Underground structure full life cycle management method and system based on fusion sensing data

Abstract

The invention relates to the technical field of underground structure management, in particular to an underground structure full life cycle management method and system based on fusion sensing data. According to the technical scheme, the system comprises a sensing layer, a data transmission layer, a data processing and analysis layer and a decision support and application layer. Through cooperation of the distributed optical fiber sensor and the multiple sensors, all-around and multi-parameter monitoring of the underground structure is achieved, the monitoring accuracy and reliability are improved, historical data and operation rules of the structure are automatically learned through a data driving prediction model, different geological environments and load conditions are adapted, a scientific basis is provided for maintenance decision making, and meanwhile, the system has good application prospects. The self-adjusting capability is realized; monitoring system application and data accumulation and utilization in each stage are considered, seamless connection and cooperative work are achieved, full-life-cycle integrated management is achieved, the overall performance of the structure is improved, the service life of the structure is prolonged, the maintenance cost is reduced, and sustainable development is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of underground structure management, and in particular to a method and system for managing the entire life cycle of underground structures based on fused sensor data. Background Art

[0002] With the acceleration of urbanization and the increasing scarcity of land resources, the development and utilization of underground space has received increasing attention. Underground structures such as subways, underground shopping malls, and integrated pipeline corridors have emerged in large numbers, making the management and maintenance of underground structures crucial. Underground structures are often located in complex geological environments and are affected by various factors such as soil pressure and groundwater. Furthermore, once safety issues arise in underground structures, they often cause serious casualties and property losses, thus placing higher demands on their safety management throughout their life cycle. Existing long-term health monitoring of underground structures still relies on periodic testing with discretely distributed sensors, making it difficult to achieve efficient, continuous, and real-time detection. Early identification of structural damage is particularly difficult in complex geodetic environments. Furthermore, existing monitoring systems lack intelligent feedback and adaptive control mechanisms, making it difficult to achieve an integrated monitoring-assessment-optimization application. Therefore, we propose a method and system for the full life cycle management of underground structures based on fused sensor data. Summary of the Invention

[0003] The purpose of the present invention is to address the problems existing in the background technology and propose a method and system for underground structure full life cycle management based on fused sensor data.

[0004] A first aspect: A full lifecycle management system for underground structures based on fused sensor data, comprising: a perception layer, the perception layer comprising a distributed optical fiber sensor network and multiple types of sensor modules;

[0005] A data transmission layer, comprising a wired and wireless hybrid transmission network and a data encryption and transmission module;

[0006] The data processing and analysis layer includes a multi-source data fusion algorithm for fusing the data collected by the perception layer, and a prediction model based on a deep learning algorithm;

[0007] The decision support and application layer includes an adaptive control module established based on the analysis results of the data processing and analysis layer.

[0008] Optionally, the distributed optical fiber sensor network includes several groups of distributed optical fiber sensors laid along key parts of the underground structure, using wavelength division multiplexing and time division multiplexing technology, wherein the wavelength division multiplexing and time division multiplexing technology are used to realize multiplexing of multiple sensors on one optical fiber;

[0009] The wavelength division multiplexing includes using a wavelength separator at the receiving end to separate optical signals of different wavelengths, and setting n sensors to be modulated at wavelengths λ1, λ2, ..., λ n The optical signal is represented as:

[0010]

[0011] Among them, A i is the optical signal amplitude, ω i =2πc / λ i , c is the speed of light, is the initial phase, and the optical filter is used to separate the sensor signals at the receiving end. Bragg The Bragg grating, whose reflected light satisfies λ Bragg =2nΛ, where n is the fiber refractive index and Λ is the grating period. The center wavelength changes with the fiber refractive index or grating period. The corresponding sensor information is obtained by detecting the change in the reflected light wavelength.

[0012] The time division multiplexing technology involves dividing the time axis into multiple time slots. Different sensors send optical signals in different time slots. The transmitter sends optical pulse signals from each sensor in sequence. The receiver receives the signals from the corresponding sensors in the corresponding time slots through time synchronization. Assuming that the length of each time slot is Δt, the optical pulse signal sent by the jth sensor in the mth time slot is expressed as:

[0013]

[0014] Among them, rect(x) is the matrix impulse function, τ j is the delay time of the jth sensor signal, ω c is the carrier angular frequency. In a time-division multiplexing system containing N sensors, the total time period T = NΔt.

[0015] Optionally, the multi-type sensor module includes an acceleration sensor, a pressure sensor, and a displacement sensor. An intelligent sensor node is used to perform local preprocessing and preliminary fusion of the sensor data. The intelligent sensor node is equipped with a microprocessor and a local storage unit. During the local preprocessing, a mean filtering algorithm is used to remove noise from the raw data. The formula is as follows:

[0016]

[0017] in, is the filtered data, x i is the original data sequence, N is the filter window size;

[0018] In the initial fusion, the weighted average fusion algorithm is used to calculate the measurement values y1, y2, ..., y n , the fusion result Y is:

[0019]

[0020] Among them, w i is the weight of the corresponding sensor data, and

[0021] Optionally, in the hybrid wired and wireless transmission network, the data collected by the distributed optical fiber sensor is transmitted using an optical fiber network, and the data collected by the multi-sensor module is transmitted using wireless transmission and wired transmission. For sensors close to the data aggregation node, wireless communication technology is used for transmission, and for sensors far from the data aggregation node, wired Ethernet is used for transmission.

[0022] The data encryption and transmission module includes encrypting data using an advanced encryption algorithm, which uses a block cipher system to group plaintext into blocks of fixed length and encrypt each block. Let the plaintext block be P and the key be K. The encrypted ciphertext block C is obtained by round transformation. The round transformation includes byte substitution, row shift, column confusion and round key addition operations. The byte substitution performs nonlinear substitution on each byte in the plaintext through an S-box. Let the plaintext byte be x and the byte after the S-box substitution be y, expressed as y=S(x);

[0023] The transmitted data is checked by a cyclic redundancy check method, which generates a check code by performing a polynomial division operation on the data. Suppose the data polynomial is D(x), the generating polynomial is G(x), and the remainder R(x) of D(x) divided by G(x) is calculated as the check code. The check code is recalculated at the receiving end and compared with the received check code. If they are consistent, the data transmission is correct.

[0024] Optionally, the multi-source data fusion algorithm extracts time series features through a bidirectional long short-term memory network and introduces an adaptive weight allocation module to optimize the contribution weights of multimodal data in real time;

[0025] The core calculation of the long short-term memory unit in the bidirectional long short-term memory network includes the update of the input gate, the forget gate, the output gate and the memory unit. The bidirectional long short-term memory network processes the input sequence simultaneously through the forward and backward long short-term memory networks, and takes the weighted sum of the hidden states in the two directions as the final output;

[0026] The adaptive weight allocation module includes a hybrid decision-making mechanism of fuzzy logic and evidence theory. The fuzzy logic is used to convert the monitoring value of multimodal data into fuzzy linguistic variables. First, the fuzzy membership function of the input data is determined, and fuzzy reasoning is performed according to the pre-established fuzzy rule base to obtain the output fuzzy set. The evidence theory is used to process uncertainty information and assign trust to each data modality through the basic probability distribution function. Let m i is the basic probability distribution function of the i-th data mode, and for all possible hypothesis sets A, it satisfies Where Θ is the recognition framework, and the basic probability distribution functions of different data modalities are then fused through the Dempster combination rule to obtain a comprehensive trust distribution, and then weights are assigned to each data modality in real time based on the trust distribution;

[0027] Let w i is the weight of the i-th data mode. After being processed by the hybrid decision-making mechanism of fuzzy logic and evidence theory, the weight update formula is:

[0028]

[0029] Among them, m i is the trustworthiness of the i-th data modality obtained by evidence theory, is the membership degree of the fuzzy set corresponding to the i-th data mode obtained by fuzzy logic, and n is the total number of data modes.

[0030] Optionally, in the prediction model based on the deep learning algorithm, the structural mechanics constitutive equation is embedded in the deep learning model as a regularization term. The deep learning model includes an input layer, multiple hidden layers and an output layer. Let the input data be X, and the weight matrix of the lth layer be W l , the bias vector is b l , the activation function is σ, then the output H of the lth layer l Expressed as:

[0031] H l =σ(W l H l-1 +b l )

[0032] Among them, H 0 =X, the output of the final output layer is the prediction result of the prediction model

[0033] The structural mechanics constitutive equation is added as a regularization term to the loss function L of the deep learning model, and the loss based on data fitting is set to L date , the regularization term corresponding to the structural mechanics constitutive equation is L physics , then the total loss function is: L = Ldate +λL physics , where λ is a hyperparameter that balances the weights of the two;

[0034] The strain results predicted by the deep learning model are substituted into the constitutive equation and the deviation from the theoretical value is calculated to construct L physics , assuming the strain predicted by the model is The theoretical stress calculated according to the constitutive equation is: The stress predicted by the model is Then L physics It can be expressed as:

[0035]

[0036] During the model training process, the weight W is adjusted through the back propagation algorithm l and bias b l , so that the total loss function L is minimized.

[0037] Optionally, the adaptive control module uses a model predictive control algorithm and a rule-based decision-making mechanism, wherein the rules are pre-established, and when the structural state is monitored to trigger specific rule conditions, corresponding control measures are executed;

[0038] The model predictive control algorithm solves a finite time domain optimization problem in each control cycle through rolling optimization to determine the current optimal control input. In the underground structure scenario, the key parameters of structural deformation and stress are used as state variables. The state space model of the underground structure is assumed to be:

[0039] x k+1 =Ax k +Bu k +w k

[0040] y k =Cx k +v k

[0041] Among them, x k is the state vector at time k, u k is the control input at time k, y k is the output vector at time k, A, B, and C are the state transfer matrix, input matrix, and output matrix respectively, w k and v k are process noise and measurement noise respectively. The prediction model is based on the current state x k And the control input sequence for the next N steps, predict the state and output for the next N steps, the prediction formula is:

[0042] x i+1|k =Axi|k +Bu i|k

[0043] y i|k =Cx i|k

[0044] Where i = k, k + 1, ..., k + N - 1, and the optimization goal is to minimize the deviation between the predicted output and the expected output and the change in the control input. The objective function J is constructed as follows:

[0045]

[0046] Among them, y ref,i is the expected output at time i, Δu i|k =u i|k -u i-1|k To control the input variation, λ1 and λ2 are weight coefficients, u nominal is the nominal control input. In each control cycle k, the optimal control input sequence is obtained by solving the above optimization problem. And the first element As the control input actually applied at the current moment.

[0047] In the second aspect, this application proposes a method for the full life cycle management of underground structures based on fused sensor data, which includes the following steps: construction of a geological environment and structural model, acquisition of geological information through geological exploration technology, construction of a three-dimensional geological model in combination with geographic information system technology, and formulation of a planning scheme for a monitoring system;

[0048] Install sensors and perform system debugging according to the planning scheme, use the monitoring system to monitor the construction process of the terrain structure in real time, and adjust the construction process and construction parameters based on the monitoring results, and establish a construction process monitoring feedback mechanism;

[0049] The monitoring system collects monitoring data of underground structures in real time and performs data analysis to evaluate the health status of the structures;

[0050] Make maintenance decisions based on the structural health assessment results and implement maintenance work based on the maintenance decisions

[0051] Optionally, in the sensor installation step, for distributed fiber optic sensors, during tunnel lining construction, the optical fiber is pre-buried in the concrete pouring layer and fixed by a clamp. During the installation process, the optical fiber is connected by a fiber optic fusion splicer, and an optical time domain reflectometer is used to perform real-time detection of the optical fiber to monitor the loss and breakpoints of the optical fiber. After the sensor is installed, the monitoring system is fully debugged, and the data acquisition and processing software is functionally tested, and the sensor is calibrated and calibrated. At the same time, a sensor calibration file is established to record the calibration time, calibration results and next calibration time information.

[0052] Optionally, the maintenance decision includes the selection of maintenance plan, determination of maintenance time, and allocation of maintenance resources. According to the maintenance decision, the maintenance work is organized and implemented. During the maintenance process, the monitoring system is used to monitor and evaluate the maintenance effect in real time, and a maintenance work file is established to record the maintenance time, maintenance content, materials used and maintenance personnel information.

[0053] In summary, this application includes at least one of the following beneficial technical effects:

[0054] The present invention uses distributed optical fiber sensors in conjunction with multiple sensors, wavelength division multiplexing, time division multiplexing technology, and intelligent sensor nodes to achieve all-round, multi-parameter monitoring of underground structures, tap the potential value of different sensor data, and improve monitoring accuracy and reliability.

[0055] The data-driven prediction model introduced in this invention can automatically learn the historical data and operating rules of the structure, adapt to different geological environments and load conditions, analyze the structural health status and predict future performance, and provide a scientific basis for maintenance decisions. At the same time, the adaptive control mechanism constructed by this invention based on monitoring data and analysis results realizes that when the structural state is abnormal, the system automatically initiates corresponding control measures, improving the structure's ability to self-regulate and respond to emergencies.

[0056] From planning and design, construction to operation and maintenance, the present invention fully considers the application of the monitoring system and the accumulation and utilization of data in each stage to achieve seamless connection and collaborative work. Through the refined construction of geological environment and structural models, monitoring system planning, construction process monitoring and quality control, and reasonable handling of operation and maintenance decision-making execution, it realizes integrated management of the entire life cycle, improves the overall performance and service life of the structure, reduces maintenance costs, and promotes sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The present invention provides a structural block diagram of an underground structure full life cycle management system based on fusion sensor data;

[0058] Figure 2A flowchart of a method for full life cycle management of underground structures based on fused sensor data is given in the present invention.

[0059] Figure 3 A schematic diagram of the execution flow of a method for full life cycle management of underground structures based on fused sensor data is provided in the present invention;

[0060] Figure 4 A creep deformation-time relationship diagram based on long-term monitoring in the present invention is provided;

[0061] Figure 5 A flow chart of the method for dynamic adjustment of underground structure parameters based on particle swarm optimization in the present invention is given;

[0062] Figure 6 A comparative verification diagram of stress-strain model prediction and constitutive relationship in a full life cycle management method of underground structures based on fused sensor data is given in the present invention. DETAILED DESCRIPTION

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0064] Example

[0065] On the one hand, the present invention proposes a full life cycle management system for underground structures based on fusion sensor data, such as Figure 1 As shown, it includes perception layer, data transmission layer, data processing and analysis layer, and decision support and application layer.

[0066] Among them, the perception layer includes a distributed fiber optic sensor network and multi-type sensor modules. The distributed fiber optic sensor network includes several groups of distributed fiber optic sensors laid along key parts of the underground structure, including tunnel linings and pipe gallery walls. A distributed optical fiber is arranged every 10 meters, and wavelength division multiplexing and time division multiplexing technologies are used. Wavelength division multiplexing and time division multiplexing technologies are used to realize the multiplexing of multiple sensors on one optical fiber.

[0067] Wavelength division multiplexing involves using a wavelength separator at the receiving end to separate optical signals of different wavelengths. Suppose n sensors are modulated at wavelengths λ1, λ2, ..., λ n The optical signal is represented as Among them, A i is the optical signal amplitude, ω i =2πc / λ i , c is the speed of light, is the initial phase, and the optical filter is used to separate the sensor signals at the receiving end. Bragg The Bragg grating, whose reflected light satisfies λ Bragg =2nΛ, where n is the fiber refractive index and Λ is the grating period. The center wavelength changes with the fiber refractive index or grating period. The corresponding sensor information is obtained by detecting the change in the reflected light wavelength. Wavelength division multiplexing technology modulates the optical signals of different sensors at different wavelengths and transmits them in the same optical fiber. At the receiving end, a wavelength separator is used to separate the optical signals of different wavelengths, thereby realizing demultiplexing of multiple sensor signals.

[0068] Time division multiplexing technology involves dividing the time axis into multiple time slots. Different sensors send optical signals in different time slots. The transmitter sends the optical pulse signal of each sensor in turn. The receiver receives the signal of the corresponding sensor in the corresponding time slot through time synchronization. Assuming that the length of each time slot is Δt, the optical pulse signal sent by the jth sensor in the mth time slot is expressed as:

[0069]

[0070] Among them, rect(x) is the matrix impulse function, τ j is the delay time of the jth sensor signal, ω c is the carrier angular frequency. In a time-division multiplexing system containing N sensors, the total time period T = NΔt. The receiving end collects signals in the corresponding time slot through the time synchronization circuit to realize the multiplexing and demultiplexing of multiple sensor signals.

[0071] In addition, the multi-type sensor modules include acceleration sensors, pressure sensors, and displacement sensors. Acceleration sensors are used to monitor the response of underground structures under dynamic loads, pressure sensors are used to measure environmental loads, and displacement sensors are used to measure displacement changes of structures. Intelligent sensor nodes are used to perform local preprocessing and preliminary fusion of sensor data to reduce data transmission volume and processing burden. Intelligent sensor nodes are equipped with microprocessors and local storage units. In local preprocessing, the mean filtering algorithm is used to remove noise from the raw data. The formula is: in, is the filtered data, x i is the original data sequence, N is the filter window size;

[0072] In the initial fusion, the weighted average fusion algorithm is used to calculate the measurement values y1, y2, ..., y n , the fusion result Y is:

[0073]

[0074] Among them, w i is the weight of the corresponding sensor data, and

[0075] The present invention realizes all-round, multi-parameter monitoring of underground structures through the collaboration of distributed optical fiber sensors and multiple sensors, and applies wavelength division multiplexing, time division multiplexing technology and intelligent sensor nodes, thereby exploring the potential value of different sensor data and improving monitoring accuracy and reliability.

[0076] The data transmission layer includes a wired and wireless hybrid transmission network and a data encryption and transmission module. In the wired and wireless hybrid transmission network, the data collected by distributed fiber optic sensors are transmitted using a fiber optic network. The fiber optic network has the advantages of large bandwidth and strong anti-interference ability, which can ensure real-time and stable transmission of data. For the data collected by the multi-sensor module, wireless transmission and wired transmission are used for transmission. Among them, for sensors close to the data aggregation node, wireless communication technology is used for transmission, and for sensors far away from the data aggregation node, wired Ethernet is used for transmission.

[0077] Among them, the data encryption and transmission module includes using an advanced encryption algorithm to encrypt data to ensure the security and integrity of the data. The advanced encryption algorithm adopts a block cipher system to group the plaintext into blocks of fixed length and encrypt each block. Let the plaintext block be P and the key be K. The encrypted ciphertext block C is obtained through round transformation. The round transformation includes byte substitution, row shift, column confusion and round key addition operation. Among them, byte substitution performs nonlinear replacement of each byte in the plaintext through an S box. Let the plaintext byte be x and the byte replaced by the S box be y, expressed as y=S(x). The transmitted data is verified by the cyclic redundancy check method. The cyclic redundancy check method generates a check code by performing a polynomial division operation on the data. Let the data polynomial be D(x) and the generating polynomial be G(x). The remainder R(x) of D(x) divided by G(x) is calculated as the check code. The check code is recalculated at the receiving end and compared with the received check code to determine whether the data transmission is correct. If they are consistent, the data transmission is correct.

[0078] The data processing and analysis layer includes a multi-source data fusion algorithm for fusing the data collected by the perception layer, and also includes a prediction model based on a deep learning algorithm.

[0079] Among them, the multi-source data fusion algorithm extracts time series features through a bidirectional long short-term memory network, and introduces an adaptive weight distribution module to optimize the contribution weights of multimodal data in real time. The core calculations of the long short-term memory unit in the bidirectional long short-term memory network include the input gate, forget gate, output gate and memory unit update. The bidirectional long short-term memory network processes the input sequence simultaneously through the forward and backward long short-term memory networks, and takes the weighted sum of the hidden states in the two directions as the final output, thereby achieving comprehensive acquisition of time series features.

[0080] The adaptive weight allocation module includes a hybrid decision-making mechanism of fuzzy logic and evidence theory. Fuzzy logic is used to convert the monitoring value of multimodal data into fuzzy linguistic variables. First, the fuzzy membership function of the input data is determined, and fuzzy reasoning is performed according to the pre-established fuzzy rule base to obtain the output fuzzy set. Evidence theory is used to process uncertainty information and assign trust to each data modality through the basic probability distribution function. Let m i is the basic probability distribution function of the i-th data mode, and for all possible hypothesis sets A, it satisfies Here, Θ is the recognition framework. The basic probability distribution functions of different data modalities are then fused through the Dempster combination rule to obtain a comprehensive trust distribution. Then, based on the trust distribution, weights are assigned to each data modality in real time to optimize the contribution of multimodal data in the fusion process.

[0081] Let w i is the weight of the i-th data mode. After being processed by the hybrid decision-making mechanism of fuzzy logic and evidence theory, the weight update formula is:

[0082]

[0083] Among them, m i is the trustworthiness of the i-th data modality obtained by evidence theory, is the membership degree of the fuzzy set corresponding to the i-th data mode obtained by fuzzy logic, and n is the total number of data modes. By continuously updating the weights in real time, the fusion algorithm can dynamically adjust the importance of each modal data in the fusion according to the actual operating status of the underground structure, thereby improving the accuracy of the fusion result in reflecting the structural status.

[0084] In the prediction model based on deep learning algorithm, the structural mechanics constitutive equation is embedded in the deep learning model as a regularization term. The deep learning model includes an input layer, multiple hidden layers and an output layer. Let the input data be X and the weight matrix of the lth layer be W. l , the bias vector is b l , the activation function is σ, then the output H of the lth layer l Expressed as: H l =σ(Wl H l-1 +b l ), where H 0 =X, the output of the final output layer is the prediction result of the prediction model

[0085] The structural mechanics constitutive equation is added as a regularization term to the loss function L of the deep learning model. The purpose is to introduce physical constraints in the model training process so that the model can not only fit the data but also follow the basic laws of structural mechanics, thereby improving the generalization ability and prediction accuracy of the model. Especially under limited data or complex working conditions, the loss based on data fitting is set to L date , the regularization term corresponding to the structural mechanics constitutive equation is L physics , then the total loss function is: L = L date +λL physics , where λ is a hyperparameter that balances the weights of the two;

[0086] The strain results predicted by the deep learning model are substituted into the constitutive equation and the deviation from the theoretical value is calculated to construct L physics , assuming the strain predicted by the model is The theoretical stress calculated according to the constitutive equation is: The stress predicted by the model is Then L physics It can be expressed as:

[0087]

[0088] During the model training process, the weight W is adjusted through the back propagation algorithm l and bias b l , so that the total loss function L is minimized, thereby integrating the structural mechanics constitutive equation into the training of the deep learning model.

[0089] The present invention introduces a data-driven prediction model that can automatically learn the historical data and operating laws of the structure, adapt to different geological environments and load conditions, analyze the structural health status and predict future performance, and provide a scientific basis for maintenance decisions.

[0090] The decision support and application layer includes an adaptive control module based on the analysis results of the data processing and analysis layer. The adaptive control module uses a model predictive control algorithm and a rule-based decision-making mechanism. The rules are pre-established. When the monitored structural state triggers specific rule conditions, the corresponding control measures are executed.

[0091] The model predictive control algorithm solves a finite-time optimization problem in each control cycle through rolling optimization to determine the current optimal control input. In the underground structure scenario, the key parameters of structural deformation and stress are used as state variables, and a state space model is established to describe the dynamic characteristics of the structure. The state space model of the underground structure is assumed to be:

[0092] x k+1 =Ax k +Bu k +w k

[0093] y k =Cx k +v k

[0094] Among them, x k is the state vector at time k, u k is the control input at time k, y k is the output vector at time k, A, B, and C are the state transfer matrix, input matrix, and output matrix respectively, w k and v k They are process noise and measurement noise respectively. The prediction model is based on the current state x k And the control input sequence for the next N steps, predict the state and output for the next N steps, the prediction formula is:

[0095] x i+1|k =Ax i|k +Bu i|k

[0096] y i|k =Cx i|k

[0097] Where i = k, k + 1, ..., k + N - 1, and the optimization goal is to minimize the deviation between the predicted output and the expected output and the change in the control input. The objective function J is constructed as follows:

[0098]

[0099] Among them, y ref,i is the expected output at time i, Δu i|k =u i|k -u i-1|k To control the input variation, λ1 and λ2 are weight coefficients, u nominal is the nominal control input. In each control cycle k, the optimal control input sequence is obtained by solving the above optimization problem. And the first element As the control input actually applied at the current moment.

[0100] In addition to model predictive control, a rule-based decision-making mechanism is introduced as a supplement. Based on common abnormal conditions of underground structures and corresponding treatment strategies, some rules are pre-established. When the monitored structural status triggers specific rule conditions, corresponding control measures are executed to improve response speed and decision reliability.

[0101] In this embodiment, rule 1 is proposed: abnormal rise in groundwater level;

[0102] Condition: When the monitored groundwater level exceeds the warning water level h for n1 consecutive cycles alert , and the rate of increase is greater than v rise ;

[0103] Control measures: Start the drainage equipment, and the drainage power P drain Determined according to the following formula:

[0104] Pd rein =α·(hh elart )·v rise

[0105] Where h is the current groundwater level and α is a coefficient related to the performance of the drainage equipment.

[0106] The present invention constructs an adaptive control mechanism based on monitoring data and analysis results. When the structural state is abnormal, the system automatically initiates corresponding control measures to improve the structure's ability to self-regulate and respond to emergencies.

[0107] On the other hand, the present invention provides a method for managing the entire life cycle of underground structures based on fused sensor data, such as Figure 2 As shown, the following steps are included:

[0108] Construction of geological environment and structural models, obtaining geological information through geological exploration technology, combining with geographic information system technology to construct a three-dimensional geological model, and formulate a planning scheme for the monitoring system;

[0109] Install sensors and conduct system debugging according to the planning scheme, use the monitoring system to monitor the construction process of the terrain structure in real time, and adjust the construction process and construction parameters based on the monitoring results, and establish a monitoring feedback mechanism for the construction process;

[0110] The monitoring system collects monitoring data of underground structures in real time and performs data analysis to evaluate the health status of the structures;

[0111] Based on the structural health status assessment results, maintenance decisions are made and maintenance work is carried out according to the maintenance decisions.

[0112] Among them, in the sensor installation step, for distributed fiber optic sensors, during the tunnel lining construction, the optical fiber is pre-buried in the concrete pouring layer and fixed by a clamp. During the installation process, the optical fiber is connected by a fiber optic fusion splicer, and the optical time domain reflectometer is used to perform real-time detection of the optical fiber to monitor the loss and breakpoints of the optical fiber. After the sensor is installed, the monitoring system is fully debugged to check whether the data transmission network is unobstructed. The transmission rate and packet loss rate indicators of the wired network are tested by a network tester. For wireless networks, the signal strength and stability are tested, and the data acquisition and processing software is functionally tested to check the accuracy, real-time performance of data acquisition, and whether the data storage and analysis functions are normal. The sensor is calibrated and calibrated. The pressure sensor is calibrated with a standard weight, and the displacement sensor is calibrated with a standard displacement stage to ensure the accuracy of the monitoring data. At the same time, a sensor calibration file is established to record the calibration time, calibration results and the next calibration time information.

[0113] Maintenance decisions include the selection of maintenance plans, determination of maintenance time, and allocation of maintenance resources. Maintenance work is organized and implemented according to maintenance decisions. During the maintenance process, the monitoring system is used to monitor and evaluate the maintenance effect in real time, and a maintenance work file is established to record maintenance time, maintenance content, materials used, and maintenance personnel information.

[0114] From planning and design, construction to operation and maintenance, the present invention fully considers the application of the monitoring system and the accumulation and utilization of data in each stage to achieve seamless connection and collaborative work. Through the refined construction of geological environment and structural models, monitoring system planning, construction process monitoring and quality control, and reasonable handling of operation and maintenance decision-making execution, it realizes integrated management of the entire life cycle, improves the overall performance and service life of the structure, reduces maintenance costs, and promotes sustainable development.

[0115] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above embodiments, those skilled in the art may make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A full life cycle management system for underground structures based on fused sensor data, characterized in that: include: A perception layer comprising a distributed optical fiber sensor network and multiple sensor modules; A data transmission layer, comprising a wired and wireless hybrid transmission network and a data encryption and transmission module; The data processing and analysis layer includes a multi-source data fusion algorithm for fusing the data collected by the perception layer, and a prediction model based on a deep learning algorithm; The decision support and application layer includes an adaptive control module established based on the analysis results of the data processing and analysis layer.

2. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: The distributed optical fiber sensor network includes several groups of distributed optical fiber sensors laid along key parts of the underground structure, using wavelength division multiplexing and time division multiplexing technology; The wavelength division multiplexing includes using a wavelength separator at the receiving end to separate optical signals of different wavelengths, and setting n sensors to be modulated at wavelengths λ1, λ2, ..., λ n The optical signal is represented as: Among them, A i is the optical signal amplitude, ω i =2πc / λ i , c is the speed of light, is the initial phase, and the optical filter is used to separate the sensor signals at the receiving end. Bragg The Bragg grating, whose reflected light satisfies λ Bragg =2nΛ, where n is the fiber refractive index and Λ is the grating period. The center wavelength changes with the fiber refractive index or grating period. The corresponding sensor information is obtained by detecting the change in the reflected light wavelength. The time division multiplexing technology involves dividing the time axis into multiple time slots. Different sensors send optical signals in different time slots. The transmitter sends optical pulse signals from each sensor in sequence. The receiver receives the signals from the corresponding sensors in the corresponding time slots through time synchronization. Assuming that the length of each time slot is Δt, the optical pulse signal sent by the jth sensor in the mth time slot is expressed as: Among them, rect(x) is the matrix impulse function, τ j is the delay time of the jth sensor signal, ω c is the carrier angular frequency. In a time-division multiplexing system containing N sensors, the total time period T = NΔt.

3. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: The multi-type sensor module includes an acceleration sensor, a pressure sensor, and a displacement sensor. An intelligent sensor node is used to perform local preprocessing and preliminary fusion of the sensor data. The intelligent sensor node is equipped with a microprocessor and a local storage unit. During the local preprocessing, the noise in the raw data is removed by a mean filtering algorithm. The formula is as follows: in, is the filtered data, x i is the original data sequence, N is the filter window size; In the initial fusion, the weighted average fusion algorithm is used to calculate the measurement values y1, y2, ..., y n , the fusion result Y is: Among them, w i is the weight of the corresponding sensor data, and 4. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: In the hybrid wired and wireless transmission network, the data collected by the distributed optical fiber sensors are transmitted using an optical fiber network, and the data collected by the multi-sensor module are transmitted using wireless and wired transmission methods. For sensors close to the data aggregation node, wireless communication technology is used for transmission, and for sensors far away from the data aggregation node, wired Ethernet is used for transmission. The data encryption and transmission module includes encrypting data using an advanced encryption algorithm, which uses a block cipher system to group plaintext into blocks of fixed length and encrypt each block. Let the plaintext block be P and the key be K. The encrypted ciphertext block C is obtained by round transformation. The round transformation includes byte substitution, row shift, column confusion and round key addition operations. The byte substitution performs nonlinear substitution on each byte in the plaintext through an S-box. Let the plaintext byte be x and the byte after the S-box substitution be y, expressed as y=S(x); The transmitted data is checked by a cyclic redundancy check method, which generates a check code by performing a polynomial division operation on the data. Suppose the data polynomial is D(x), the generating polynomial is G(x), and the remainder R(x) of D(x) divided by G(x) is calculated as the check code. The check code is recalculated at the receiving end and compared with the received check code. If they are consistent, the data transmission is correct.

5. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: The multi-source data fusion algorithm extracts time series features through a bidirectional long short-term memory network and introduces an adaptive weight allocation module to optimize the contribution weights of multimodal data in real time; The core calculation of the long short-term memory unit in the bidirectional long short-term memory network includes the update of the input gate, the forget gate, the output gate and the memory unit. The bidirectional long short-term memory network processes the input sequence simultaneously through the forward and backward long short-term memory networks, and takes the weighted sum of the hidden states in the two directions as the final output; The adaptive weight allocation module includes a hybrid decision-making mechanism of fuzzy logic and evidence theory. The fuzzy logic is used to convert the monitoring value of multimodal data into fuzzy linguistic variables. First, the fuzzy membership function of the input data is determined, and fuzzy reasoning is performed according to the pre-established fuzzy rule base to obtain the output fuzzy set. The evidence theory is used to process uncertainty information and assign trust to each data modality through the basic probability distribution function. Let m i is the basic probability distribution function of the i-th data mode, and for all possible hypothesis sets A, it satisfies Where Θ is the recognition framework, and the basic probability distribution functions of different data modalities are then fused through the Dempster combination rule to obtain a comprehensive trust distribution, and then weights are assigned to each data modality in real time based on the trust distribution; Let w i is the weight of the i-th data mode. After being processed by the hybrid decision-making mechanism of fuzzy logic and evidence theory, the weight update formula is: Among them, m i is the trustworthiness of the i-th data modality obtained by evidence theory, is the membership degree of the fuzzy set corresponding to the i-th data mode obtained by fuzzy logic, and n is the total number of data modes.

6. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: In the prediction model based on the deep learning algorithm, the structural mechanics constitutive equation is embedded in the deep learning model as a regularization term. The deep learning model includes an input layer, multiple hidden layers and an output layer. Let the input data be X and the weight matrix of the first layer be W. l , the bias vector is b l , the activation function is σ, then the output H of the lth layer l Expressed as: H l =σ(W l H l-1 +b l ) Among them, H 0 =X, the output of the final output layer is the prediction result of the prediction model The structural mechanics constitutive equation is added as a regularization term to the loss function L of the deep learning model, and the loss based on data fitting is set to L date , the regularization term corresponding to the structural mechanics constitutive equation is L physics , then the total loss function is: L = L date +λL physics , where λ is a hyperparameter that balances the weights of the two; The strain results predicted by the deep learning model are substituted into the constitutive equation and the deviation from the theoretical value is calculated to construct L physics , assuming the strain predicted by the model is The theoretical stress calculated according to the constitutive equation is: The stress predicted by the model is Then L physics It can be expressed as: During the model training process, the weight W is adjusted through the back propagation algorithm l and bias b l , so that the total loss function L is minimized.

7. The underground structure full life cycle management system based on fusion sensor data according to claim 1 is characterized in that: The adaptive control module uses a model predictive control algorithm and a rule-based decision-making mechanism. The rules are pre-established. When the structural state is monitored to trigger specific rule conditions, the corresponding control measures are executed; The model predictive control algorithm solves a finite time domain optimization problem in each control cycle through rolling optimization to determine the current optimal control input. In the underground structure scenario, the key parameters of structural deformation and stress are used as state variables. The state space model of the underground structure is assumed to be: x k+1 =Ax k +Bu k +w k y k =Cx k +v k Among them, x k is the state vector at time k, u k is the control input at time k, y k is the output vector at time k, A, B, and C are the state transfer matrix, input matrix, and output matrix respectively, w k and v k are process noise and measurement noise respectively. The prediction model is based on the current state x k And the control input sequence for the next N steps, predict the state and output for the next N steps, the prediction formula is: x i+1|k =Ax i|k +Bu i|k y i|k =Cx i|k Where i = k, k + 1, ..., k + N - 1, and the optimization goal is to minimize the deviation between the predicted output and the expected output and the change in the control input. The objective function J is constructed as follows: Among them, y ref,i is the expected output at time i, Δu i|k =u i|k -u i-1|k To control the input variation, λ1 and λ2 are weight coefficients, u nominal is the nominal control input. In each control cycle k, the optimal control input sequence is obtained by solving the above optimization problem. And the first element As the control input actually applied at the current moment.

8. A method for managing the entire life cycle of underground structures based on fused sensor data, characterized in that: The following steps are involved: Construction of geological environment and structural models, obtaining geological information through geological exploration technology, combining with geographic information system technology to construct a three-dimensional geological model, and formulate a planning scheme for the monitoring system; Install sensors and perform system debugging according to the planning scheme, use the monitoring system to monitor the construction process of the terrain structure in real time, and adjust the construction process and construction parameters based on the monitoring results, and establish a construction process monitoring feedback mechanism; The monitoring system collects monitoring data of underground structures in real time and performs data analysis to evaluate the health status of the structures; Based on the structural health status assessment results, maintenance decisions are made and maintenance work is carried out according to the maintenance decisions.

9. The method for managing the entire life cycle of underground structures based on fused sensor data according to claim 8, characterized in that: In the sensor installation step, for distributed fiber optic sensors, during tunnel lining construction, the optical fiber is pre-buried in the concrete pouring layer and fixed by a clamp. During the installation process, the optical fiber is connected by a fiber optic fusion splicer, and an optical time domain reflectometer is used to perform real-time detection of the optical fiber to monitor the loss and breakpoints of the optical fiber. After the sensor is installed, the monitoring system is fully debugged, and the data acquisition and processing software is functionally tested. The sensor is calibrated and calibrated. At the same time, a sensor calibration file is established to record the calibration time, calibration results and the next calibration time information.

10. The method for managing the entire life cycle of underground structures based on fused sensor data according to claim 8, characterized in that: The maintenance decision includes the selection of maintenance plan, determination of maintenance time, and allocation of maintenance resources. According to the maintenance decision, the maintenance work is organized and implemented. During the maintenance process, the monitoring system is used to monitor and evaluate the maintenance effect in real time, and a maintenance work file is established to record the maintenance time, maintenance content, materials used and maintenance personnel information.

Images