection and fusion analysis method and system for structural health monitoring.

[0005] According to one aspect of the present invention, a multivariate data collection and fusion analysis method for structural health monitoring is provided, comprising:

[0006] Using integrated multi-sensors to obtain monitoring data of target buildings or structures, the monitoring data including acceleration monitoring data, inclination monitoring data, and temperature monitoring data;

[0007] Transmitting the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data;

[0008] Performing structural modal parameter identification on the acceleration monitoring data to obtain modal parameter identification results, and obtaining modal parameter historical monitoring data; processing the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data;

[0009] Establishing an LSTM neural network modal parameter prediction model, and using the modal parameter historical monitoring data as training samples for training; establishing an LSTM neural network corrected inclination prediction model, and using the corrected inclination historical monitoring data as training samples for training;

[0010] Obtain a modal parameter sequence to be predicted, input the modal parameter sequence into a trained LSTM neural network modal parameter prediction model, and output a modal parameter prediction sequence of a target building or structure; obtain a corrected inclination angle sequence to be predicted, input the corrected inclination angle sequence into a trained LSTM neural network inclination angle prediction model, and output an inclination angle prediction sequence of a target building or structure;

[0011] Trend determination and early warning are performed based on the predicted modal parameter prediction sequence and the inclination angle prediction sequence.

[0012] Preferably, the use of integrated multi-sensors to obtain the monitoring data of the target building or structure is to use sensors for regular collection;

[0013] When the sensor performs a single acquisition, the collected data includes acceleration time history, single or multiple inclination data, and single or multiple temperature data collected at a fixed sampling frequency within a fixed time period.

[0014] Preferably, the SSI method is used to perform modal parameter identification on the acceleration monitoring data of the target building or structure, and the modal parameters identified include: modal frequency, modal vibration shape and modal damping;

[0015] A stability diagram is also generated during the identification process, wherein the stability diagram includes a stability diagram whose abscissa is the modal frequency and whose ordinate is the modal order;

[0016] Based on the stability diagram, unstable frequencies in the identified modal parameters are filtered to obtain frequency-stable modal frequencies, modal vibration shapes and modal damping.

[0017] Preferably, the processing of the inclination monitoring data and the temperature monitoring data to obtain the corrected inclination historical monitoring data and the temperature historical monitoring data includes:

[0018] Normalizing the inclination data and the temperature data and constructing a database to obtain inclination historical monitoring data and temperature historical monitoring data;

[0019] The XGBoost method is used to perform segmented component decomposition on the historical monitoring data of the inclination angle, and linear fitting is used to obtain the trend item in the component decomposition;

[0020] The temperature historical monitoring data is regarded as an exogenous variable, and the least squares fitting is adopted to obtain the exogenous term caused by the exogenous variable;

[0021] The trend term and the exogenous term are removed from the inclination historical monitoring data to obtain the corrected inclination historical monitoring data.

[0022] Preferably, the normalization process is specifically: E[x] is the mean of the data sample, and Var[x] is the variance of the data sample.

[0023] Preferably, the training process of the LSTM neural network modal parameter prediction model is the same as that of the LSTM neural network inclination angle prediction model, and both include the following process:

[0024] Determine the neural network hyperparameters based on the number of input variables;

[0025] Determine the output variable of the LSTM neural network modal parameter prediction model as a modal parameter prediction sequence; determine the output variable of the LSTM neural network inclination angle prediction model as an inclination angle prediction sequence;

[0026] Training is performed by minimizing the difference between the modal parameter prediction value / minimizing the inclination angle prediction value and the measured value to obtain the parameters of each neuron node of the LSTM neural network modal parameter prediction model / LSTM neural network inclination angle prediction model.

[0027] Preferably, the trend determination and early warning based on the modal parameter prediction sequence and the inclination angle prediction sequence obtained by prediction include:

[0028] Obtaining an inclination trend based on the modal parameters and the inclination prediction value;

[0029] According to the set upper and lower limit thresholds of modal parameters and inclination trend, judge whether the modal parameters and inclination trend exceed the limit:

[0030] An early warning is issued when the inclination angle trend or at least one modal parameter exceeds the limit; or, an early warning is issued when both the inclination angle trend and at least one modal parameter exceed the limit.

[0031] According to a second aspect of the present invention, there is provided a multi-data collection and fusion analysis system for structural health monitoring, comprising:

[0032] A data acquisition module, which uses an integrated multi-sensor to acquire monitoring data of a target building or structure, wherein the monitoring data includes acceleration monitoring data, inclination monitoring data, and temperature monitoring data;

[0033] A data transmission module, for transmitting the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data;

[0034] A data processing module performs structural modal parameter identification on the acceleration monitoring data to obtain modal parameter identification results, and obtains modal parameter historical monitoring data; processes the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data;

[0035] A prediction model training module is used to establish an LSTM neural network modal parameter prediction model, and the modal parameter historical monitoring data is used as a training sample for training; an LSTM neural network corrected inclination prediction model is established, and the corrected inclination historical monitoring data is used as a training sample for training;

[0036] The settlement prediction module obtains a modal parameter sequence to be predicted, inputs the modal parameter sequence into a trained LSTM neural network modal parameter prediction model, and outputs a modal parameter prediction sequence of a target building or structure; obtains a corrected inclination angle sequence to be predicted, inputs the corrected inclination angle sequence into a trained LSTM neural network inclination angle prediction model, and outputs an inclination angle prediction sequence of a target building or structure;

[0037] The early warning module performs trend determination and early warning based on the predicted modal parameter prediction sequence and the inclination angle prediction sequence.

[0038] According to a third aspect of the present invention, there is provided a terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can be used to execute the method described, or to run the system described, when executing the program.

[0039] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to execute the method described, or to run the system described.

[0040] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0041] 1. The multivariate data collection and fusion analysis method and system for structural health monitoring in the embodiment of the present invention effectively removes the influence of environmental parameters on the measured structural parameters and prediction results by processing and correcting the acceleration monitoring data, inclination monitoring data, and temperature monitoring data. The LSTM neural network is used to train and fit the measured data of buildings or structures under different conditions, and the nonlinear evolution characteristics of the measured structure time series data can be expressed with high precision.

[0042] 2. The multivariate data collection and fusion analysis method and system for structural health monitoring in the embodiments of the present invention comprehensively consider the different characteristics of the target buildings or structures and the actual environmental conditions, and can accurately predict the structural health and safety indicator data under different site conditions. The interactive warning information it provides can provide relevant operation and maintenance personnel with a low-cost, high-efficiency, and high-precision real-time structural safety management solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0044] Figure 1 It is a schematic flow chart of a multi-data collection and fusion analysis method for structural health monitoring according to an embodiment of the present invention;

[0045] Figure 2 A stability diagram of a preferred embodiment of the present invention, wherein the horizontal axis is the modal frequency and the vertical axis is the modal order;

[0046] Figure 3 Schematic diagram of the internal structure of a neural network model according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several variations and improvements may be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0048] With the development of integrated small multi-sensors and the increasing innovation of artificial intelligence technology and their extensive application in the early perception of building risk sources, it is possible to simultaneously collect structural data and environmental parameters in different scenarios. Based on this, the embodiment of the present invention provides a multi-data collection and fusion analysis method for structural health monitoring, which is suitable for the prediction and evaluation of the structural health and safety of different types of buildings or structures. Figure 1 , the method comprising:

[0049] Step 1: Use integrated multi-sensors to obtain monitoring data of target buildings or structures, including acceleration monitoring data, inclination monitoring data, and temperature monitoring data;

[0050] Step 2: Transmit the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data;

[0051] Step three, perform structural modal parameter identification on the acceleration monitoring data, obtain the first modal parameter identification result of the target building or structure, and obtain modal parameter historical monitoring data; process the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data;

[0052] Step 4: Establish an LSTM neural network modal parameter prediction model based on the historical monitoring data of modal parameters, use the historical monitoring data of modal parameters as training samples, and train the LSTM neural network modal parameter prediction model; establish an LSTM neural network corrected inclination prediction model based on the corrected inclination historical monitoring data, use the corrected inclination historical monitoring data as training samples, and train the LSTM neural network inclination prediction model;

[0053] Step 5: determine the modal parameter sequence for prediction, input the modal parameter sequence into the trained LSTM neural network modal parameter prediction model, and output the modal parameter prediction sequence of the target building or structure; determine the corrected inclination angle sequence for prediction, input the corrected inclination angle sequence into the trained LSTM neural network inclination angle prediction model, and output the inclination angle prediction sequence of the target building or structure;

[0054] Step six: performing trend determination and early warning on the predicted modal parameter prediction sequence and inclination angle prediction sequence.

[0055] The embodiment of the present invention comprehensively analyzes multi-sensory data such as historical modal parameter information, historical inclination information and historical temperature information, removes environmental impacts, and uses an LSTM neural network to train and fit the measured structural parameters of buildings or structures under different conditions, such as modal parameters and inclination data. The nonlinear evolution characteristics of the measured structural parameter time series data can be expressed with high precision, thereby improving the accuracy of structural parameter prediction in different scenarios.

[0056] In a preferred embodiment of the present invention, step one can obtain historical monitoring data through a data acquisition network. The data acquisition network comes from an integrated miniaturized multi-sensor, which includes an acceleration acquisition module, an inclination acquisition module, and a temperature acquisition module to collect the above data, and uses an integrated sensor transmission module to realize the unified timed transmission of multi-sensor data; wherein, the acceleration data collected once is composed of the acceleration time course collected at a fixed sampling frequency within a fixed time period, and the inclination and temperature data collected once are composed of single or multiple collected data.

[0057] In some other embodiments, the data acquisition network may also be composed of different measuring devices, including accelerometers, inclinometers, thermometers and hygrometers and other devices arranged on the target building or structure. The data collected by the above-mentioned acquisition devices are collected and transmitted simultaneously by a unified gateway and a timing module arranged on the gateway.

[0058] In a preferred embodiment of the invention, the data transmission method in step two includes an on-site distributed wired transmission network composed of a sensor data acquisition terminal, a distributed gateway, a public network data transmission protocol, a cloud or local data receiving port, etc., to realize data transmission of a sensor network that uses a wired data transmission method for data transmission.

[0059] In some other embodiments, the data transmission method can also adopt an on-site distributed wireless transmission network composed of a sensor wireless sending device, a wireless transmission protocol (Internet of Things, WiFi, Lora, ZigBee, etc.), a distributed gateway, a public network data transmission protocol, a cloud or local data receiving port, etc., to realize data transmission of a sensor network that adopts a wireless data transmission method for data transmission.

[0060] In order to achieve high-precision expression of the nonlinear evolution characteristics of the measured structure time series data, it is necessary to effectively process the obtained data. In a preferred embodiment of the present invention, the random subspace method SSI is used to identify the structural modal parameters of the acceleration monitoring data, and the identified modal parameters include: modal frequency, modal vibration shape and modal damping. In addition, in the process of using the SSI algorithm to identify the modal parameters of the acceleration time series, a stability diagram is also generated, such as Figure 2 As shown, the stability diagram includes a stability diagram with the horizontal axis being the modal frequency and the vertical axis being the modal order. The stable frequency is identified based on the stability diagram, and the unstable frequency caused by the noise interference of the original data is filtered out, so as to identify the real mode of the target building or structure, and obtain the modal frequency, vibration shape and damping identification results of the stable frequency as the input data of the structural modal parameter prediction model in step four.

[0061] In the above embodiment, the acceleration monitoring data is processed. Similarly, in another embodiment of the present invention, the inclination monitoring data and the temperature monitoring data are also processed to obtain the corrected inclination historical monitoring data and the temperature historical monitoring data, which mainly include two parts:

[0062] First, considering the dimensional differences between multivariate data, it is necessary to normalize the collected multi-source data to achieve better neural network training results. The normalized data can be expressed as: E[x ] is the sample mean of the data, Var[x] is the sample variance of the data, and after normalization, variables of different dimensions can be processed uniformly.

[0063] Second, the XGBoost algorithm is used to perform segmented component decomposition on the historical monitoring data of inclination. The historical monitoring data of temperature is regarded as exogenous environmental variables, and the possible data linear drift caused by sensor hardware is considered. The purpose is to remove the components related to environmental variables and linear data drift in the original monitoring data to obtain the corrected true structural inclination. The trend term caused by linear drift of data in the component decomposition is fitted linearly, and the exogenous term caused by exogenous environmental variables is fitted least squares. The trend term and exogenous term of the historical monitoring data of inclination are removed to obtain the corrected historical monitoring data of inclination, thereby improving the prediction accuracy of the model.

[0064] After obtaining the processed data, these data can be used as input for model training. In a preferred embodiment of the present invention, step 4 is implemented to use the historical monitoring data of modal parameters as training samples to implement the training of the LSTM neural network modal parameter prediction model; and the historical monitoring data of the corrected inclination angle is used as training samples to implement the training of the LSTM neural network inclination angle prediction model.

[0065] In a preferred embodiment, the training process of the LSTM neural network modal parameter prediction model is as follows:

[0066] Determine the neural network hyperparameters based on the number of input variables;

[0067] The output variable of the LSTM neural network modal parameter prediction model is a modal parameter prediction sequence;

[0068] The training is performed by minimizing the difference between the predicted value and the measured value of the modal parameters to obtain the parameters of each neuron node of the LSTM neural network modal parameter prediction model.

[0069] In a preferred implementation, the LSTM neural network inclination prediction model is trained in the following steps:

[0070] Determine the neural network hyperparameters based on the number of input variables;

[0071] The output variable of the LSTM neural network inclination prediction model is the inclination prediction sequence;

[0072] The training is performed by minimizing the difference between the predicted inclination value and the measured value to obtain the parameters of each neuron node of the LSTM neural network inclination prediction model.

[0073] Through the training process of the above embodiment, a trained LSTM neural network inclination prediction model and an LSTM neural network inclination prediction model are obtained. The LSTM neural network divides the input time series data into sub-units of the same time slice for processing, and extracts and characterizes the information of the front sub-unit through the learning iteration of the input gate vector, the forget gate vector, and the output gate vector.

[0074] Further, such as Figure 3 The figure shows the internal structure of each LSTM neural network prediction model. The LSTM neural network realizes the transmission of information in time sequence through its chain structure, and is trained and learned through the update of gate vectors and hidden node states. It is suitable for processing and predicting time series data. The neural network divides the time series data in the historical data sequence into sub-units of the same time slice for processing. The tth time slice sub-unit processes the tth input data. Consider the unit state C transmitted by the sub-unit in the t-1th time slice t-1 With the hidden node state h t-1 , and the unit state C calculated by this subunit t With the hidden node state h t Pass to the next subunit.

[0075] For the tth input data and the hidden node state h transmitted by the t-1th time slice sub-unit t-1 In this embodiment, the neural network model obtains the forget gate vector f through formula (1) t To represent the degree of retention of the information learned by the previous subunit in this subunit, each element of which is within the range of [0,1], f t 0 means "completely forgotten", f t 1 represents “completely retained”. f and b f are the corresponding weight coefficient matrix and bias term respectively, and σ is the sigmoid activation function. At the same time, the neural network model in this embodiment obtains the input gate vector i through equations (2) and (3): t And the unit status update value The input gate vector i t Each element is in the range [0,1] and is used to control the value of the control state update. In updating this subunit status C t The retention level at time i t 0 means "no update at all", i t 1 represents “complete update”. i and b i are the corresponding weight coefficient matrix and bias term, and σ is the sigmoid activation function. In formula (3), WC and b C are the corresponding weight coefficient matrix and bias term respectively.

[0076]

[0077] Input data passes through the forget gate vector f t With the input gate vector i t After that, valuable historical information will be retained and worthless information will be removed. The state of this subunit is combined with the historical unit state and the updated state through formula (4) to complete the unit state C t Updates.

[0078]

[0079] To obtain the predicted value And complete the hidden node state h t Update, and obtain the output gate vector o through equations (5) and (6) t And the hidden node state h of this subunit t The output gate vector o t Each element is in the range [0,1] and is used to control the state h of the unit hidden node. t Update degree, o t 0 means "no update at all", o t 1 represents “complete update”. In formula (5), W o and b o are the corresponding weight coefficient matrix and bias term respectively, and σ is the sigmoid activation function.

[0080]

[0081] h t =o t *tanh(C t )(6)

[0082] Through the above LSTM structure and data flow, the modal parameter and inclination angle prediction of the LSTM neural network prediction model is realized.

[0083] In a preferred embodiment of the present invention, step five is implemented. When modal parameter prediction is performed, a modal parameter sequence for prediction is determined, the above data sequence is input as input data into a trained LSTM neural network model, and the modal parameter prediction value of the monitoring target building or structure is output. When inclination angle prediction is performed, a corrected inclination angle sequence for prediction is determined, the above data sequence is input as input data into a trained LSTM neural network model, and the inclination angle prediction value of the monitoring target building or structure is output.

[0084] After obtaining the predicted values ​​of the modal parameters and the predicted values ​​of the inclination angle, further, step six is ​​implemented in the embodiment to perform trend judgment and generate early warning information. Specifically, the modal parameter and inclination angle early warning processing process is assembled in the cloud or local server where the data analysis is located. Based on the predicted modal parameters and inclination angles, the slope is obtained by linear fitting, and the slope is used as the inclination angle trend. When the predicted values ​​of the target building or structure obtained by the above-mentioned prediction methods exceed the settlement threshold set by the system, or the slope obtained by the linear fitting of the modal parameters and the inclination angle increases significantly, it is considered that the monitored settlement has deteriorated, and early warning information is sent to relevant operation and maintenance management personnel through text messages, emails, etc. The above-mentioned early warning information can be hierarchically managed and warned for different thresholds, thereby realizing fully automatic monitoring and early warning of settlement. For example, an alarm is triggered when the inclination angle trend or at least one modal parameter exceeds the threshold, or an alarm is triggered when the inclination angle trend and at least one modal parameter exceed the threshold at the same time, so as to realize data fusion analysis and reduce the probability of false alarms.

[0085] The above-mentioned embodiments of the present invention comprehensively consider the different characteristics of the target buildings or structures and the actual environmental conditions, and can accurately predict the structural health and safety index data under different site conditions. The interactive information provided can provide relevant operation and maintenance personnel with a low-cost, high-efficiency, and high-precision real-time structural safety management solution.

[0086] Based on the same inventive concept, another embodiment of the present invention further provides a multivariate data fusion analysis system for structural health monitoring, comprising:

[0087] A data acquisition module, which uses an integrated multi-sensor to acquire monitoring data of a target building or structure, wherein the monitoring data includes acceleration monitoring data, inclination monitoring data, and temperature monitoring data;

[0088] A data transmission module, for transmitting the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data;

[0089] A data processing module performs structural modal parameter identification on the acceleration monitoring data, obtains a first modal parameter identification result of the target building or structure, and obtains modal parameter historical monitoring data; processes the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data;

[0090] A prediction model training module is used to establish an LSTM neural network modal parameter prediction model, use the modal parameter historical monitoring data as training samples, and train the LSTM neural network modal parameter prediction model; establish an LSTM neural network inclination prediction model, use the corrected inclination historical monitoring data as training samples, and train the LSTM neural network inclination prediction model;

[0091] The settlement prediction module determines a modal parameter sequence for prediction, inputs the modal parameter sequence into a trained LSTM neural network modal parameter prediction model, and outputs a modal parameter prediction sequence of a target building or structure; determines a corrected inclination angle sequence for prediction, inputs the corrected inclination angle sequence into a trained LSTM neural network inclination angle prediction model, and outputs an inclination angle prediction sequence of a target building or structure;

[0092] The early warning module performs trend determination and early warning on the predicted modal parameter prediction sequence and inclination angle prediction sequence.

[0093] The modules / units in the above examples of the present invention may specifically refer to the implementation techniques of the corresponding steps of the multivariate data collection and fusion analysis method for structural health monitoring in the above embodiments, which will not be described in detail here.

[0094] Based on the same inventive concept, in other embodiments of the present invention, a terminal is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can be used to execute the above-mentioned method or to run the above-mentioned system when executing the program.

[0095] Optionally, the memory is used to store programs; the memory may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM), such as static random-access memory (English: static random-access memory, abbreviated: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviated: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory is used to store computer programs (such as applications, functional modules, etc. that implement the above method), computer instructions, etc., and the above-mentioned computer programs, computer instructions, etc. can be partitioned and stored in one or more memories.

[0096] The processor is used to execute the computer program stored in the memory to implement the various steps in the method involved in the above embodiment. For details, please refer to the relevant description in the above method embodiment.

[0097] The processor and the memory may be independent structures or integrated structures. When the processor and the memory are independent structures, the memory and the processor may be coupled and connected via a bus.

[0098] Based on the same inventive concept, in other embodiments of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, it can be used to execute the above method, or run the above system.

[0099] Among them, computer-readable media include computer storage media and communication media, wherein the communication media include any media that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general or special-purpose computer. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in a user device. Of course, the processor and the storage medium can also be present in a communication device as discrete components.

[0100] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0101] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0104] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A multivariate data collection and fusion analysis method for structural health monitoring, characterized in that: include: Using integrated multi-sensors to obtain monitoring data of target buildings or structures, the monitoring data including acceleration monitoring data, inclination monitoring data, and temperature monitoring data; Transmitting the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data; Performing structural modal parameter identification on the acceleration monitoring data to obtain modal parameter identification results, and obtaining modal parameter historical monitoring data; processing the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data; Establishing an LSTM neural network modal parameter prediction model, and using the modal parameter historical monitoring data as training samples for training; establishing an LSTM neural network corrected inclination prediction model, and using the corrected inclination historical monitoring data as training samples for training; Obtain a modal parameter sequence to be predicted, input the modal parameter sequence into a trained LSTM neural network modal parameter prediction model, and output a modal parameter prediction sequence of a target building or structure; obtain a corrected inclination angle sequence to be predicted, input the corrected inclination angle sequence into a trained LSTM neural network inclination angle prediction model, and output an inclination angle prediction sequence of a target building or structure; Trend determination and early warning are performed based on the predicted modal parameter prediction sequence and the inclination angle prediction sequence.

2. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 1 is characterized in that: The use of integrated multi-sensors to obtain monitoring data of target buildings or structures is to use sensors for regular collection; When the sensor performs a single acquisition, the collected data includes acceleration time history, single or multiple inclination data, and single or multiple temperature data collected at a fixed sampling frequency within a fixed time period.

3. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 1 is characterized in that: The SSI method is used to identify the modal parameters of the acceleration monitoring data of the target building or structure, and the modal parameters are identified, including: modal frequency, modal vibration shape and modal damping; A stability diagram is also generated during the identification process, wherein the stability diagram includes a stability diagram whose abscissa is the modal frequency and whose ordinate is the modal order; Based on the stability diagram, unstable frequencies in the identified modal parameters are filtered to obtain frequency-stable modal frequencies, modal vibration shapes and modal damping.

4. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 1 is characterized in that: The processing of the inclination monitoring data and the temperature monitoring data to obtain the corrected inclination historical monitoring data and the temperature historical monitoring data includes: Normalizing the inclination data and the temperature data and constructing a database to obtain inclination historical monitoring data and temperature historical monitoring data; The XGBoost method is used to perform segmented component decomposition on the historical monitoring data of the inclination angle, and linear fitting is used to obtain the trend item in the component decomposition; The temperature historical monitoring data is regarded as an exogenous variable, and the least squares fitting is adopted to obtain the exogenous term caused by the exogenous variable; The trend term and the exogenous term are removed from the inclination historical monitoring data to obtain the corrected inclination historical monitoring data.

5. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 4 is characterized in that: The normalization process is specifically as follows: E[x] is the mean of the data sample, and Var[x] is the variance of the data sample.

6. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 1 is characterized in that: The training process of the LSTM neural network modal parameter prediction model is the same as that of the LSTM neural network inclination angle prediction model, and both include the following process: Determine the neural network hyperparameters based on the number of input variables; Determine the output variable of the LSTM neural network modal parameter prediction model as a modal parameter prediction sequence; determine the output variable of the LSTM neural network inclination angle prediction model as an inclination angle prediction sequence; Training is performed by minimizing the difference between the modal parameter prediction value / minimizing the inclination angle prediction value and the measured value to obtain the parameters of each neuron node of the LSTM neural network modal parameter prediction model / LSTM neural network inclination angle prediction model.

7. The multivariate data collection and fusion analysis method for structural health monitoring according to claim 1 is characterized in that: The performing trend determination and early warning based on the modal parameter prediction sequence and the inclination angle prediction sequence obtained by prediction includes: Obtaining an inclination trend based on the modal parameters and the inclination prediction value; According to the set upper and lower limit thresholds of modal parameters and inclination trend, judge whether the modal parameters and inclination trend exceed the limit: An early warning is issued when the inclination angle trend or at least one modal parameter exceeds the limit; or, an early warning is issued when both the inclination angle trend and at least one modal parameter exceed the limit.

8. A multi-data collection and fusion analysis system for structural health monitoring, characterized in that: include: A data acquisition module, which uses an integrated multi-sensor to acquire monitoring data of a target building or structure, wherein the monitoring data includes acceleration monitoring data, inclination monitoring data, and temperature monitoring data; A data transmission module, for transmitting the acceleration monitoring data, the inclination monitoring data, and the temperature monitoring data; A data processing module performs structural modal parameter identification on the acceleration monitoring data to obtain modal parameter identification results, and obtains modal parameter historical monitoring data; processes the inclination monitoring data and the temperature monitoring data to obtain corrected inclination historical monitoring data and temperature historical monitoring data; A prediction model training module is used to establish an LSTM neural network modal parameter prediction model, and the modal parameter historical monitoring data is used as a training sample for training; an LSTM neural network corrected inclination prediction model is established, and the corrected inclination historical monitoring data is used as a training sample for training; The settlement prediction module obtains a modal parameter sequence to be predicted, inputs the modal parameter sequence into a trained LSTM neural network modal parameter prediction model, and outputs a modal parameter prediction sequence of a target building or structure; obtains a corrected inclination angle sequence to be predicted, inputs the corrected inclination angle sequence into a trained LSTM neural network inclination angle prediction model, and outputs an inclination angle prediction sequence of a target building or structure; The early warning module performs trend determination and early warning based on the predicted modal parameter prediction sequence and the inclination angle prediction sequence.

9. A terminal 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, it can be used to execute the method described in any one of claims 1 to 7, or to run the system described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it can be used to execute the method described in any one of claims 1 to 7, or to run the system described in claim 8.

Images