Foundation pit monitoring system and method of self-repairing wireless sensor network

Abstract

The present application relates to the technical field of foundation pit monitoring, in particular to a self-repairing wireless sensor network foundation pit monitoring system and method, which integrates multi-source heterogeneous data collection, self-diagnosis processing, predictive failure management, network topology management and intelligent resource scheduling modules to achieve high-precision (±0.5mm) displacement monitoring. The system uses multi-dimensional information fusion and preprocessing to improve sensing capability, and uses a hierarchical health assessment system to comprehensively monitor node and network status, laying a foundation for fault prediction and active maintenance. The system can dynamically adjust network topology structure and optimize resource allocation using reinforcement learning algorithms, effectively performing node task allocation, communication power control and sampling frequency adjustment, and significantly improving the efficiency and reliability of foundation pit monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of foundation pit monitoring, in particular to a self-repairing wireless sensor network-based foundation pit monitoring system and method. BACKGROUND

[0002] With the rapid development of infrastructure construction in China, building structures are becoming larger and more complex, which puts higher requirements on the safety of building foundation pits. Although foundation pit construction safety and quality have been widely recognized in China, due to the complexity of the project, accidents in foundation pit construction still occur from time to time.

[0003] In the prior art, a foundation pit deformation monitoring method and system based on machine vision are disclosed in Chinese patent application No. CN117073559A. The method adopts a combination of binocular monitoring host and positioning camera, solves the problem of binocular monitoring host pose disturbance through positioning camera, and solves the problem of deviation between main target displacement and actual foundation pit displacement by setting main target and positioning target on the target vertical rod regarded as a rigid body. Although this technology improves the monitoring accuracy and stability to a certain extent, it still has the following shortcomings:

[0004] 1. The system has a single point failure risk, and any key device failure will cause the entire monitoring system to fail, lacking a redundant backup mechanism;

[0005] 2. Low energy management efficiency, the target device uses a battery for power supply, the battery has a limited life, and the battery needs to be frequently replaced or charged, increasing maintenance costs and downtime;

[0006] 3. Insufficient communication reliability, relying only on a single 4G communication method, which is prone to data transmission interruption in unstable signal areas;

[0007] 4. Lack of data integrity protection mechanism, data loss cannot be recovered when device failure or communication interruption occurs, lacking data reconstruction and compensation algorithms;

[0008] 5. Low system intelligence, unable to automatically adjust monitoring parameters according to environmental conditions, lacking active monitoring and early warning of device health status;

[0009] 6. Fixed network topology, difficult to adjust once deployed, unable to reconstruct the communication network according to changes in site conditions.

[0010] Therefore, there is an urgent need for a foundation pit monitoring system with self-repairing capability, high reliability, and high intelligence to overcome the shortcomings of the prior art. SUMMARY

[0011] The application aims to provide a foundation pit monitoring system of a self-repairing wireless sensor network and a method thereof, and solve the technical problems of single-point failure risk, low energy management efficiency, insufficient communication reliability, and lack of data integrity protection mechanism in the prior art.

[0012] The application provides a foundation pit monitoring system of a self-repairing wireless sensor network, comprising:

[0013] A multi-source heterogeneous data acquisition module is configured to acquire displacement data, environmental parameters, and structural stress data in a foundation pit monitoring area.

[0014] A self-diagnosis processing module is in communication connection with the multi-source heterogeneous data acquisition module, and is configured to pre-process, detect abnormalities, and evaluate health degrees of the multi-source heterogeneous data.

[0015] A predictive failure management module is in communication connection with the self-diagnosis processing module, and is configured to predict equipment failure and generate a data reconstruction strategy based on the health degree evaluation result.

[0016] A network topology management module is in communication connection with the predictive failure management module, and is configured to dynamically adjust a network topology structure according to the equipment failure prediction result.

[0017] An intelligent resource scheduling module is in communication connection with the predictive failure management module and the network topology management module, and is configured to optimize system resource allocation based on a reinforcement learning algorithm, and perform node task allocation, communication power control, and sampling frequency adjustment according to the network topology structure adjustment and the data reconstruction strategy.

[0018] Preferably, the multi-source heterogeneous data acquisition module comprises:

[0019] A displacement sensing unit is configured to acquire displacement data of characteristic points of a foundation pit.

[0020] An environmental parameter acquisition unit is configured to acquire temperature, humidity, air pressure, and illumination intensity.

[0021] A structural stress acquisition unit is configured to acquire vibration and strain data.

[0022] A data preprocessing unit is configured to filter, denoise, and standardize the displacement data, the environmental parameters, and the structural stress data.

[0023] Preferably, the self-diagnosis processing module comprises:

[0024] A hardware self-checking unit is configured to monitor sensor calibration state, communication link quality, and power supply state.

[0025] A health degree calculation unit is configured to generate a health degree score based on a multi-dimensional health index and a service duration attenuation factor.

[0026] an anomaly detection unit configured to detect data anomalies based on an adaptive threshold algorithm and an Isolation Forest method.

[0027] Preferably, the health degree calculation unit calculates the node health degree by the following formula:

[0028] H = (∑(W i · S i )) · (1 - αD),

[0029] wherein H is the node health degree, W i is the weight of each dimension health indicator, S i is the score of each dimension, D is the service duration decay factor, and α is an adjustable parameter with a value range of 0.05-0.15.

[0030] Preferably, the predictive failure management module comprises:

[0031] a failure prediction unit configured to generate future health degree trajectories and failure probabilities based on a time series prediction network with a Transformer architecture;

[0032] a knowledge distillation unit configured to convert a complex model trained in the cloud into a lightweight model suitable for edge deployment; and

[0033] a data reconstruction unit configured to reconstruct missing data based on spatial correlation modeling and multivariate spatiotemporal tensor decomposition algorithms.

[0034] Preferably, the network topology management module comprises:

[0035] a topology awareness unit configured to discover network topology based on a link quality evaluation matrix and node criticality calculation;

[0036] a dynamic topology optimization unit configured to generate an optimized network topology scheme based on node health degree and link quality; and

[0037] a topology migration execution unit configured to implement seamless topology switching through a two-stage commit mechanism.

[0038] Preferably, the intelligent resource scheduling module comprises:

[0039] a resource state awareness unit configured to monitor the states of computing resources, communication resources, and energy resources;

[0040] a multi-objective resource optimization unit configured to generate resource scheduling strategies based on a deep reinforcement learning network; and

[0041] a self-optimization execution unit configured to execute resource allocation decisions through a hierarchical scheduling mechanism and an adaptive control loop.

[0042] As preferred, the resource scheduling strategy includes sampling frequency control, computing task allocation, communication power control and sleep scheduling, and the monitoring accuracy, energy efficiency and system life are balanced by a multi-objective optimization function.

[0043] As preferred, the reconstructed data generated by the data reconstruction unit is subjected to displacement continuity and maximum deformation rate constraints by a Bayesian inference engine, ensuring that the reconstructed data conforms to physical laws.

[0044] The foundation pit monitoring method of the self-repairing wireless sensor network comprises the following steps:

[0045] Collecting displacement data, environmental parameters and structural stress data in the foundation pit monitoring area;

[0046] Pretreating, anomaly detecting and health degree evaluating the displacement data, the environmental parameters and the structural stress data;

[0047] Predicting equipment failure and generating data reconstruction strategy based on the health degree evaluation results;

[0048] Dynamically adjusting the network topology structure according to the equipment failure prediction results; and

[0049] Optimizing system resource allocation based on a reinforcement learning algorithm, and performing node task allocation, communication power control and sampling frequency adjustment according to the network topology structure adjustment and the data reconstruction strategy.

[0050] The beneficial effects of the present application include:

[0051] 1. Through multi-source heterogeneous data fusion and preprocessing mechanism, multi-dimensional information acquisition and analysis are realized, the system perception ability is enhanced, and the displacement monitoring accuracy is improved from ±2mm of the traditional method to ±0.5mm;

[0052] 2. A hierarchical health degree evaluation system is adopted to realize comprehensive monitoring of the node and network state, providing a basis for fault prediction and active maintenance;

[0053] 3. The predictive failure management and data reconstruction framework is introduced, which can predict equipment failure 3 hours in advance and automatically take measures to ensure that the data integrity rate reaches 99.9%;

[0054] 4. An adaptive network topology reconstruction mechanism is constructed, so that the system can dynamically adjust the communication structure, effectively improving the reliability and anti-interference ability of the network;

[0055] 5. Intelligent resource scheduling and self-optimizing system are adopted, and resource allocation is continuously optimized based on reinforcement learning method, so that the average power consumption of the node is reduced by 40%, and the battery life is prolonged by 65%;

[0056] 6. The overall system continuous working ability is improved from 85% of the traditional scheme to 99.7%, the field maintenance frequency is reduced by 68%, and the total life cycle total cost is reduced by 35%. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is the overall architecture schematic diagram of the foundation pit monitoring system of the self-repairing wireless sensor network of the present application.

[0058] Figure 2 is the structural schematic diagram of the multi-source heterogeneous data acquisition module of the present application.

[0059] Figure 3 is the structural schematic diagram of the self-diagnosis processing module of the present application.

[0060] Figure 4 is the working flow chart of the health degree calculation unit of the present application.

[0061] Figure 5 is the structural schematic diagram of the predictive failure management module of the present application.

[0062] Figure 6 is the structural schematic diagram of the network topology management module of the present application.

[0063] Figure 7 is the structural schematic diagram of the intelligent resource scheduling module of the present application.

[0064] Figure 8 is the foundation pit monitoring method flow chart of the self-repairing wireless sensor network of the present application.

[0065] Figure 9 is the monitoring effect comparison chart of the system in a certain project of the present application. DETAILED DESCRIPTION

[0066] Please refer to the accompanying Figures 1-9 , the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application.

[0067] Embodiment 1

[0068] As shown in Figure 1 , the foundation pit monitoring system of the self-repairing wireless sensor network provided by the present application comprises a multi-source heterogeneous data acquisition module 1, a self-diagnosis processing module 2, a predictive failure management module 3, a network topology management module 4 and an intelligent resource scheduling module 5.

[0069] The multi-source heterogeneous data acquisition module 1 is used for collecting displacement data, environmental parameters and structural stress data in the foundation pit monitoring area. The self-diagnosis processing module 2 is in communication connection with the multi-source heterogeneous data acquisition module 1, and is used for pre-processing, abnormality detection and health degree evaluation of the multi-source heterogeneous data. The predictive failure management module 3 is in communication connection with the self-diagnosis processing module 2, and is used for predicting equipment failure and generating data reconstruction strategies based on the health degree evaluation result. The network topology management module 4 is in communication connection with the predictive failure management module 3, and is used for dynamically adjusting the network topology structure according to the equipment failure prediction result. The intelligent resource scheduling module 5 is in communication connection with the predictive failure management module 3 and the network topology management module 4 respectively, and is used for optimizing system resource allocation based on a reinforcement learning algorithm, and performing node task allocation, communication power control and sampling frequency adjustment according to the network topology structure adjustment and the data reconstruction strategy.

[0070] The above modules are interconnected through a wireless communication network or a wired manner to form a complete monitoring system. The system adopts a layered distributed architecture, and each module can be flexibly configured and deployed according to actual needs. In the foundation pit monitoring site, the multi-source heterogeneous data acquisition module 1 is deployed at key monitoring points and connected with the self-diagnosis processing module 2 and the predictive failure management module 3 through a self-organizing network. The network topology management module 4 and the intelligent resource scheduling module 5 can be deployed in the on-site edge computing device, and are responsible for the coordinated management of the overall network and resources.

[0071] The working process of the system is as follows: the multi-source heterogeneous data acquisition module 1 continuously collects foundation pit monitoring data; the self-diagnosis processing module 2 processes the collected data in real time and evaluates the node health state; the predictive failure management module 3 predicts possible failures and formulates countermeasures according to the health state; the network topology management module 4 dynamically adjusts the network structure to adapt to changes; and the intelligent resource scheduling module 5 optimizes resource allocation to ensure efficient operation of the system. This intelligent closed-loop architecture enables the system to have the ability of self-monitoring, self-diagnosis and self-repair, significantly improving the reliability and efficiency of foundation pit monitoring.

[0072] Embodiment 2

[0073] As shown in Figure 2 The multi-source heterogeneous data acquisition module 1 includes a displacement sensing unit 11, an environmental parameter acquisition unit 12, a structural stress acquisition unit 13 and a data preprocessing unit 14.

[0074] The displacement sensing unit 11 uses a micro millimeter wave displacement sensor array instead of a traditional visual target, and each foundation pit feature point is configured with 3 sensors at an angle of 120°, with a working frequency of 76-81 GHz, a measurement accuracy of ±0.5 mm and a measurement range of 0-50 mm. Preferably, the displacement sensing unit 11 uses a TI IWR6843 chip, which has the characteristics of being resistant to light changes and being able to work normally in harsh weather conditions.

[0075] The environmental parameter acquisition unit 12 is used to collect environmental parameters such as temperature, humidity, air pressure and light intensity. Preferably, the temperature sensor has a measurement range of -40°C to 85°C and an accuracy of ±0.5°C; the humidity sensor has a measurement range of 0-100% RH and an accuracy of ±3% RH; the air pressure sensor has a measurement range of 30-110 kPa and an accuracy of ±0.1 kPa; and the light intensity sensor has a measurement range of 0-65535 lux.

[0076] The structural stress acquisition unit 13 is used to collect vibration and strain data. Preferably, the three-axis accelerometer has a measurement range of ±16g and a resolution of 0.001g; and the strain gauge has a measurement range of ±3000με and a resolution of 1με. These data can be used to evaluate the dynamic response characteristics of the foundation pit structure and provide supplementary information for monitoring.

[0077] The data preprocessing unit 14 performs filtering, noise reduction and standardization processing on the collected displacement data, environmental parameters and structural stress data. Preferably, the wavelet transform denoising algorithm is used to eliminate environmental interference, and a sliding window is used for data smoothing processing, with the window size being adaptively adjusted between 7-31. At the same time, local linear regression-based missing value interpolation is implemented, and time domain features (mean, variance, kurtosis), frequency domain features (power spectral density) and change rate features are extracted.

[0078] The data sampling frequency adopts an adaptive adjustment mechanism: in normal state, it is 5 minutes / time, and when an abnormal state is detected, it can be increased to a maximum of 10 seconds / time. The triggering mode of data acquisition includes three modes of timing trigger, event trigger and remote instruction trigger, which can be flexibly switched according to the field requirements.

[0079] Each acquisition unit is connected with the data preprocessing unit 14 through an internal bus, and the preprocessed data is transmitted to the self-diagnosis processing module 2 through a low-power wireless communication module (such as LoRa or ZigBee) for further analysis.

[0080] Example 3

[0081] As shown in Figure 3 The self-diagnosis processing module 2 includes a hardware self-checking unit 21, a health degree calculation unit 22 and an abnormality detection unit 23.

[0082] The hardware self-checking unit 21 is used to monitor the sensor calibration state, the communication link quality and the power supply state. Preferably, the hardware layer self-checking includes sensor calibration state checking, analog-to-digital converter (ADC) self-checking, communication module link testing; resource monitoring includes CPU load (threshold 80%), memory usage (threshold 75%), storage space (threshold 85%); power management includes voltage fluctuation monitoring (±0.2V alarm), charge and discharge cycle count, power prediction algorithm; communication quality evaluation includes received signal strength indication (RSSI) monitoring, packet loss rate statistics, channel quality index (CQI) analysis. The self-checking period is configurable, generally set to full self-checking once an hour and fast self-checking once a minute.

[0083] The health degree calculation unit 22 generates a health degree score based on the multi-dimensional health indicators and the service duration attenuation factor. As shown in Figure 4 , the health degree calculation process includes four steps of indicator collection, weight distribution, health degree calculation and timing update.

[0084] Preferably, the health degree calculation unit 22 adopts a multi-dimensional health indicator weight distribution matrix W = [0.35, 0.25, 0.20, 0.20], respectively corresponding to the four dimensions of sensor function, computing resource, power supply state and communication capability. The health degree calculation formula is:

[0085] H = (W i -S i )(1-D),

[0086] Wherein, H is the node health degree, W i is the weight of each dimension health indicator, S i is the score of each dimension (value range 0-1), D is the service duration attenuation factor, which is an adjustable parameter, and the preferred value range is 0.05-0.15.

[0087] In order to ensure the timing continuity of health degree evaluation, an exponential moving average updating mechanism is adopted:

[0088] H t = αH t-1 +(1-α)H new ,

[0089] Wherein, α is preferably set to 0.7-0.9, H t is the health degree score at time t, H t -1 is the health degree score at the previous time, and H new is the newly calculated health degree value.

[0090] The health score ranges from 0 to 1 and is divided into five levels: excellent (0.9-1.0), good (0.8-0.9), general (0.7-0.8), attention (0.6-0.7), and warning (0.6 or below).

[0091] The anomaly detection unit 23 detects data anomalies based on an adaptive threshold algorithm and the Isolation Forest method. Preferably, the adaptive threshold calculation uses the formula μ+kσ, where μ is the historical mean, σ is the standard deviation, and k is a tunable coefficient ranging from 1.5 to 3.5.

[0092] The anomaly level is divided into three levels: attention (0.7-0.85), warning (0.85-0.95), and emergency (0.95). In addition, the Z-score standardization process is used with the Isolation Forest anomaly point detection algorithm, and the pollution rate parameter is set to 0.03 to effectively identify abnormal patterns in multi-dimensional data.

[0093] The self-diagnosis processing module 2 exchanges data with the multi-source heterogeneous data acquisition module 1 through a secure wireless communication protocol, supports data encryption and integrity verification. The processing results are transmitted to the predictive failure management module 3 in the form of health reports and abnormal events, providing basic data for subsequent failure prediction and management.

[0094] Example 4

[0095] The health calculation unit 22 calculates the node health degree by the following formula:

[0096] H=(W i -S i )(1-D),

[0097] In actual implementation, each parameter of the formula has a clear physical meaning and value method. Each dimension score Si is normalized to have a value range of 0-1. The specific calculation method is as follows:

[0098] The sensing function score S1 consists of three sub-indicators: calibration state score (40% contribution), signal quality score (35% contribution), and responsiveness score (25% contribution).

[0099] The calibration state score is based on the time interval and drift of the last calibration; the signal quality score is based on the signal-to-noise ratio and stability; and the responsiveness score is based on the sensor response time and sampling consistency.

[0100] The computing resource score S2 consists of CPU load, memory usage, and storage state, each with equal weight. The calculation formula is:

[0101]

[0102] where CPUload, MEMuse and STORuse represent the percentage of CPU, memory and storage usage, respectively.

[0103] The power status score S3 considers three factors: battery level, charge cycle count and voltage stability. The calculation formula is:

[0104]

[0105] where BAT level is the battery level percentage, CYCcount is the current charge cycle count, CYCmax is the maximum charge cycle count (usually 500-1000 times), and VOLstability is the voltage stability score (0-1).

[0106] The communication capability score S4 considers three factors: signal strength, packet loss rate and channel quality. The calculation formula is:

[0107] where RSSI score is the normalized RSSI score (0-1), PLOSS rate is the packet loss rate (0-1), and CQI is the channel quality index (0-15).

[0108] The service duration decay factor D represents the ratio of device usage time to design life, and the calculation formula is:

[0109] where T current is the current service time (hours), T design is the design service life (hours). The parameter a is used to adjust the influence of service duration on health, and the larger the value of a, the more obvious the decay effect of usage time on health.

[0110] During the long-term operation of the system, the parameters for health calculation can be fine-tuned according to actual conditions to adapt to different engineering environments and requirements. The health score, as an important basis for system self-repair decision-making, can effectively reflect the overall working state and potential risks of the node.

[0111] Example 5

[0112] As shown in FIG. 3, the predictive failure management module 3 includes a failure prediction unit 31, a knowledge distillation unit 32 and a data reconstruction unit 33. Figure 5

[0113] ​The failure prediction unit 31 generates future health degree trajectories and failure probabilities based on a Transformer architecture-based time series prediction network. Preferably, the network consists of an input layer, an encoding layer, a decoding layer, and an output layer. The input layer receives the health degree time series of the node, with an input window length of 168 hours (one week); the encoding layer contains 8 attention heads with a hidden layer dimension of 512; the decoding layer consists of 3 layers of GRU (Gated Recurrent Unit) with 128 neurons per layer; the output layer predicts the health degree trajectories and failure probabilities for the next 72 hours. In addition, the failure prediction unit 31 also contains a failure type classifier that uses a SoftMax function to classify 7 common failure types (sensor drift, communication failure, power problem, memory corruption, physical damage, software defect, unknown). The early warning mechanism is triggered when the predicted health degree is less than 0.6 or decreases by more than 20% within 3 hours.

[0114] The knowledge distillation unit 32 converts the complex model trained in the cloud into a lightweight model suitable for edge deployment. Preferably, the teacher model is a complex Transformer network trained in the cloud with approximately 10M parameters; the student model is a lightweight network suitable for edge deployment with approximately 300K parameters. The distillation temperature T is set to 2.5 and the soft label weight α is set to 0.7. Through model pruning and quantization (4-bit fixed-point), the computational complexity is reduced by 85%, while the prediction performance loss is kept within an acceptable range (accuracy reduction of no more than 5%). This enables the complex failure prediction algorithm to run efficiently on resource-constrained edge devices.

[0115] The data reconstruction unit 33 reconstructs missing data based on spatial correlation modeling and multivariate spatio-temporal tensor decomposition algorithms. Preferably, first, the proximity relationship matrix is constructed by the Euclidean distance between nodes and the measurement correlation to identify the donation nodes suitable for data reconstruction reference. Then, the low-rank tensor completion algorithm based on CP decomposition is used to process multi-dimensional missing data, which can be represented by the formula:

[0116]

[0117] where X is the original data tensor, W is the mask tensor indicating missing values, is the reconstructed complete tensor, ⊙ represents the Hadamard product, ||·||F F is the Frobenius norm, ||·||K * is the kernel norm, and λ is the regularization parameter.

[0118] In addition, the data reconstruction unit 33 also contains a Bayesian inference engine, combined with prior knowledge of physical constraints (displacement continuity, maximum deformation rate), to ensure that the reconstructed data conforms to the physical law. The quality evaluation of the reconstructed data is based on the consistency test of the historical data pattern, and the confidence interval is set to 95%. For data points that do not meet the requirements of the reconstruction quality, the system will mark and notify the operation and maintenance personnel to conduct on-site inspection.

[0119] The predictive failure management module 3 establishes a bidirectional communication connection with the self-diagnosis processing module 2, the network topology management module 4, and the intelligent resource scheduling module 5. On the one hand, it receives the health degree evaluation results provided by the self-diagnosis processing module 2; on the other hand, it sends the failure prediction and data reconstruction strategies to the network topology management module 4 and the intelligent resource scheduling module 5, and cooperates to complete the system self-repair.

[0120] Embodiment 6

[0121] As shown in Figure 6 The network topology management module 4 includes a topology awareness unit 41, a dynamic topology optimization unit 42, and a topology migration execution unit 43.

[0122] The topology awareness unit 41 discovers the network topology structure based on the link quality evaluation matrix and the node criticality calculation. Preferably, a distributed network topology discovery protocol is adopted, based on the improved RPL (Routing Protocol for Low-Power and Lossy Networks) protocol, supporting the 6LoWPAN (IPv6 Low-Power Wireless Personal Area Network) standard. The link quality evaluation matrix integrates the ETX (Expected Transmission Count) and PRR (Packet Reception Rate) indicators, and the calculation formula is:

[0123]

[0124] Among them, LQI is the link quality index (0-1), and ω is the weight coefficient, preferably 0.6.

[0125] The node criticality is calculated based on the Betweenness Centrality and the Degree Centrality, which is used to identify the key nodes in the network. The network bottleneck identification algorithm is based on the Max-Flow Min-Cut theorem, which is used to discover potential network weak links.

[0126] The dynamic topology optimization unit 42 generates an optimized network topology scheme based on the node health degree and the link quality. Preferably, a graph reconstruction algorithm based on the health degree and the link quality is adopted, and the objective function is:

[0127] min(w1·E+w2·(1-R)),

[0128] Among them, E represents the network energy consumption, R represents the data transmission reliability, and w1 and w2 are weight coefficients (w1+w2=1).

[0129] The constraints include: node health degree is not lower than a threshold (default 0.6), hop limit (maximum 5 hops), delay requirement (lower than 250 ms). In addition, the dynamic topology optimization unit 42 also implements a role dynamic switching mechanism, supports the promotion strategy of ordinary node → relay node → cluster head node, and calculates the role adaptation score based on the energy level, processing capacity and position. Redundant path planning ensures that any node has at least 2 independent paths connected to the gateway, improving the fault tolerance of the network.

[0130] The topology migration execution unit 43 realizes seamless topology switching through a two-stage commit mechanism. Preferably, the topology migration is divided into a preparation phase and an execution phase. The preparation phase verifies the feasibility of the new topology, ensuring that all nodes can correctly receive and process topology update instructions; the execution phase adopts a synchronous switching strategy to uniformly update the routing table and role settings of all nodes at a preset time point. The rollback guarantee mechanism can automatically restore the original topology when switching fails, avoiding network paralysis. The software-defined network control plane realizes centralized policy distribution and distributed execution, improving the flexibility and efficiency of topology adjustment.

[0131] The network topology management module 4 maintains real-time data exchange with the predictive failure management module 3 and the intelligent resource scheduling module 5. It receives failure warning information from the predictive failure management module 3 and sends topology adjustment results to the intelligent resource scheduling module 5 for corresponding resource optimization. The frequency of topology adjustment is dynamically determined according to the network state, and under normal circumstances, it is checked once every 12 hours, and when potential failure or network performance degradation is detected, it triggers immediate checking and adjustment.

[0132] Example 7

[0133] As shown in Figure 7 The intelligent resource scheduling module 5 includes a resource state perception unit 51, a multi-objective resource optimization unit 52, and a self-optimization execution unit 53.

[0134] The resource state perception unit 51 is used to monitor the state of computing resources, communication resources, and energy resources. Preferably, the computing resource monitoring includes CPU usage, memory occupation, and storage space; the communication resource monitoring includes bandwidth occupation, channel utilization, and communication window allocation; the energy resource monitoring includes battery power, energy harvesting efficiency, and power consumption distribution. Task load modeling is based on historical task execution data to build a resource consumption prediction model. The environment perception adaptation mechanism dynamically adjusts the resource strategy according to environmental factors such as light and temperature, for example, increasing the sampling frequency when the light is sufficient, and reducing the energy consumption when the light is insufficient.

[0135] The multi-objective resource optimization unit 52 generates a resource scheduling strategy based on a deep reinforcement learning network. Preferably, a deep Q-learning based method is adopted, the state space includes node health, network topology state, environmental conditions and task queue; the action space includes sampling frequency adjustment, processing task allocation, communication power control and sleep strategy; the reward function is:

[0136] R = w1 Accuracy + w2 EnergyEfficiency + w3 SystemLifespan - w4 RiskFactor,

[0137] where w1, w2, w3, w4 are weight coefficients, preferably set to [0.45, 0.30, 0.20, 0.05]. The deep Q-learning network structure includes: input layer (state vector) → 3 fully connected layers (256→128→64) → output layer (action value). The experience replay buffer capacity is 10000, the batch training size is 64, and the target network update frequency is 100 steps. The system initially adopts a rule-based scheduling strategy, and gradually transitions to a reinforcement learning-based optimization strategy as experience accumulates.

[0138] The self-optimizing execution unit 53 executes resource allocation decisions through a hierarchical scheduling mechanism and an adaptive control loop. Preferably, the hierarchical scheduling mechanism includes three levels of system level (global resource allocation and load balancing, 5 minute cycle), cluster level (regional node cooperation and task sharing, 1 minute cycle) and node level (fine tuning of internal functional modules, 10 second cycle). The adaptive control loop realizes closed-loop feedback regulation based on a PID controller, and the control parameters are dynamically adjusted according to the system response. The conflict resolution mechanism solves scheduling conflicts based on priority and resource dependency graph, ensuring that critical tasks are executed first. Dynamic strategy evaluation continuously monitors the effectiveness of strategy execution and adjusts strategies in real time that do not meet expectations.

[0139] The intelligent resource scheduling module 5 is the core execution unit of the system self-repair engine, which receives the fault management strategy from the predictive fault management module 3 and the topology adjustment result from the network topology management module 4, considers multiple factors to generate the optimal resource allocation scheme, and issues execution instructions to each functional module. Through continuous self-optimization process, the system can maximize energy efficiency and system lifespan while ensuring monitoring accuracy.

[0140] Example 8

[0141] The resource scheduling strategy includes sampling frequency control, computing task allocation, communication power control and sleep scheduling, which balances monitoring accuracy, energy efficiency and system lifespan through a multi-objective optimization function.

[0142] In a specific implementation, the sampling frequency control strategy dynamically adjusts the data acquisition frequency for different types of sensors and different monitoring scenarios. The basic sampling period is set to 300 seconds (5 minutes), supporting event-driven adaptive adjustment, with a sampling frequency range of 60 seconds (high frequency) to 1800 seconds (low frequency). Trigger conditions include: increasing the sampling rate when the node health is below 0.7; increasing the sampling rate when the displacement change rate exceeds 0.2 mm / h; reducing the sampling rate when the battery level is below 30%. This strategy can ensure the collection density of key data and effectively prolong the battery life.

[0143] The computing task allocation strategy reasonably allocates data processing tasks according to the computing capacity and health status of the nodes. Computing-intensive tasks (such as tensor completion) are offloaded to edge gateway processing when CPU utilization exceeds 80%; lightweight tasks (such as anomaly detection) are processed locally when node health is above 0.75 and battery level is above 40%. This differentiated allocation strategy can balance system load and avoid overloading individual nodes.

[0144] The communication power control strategy dynamically adjusts the transmission power according to the link quality, minimizing energy consumption while ensuring communication reliability. When the link quality is high, the power is reduced to 60% of the maximum power; when the link quality is medium, the power is set to 80% of the maximum power. Adaptive power boosting is also supported, temporarily increasing power when packet loss rate is detected to increase. This strategy saves an average of 25% of communication energy compared to the fixed power scheme.

[0145] The sleep scheduling strategy maximizes standby power consumption by reasonably arranging the active period and sleep period of the nodes. Typical sleep cycle parameters are: active period 3 seconds, sleep period 27 seconds, and 0.1 seconds of listening before sleep. When critical events are detected, the system will temporarily extend the active period (up to 300 seconds); when low activity is detected, the sleep period will be increased (with a prolongation factor of 1.5). This strategy reduces the average energy consumption of the node by more than 40%.

[0146] The above four types of strategies are adjusted through a multi-objective optimization function to ensure the overall performance of the system is optimal. The weights of the optimization objectives can be dynamically adjusted according to engineering requirements, such as increasing the monitoring accuracy weight during critical monitoring periods and increasing the system life weight during long-term unattended periods. Tests show that compared with the fixed strategy, this adaptive resource scheduling scheme reduces power consumption by an average of 22.7%, increases data transmission rate by 6.2%, and reduces communication delay by 35.9%.

[0147] Embodiment 9

[0148] The reconstructed data generated by the data reconstruction unit is subjected to displacement continuity and maximum deformation rate constraints through the Bayesian inference engine, ensuring that the reconstructed data conforms to physical laws.

[0149] In a specific implementation, when the system predicts that a node is about to fail or has detected data loss, the data reconstruction process includes four steps: spatial correlation modeling, multivariate spatiotemporal tensor decomposition, application of physical constraints, and reconstruction quality assessment.

[0150] Spatial correlation modeling first calculates the positional relationship and data correlation between nodes in the monitoring network. The positional relationship is quantified by the Euclidean distance; the data correlation is calculated by the Pearson correlation coefficient, and nodes with a correlation coefficient greater than 0.7 are selected as candidate donor nodes. Preferably, the system selects 3-5 donor nodes for each node to build the basis for data reconstruction.

[0151] Multivariate spatiotemporal tensor decomposition uses an improved tensor CP decomposition algorithm to process a third-order tensor containing time, space, and monitoring parameter dimensions. Preferably, the tensor rank is set to where I, J, and K are the sizes of the three dimensions of the tensor. To improve computational efficiency, the alternating least squares method is used to solve the decomposition problem, and the number of iterations is set to 100 or the convergence error is less than 1e -4 .

[0152] For a third-order tensor its CP decomposition form is:

[0153]

[0154] where λ r is the weight coefficient, a r ∈R I , b r ∈R J , c r ∈R K are the factor vectors obtained by decomposition, denotes the vector outer product, and R is the rank of the tensor.

[0155] The optimization objective is to minimize the reconstruction error:

[0156]

[0157] where ||·||F F is the Frobenius norm, is the tensor reconstructed by CP decomposition.

[0158] The solution process uses the alternating least squares (ALS) method, which alternately optimizes each factor matrix (such as A, B, C) until one of the following conditions is met: the number of iterations reaches 100; the convergence error is less than 1×10 -4 .

[0159] The setting can effectively improve the calculation efficiency while ensuring the decomposition accuracy, and is suitable for analysis and processing of large-scale multivariate space-time data.

[0160] The physical constraint is realized by a Bayesian inference engine, and a physical law is introduced into the reconstruction process as prior information. The displacement continuity constraint ensures smooth change of displacement at adjacent time points, and the maximum change rate is usually set to 0.3 mm / h; the maximum deformation rate constraint is set based on engineering experience, and is usually 0.1%-0.3% of the depth of the foundation pit. The Bayesian inference formula is:

[0161] P(D|X)∝P(X|D)·P(D),

[0162] wherein D is the reconstructed data, X is the observed data, P(D) is the prior distribution (including the physical constraint), P(X|D) is the likelihood function, and P(D|X) is the posterior distribution. The reconstructed data conforming to the physical law is obtained by maximum a posteriori probability (MAP) estimation.

[0163] The reconstruction quality evaluation adopts a cross-validation method, and a part of points are randomly shielded on the known data to compare the deviation of the reconstructed value and the true value. The evaluation indexes include root mean square error (RMSE), mean absolute error (MAE) and determination coefficient (R 2 ). Preferably, when the RMSE is less than 0.5 mm, the MAE is less than 0.3 mm, and the R 2 is greater than 0.9, it is considered that the reconstruction quality meets the requirements.

[0164] Practice shows that the data reconstruction method combined with the physical constraint has a reconstruction accuracy improved by about 35% than that of the pure statistical method, and the advantage is more obvious in the case of a higher data missing rate (30%). The method has been verified effective in multiple actual projects, and can guarantee the continuity and reliability of the monitoring data in the case of sensor node failure.

[0165] Embodiment 10

[0166] As shown in Figure 8 , the foundation pit monitoring method of the self-repairing wireless sensor network of the application comprises the following steps:

[0167] Step S1: collecting displacement data, environmental parameters and structural stress data in the foundation pit monitoring area. Specifically, a micro millimeter wave displacement sensor array is used to collect displacement data of characteristic points of the foundation pit; a temperature and humidity sensor, a barometric pressure sensor and an illumination intensity sensor are used to collect environmental parameters; an accelerometer and a strain gauge are used to collect vibration and strain data. The collection frequency is adaptively adjusted according to the system state, and is 5 minutes / time under normal state.

[0168] Step S2: Pre-process, anomaly detection and health degree evaluation on displacement data, environmental parameters and structural stress data. Pre-processing includes filtering, noise reduction and standardization; anomaly detection uses adaptive threshold algorithm and IsolationForest method; health degree evaluation is based on multi-dimensional health index weight matrix and service time decay factor, generating node health score. Health degree calculation formula is H=(∑(W i ·S i ))·(1-αD), where W i is the weight of each dimension health index, S i is the score of each dimension, D is the service time decay factor, and a is the adjustable parameter. This formula clearly shows how to calculate the health degree of the node by weighted sum of each dimension score and considering the service time decay factor.

[0169] Step S3: Predict equipment failure and generate data reconstruction strategy based on health degree evaluation results. Use the time series prediction network of Transformer architecture to generate future health trajectory and failure probability; convert complex model to lightweight edge model through knowledge distillation technology; formulate data reconstruction strategy according to spatial correlation and multivariate spatio-temporal tensor decomposition algorithm, and apply physical constraints through Bayesian inference engine. The prediction time window is set to 72 hours to provide sufficient warning time.

[0170] Step S4: Dynamically adjust network topology according to equipment failure prediction results. Based on link quality evaluation matrix and node criticality calculation, the current network topology is found; generate optimized network topology scheme according to node health degree and link quality; realize seamless topology switching through two-stage commit mechanism, and maintain rollback ability. The optimization goal of network topology is to minimize energy consumption while maximizing data transmission reliability.

[0171] Step S5: Optimize system resource allocation based on reinforcement learning algorithm, and perform node task allocation, communication power control and sampling frequency adjustment according to network topology adjustment and data reconstruction strategy. Generate resource scheduling strategy through deep Q learning network; adopt hierarchical scheduling mechanism and adaptive control loop to execute resource allocation decision; continuously evaluate the effect of strategy execution and optimize. Resource scheduling strategy includes sampling frequency control, computing task allocation, communication power control and sleep scheduling.

[0172] The above steps constitute a closed-loop self-repairing process, and the system continuously executes the five steps to form a self-adaptive monitoring mechanism of perception-diagnosis-prediction-adjustment-optimization. There is a close data dependency between each step: the multi-source data generated in step S1 is input into step S2; the health assessment result of step S2 is input into step S3; the fault prediction and data reconstruction strategy of step S3 is input into steps S4 and S5; the topology adjustment result of step S4 is input into step S5; and the resource optimization result of step S5 affects the data acquisition process of step S1 in turn, forming a complete feedback loop.

[0173] According to engineering requirements, the system can be configured in a fully automatic mode or a semi-automatic mode. In the fully automatic mode, the system autonomously completes fault prediction, topology adjustment and resource optimization; in the semi-automatic mode, the system generates a recommended scheme which is executed after being confirmed by an administrator. This flexible configuration mode adapts to different engineering safety levels and management requirements.

[0174] The self-repairing wireless sensor network-based foundation pit monitoring system of the application has been applied in a large underground engineering, the foundation pit of which has a depth of 36 meters, a circumference of about 580 meters and a dense building area around it. The system deploys 58 monitoring nodes to cover key positions around the foundation pit. The application results are as follows:

[0175] 1. The monitoring accuracy is significantly improved: the displacement monitoring accuracy of the system reaches ±0.5 mm, which is 75% higher than that of the traditional visual scheme (±2 mm). During a rainstorm, the system successfully captured the slight displacement change (1.2 mm / 12h) of the northeast corner of the foundation pit and warned of the potential risk in advance.

[0176] 2. Verification of self-repairing capability: During the 6-month monitoring, the system predicted and handled 13 node faults, with a success rate of 92%. In one typical case, the system predicted a node power failure 3.5 hours in advance, automatically triggered network topology reconstruction and started the data reconstruction algorithm, with no manual intervention throughout the process, and the monitoring data continuity remained 100%.

[0177] 3. Energy efficiency optimization: The average power consumption of the system is reduced by 42% compared with the traditional scheme, and the node battery life is extended from 45 days to 78 days. Through intelligent power control and sleep scheduling, the system significantly reduces the maintenance frequency while ensuring data quality, and the maintenance workload is reduced by 65%.

[0178] 4. Improved communication reliability: The data transmission success rate of the system reaches 99.8%, which is much higher than the 92% of the traditional scheme. In the case of serious electromagnetic interference caused by adjacent construction, the system maintains communication stability through adaptive communication power control and dynamic routing adjustment.

[0179] 5. Overall benefit analysis: the system deployment cost is reduced by about 25% compared with the traditional scheme, and the total life cycle total cost is reduced by 35%. At the same time, the monitoring data quality and reliability are significantly improved, which provides strong support for engineering safety management, avoids a potential safety accident, and produces significant economic and social benefits.

[0180] As shown in Figure 9 The performance of the system of the application and the traditional monitoring system in a certain external interference event of the project is shown. It can be seen that when communication interference and node failure occur, the data quality of the traditional system decreases rapidly, while the system of the application maintains stable monitoring performance through the self-repairing mechanism.

[0181] In summary, the self-repairing wireless sensor network foundation pit monitoring system and method provided by the application break through the technical bottlenecks of the traditional monitoring system such as single point failure risk, low energy management efficiency, and insufficient communication reliability, and provide a new idea for the development of foundation pit monitoring technology.

[0182] The above only describes the preferred embodiments of the application and is not intended to limit the application. The application can have various changes and variations for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A foundation pit monitoring system of a self-repairing wireless sensor network, characterized in that, Comprising: a multi-source heterogeneous data acquisition module for collecting displacement data, environmental parameters, and structural stress data within the foundation pit monitoring area; A self-diagnosis processing module, in communication connection with the multi-source heterogeneous data collection module, is configured to pre-process, detect abnormalities and evaluate health degree of the multi-source heterogeneous data, wherein the health degree evaluation is performed through a multi-dimensional health index weight matrix Computing node health degree wherein , , , are weights of dimensions of sensing function, computing resource, power state and communication capability respectively, , , , are scores of each dimension respectively, and a health degree calculation formula is ， wherein, is the node health, is the weight of each dimension health indicator, is the score of each dimension, is the service time decay factor, , is the adjustable parameter, with a value range of 0.05-0.15, is the current service time, is the design service life; a predictive failure management module in communication with the self-diagnosis processing module, for predicting equipment failure based on the health degree evaluation results through a time series prediction network of a Transformer architecture, the input window length of the time series prediction network being 168 hours, predicting the health degree trajectory and failure probability for the next 72 hours, converting a complex model trained in the cloud into a lightweight model suitable for edge deployment through a knowledge distillation unit, and generating a data reconstruction strategy based on spatial correlation modeling and a multivariate spatio-temporal tensor decomposition algorithm, the data reconstruction strategy imposing displacement continuity constraints and maximum deformation rate constraints through a Bayesian inference engine to ensure that the reconstructed data conforms to the physical laws of foundation pit displacement; The network topology management module is connected in communication with the predictive failure management module, and is configured to dynamically adjust a network topology structure according to the device failure prediction result and a link quality evaluation matrix, the link quality evaluation matrix integrating an expected transmission number ETX and a packet reception rate PRR index, by a formula The link quality index is calculated, wherein is the link quality index, and the value range is 0-1, is the expected transmission number, is the maximum value of the expected transmission number, is the packet reception rate, is a weight coefficient, and the value is 0.6, and seamless topology switching is realized through a two-stage commit mechanism; and the intelligent resource scheduling module is connected in communication with the predictive failure management module and the network topology management module, and is configured to optimize system resource allocation based on a deep reinforcement learning algorithm, the deep reinforcement learning algorithm adopting a deep Q learning network, a state space including node health degree, network topology state, environmental conditions and task queue, an action space including sampling frequency adjustment, processing task allocation, communication power control and sleep strategy, and a reward function being , wherein is the reward value, is the monitoring accuracy, is the energy consumption, is the communication delay, is the system life, , , , is a weight coefficient, and is set to According to the network topology structure adjustment and the data reconstruction strategy, node task allocation, communication power control and sampling frequency adjustment are performed through a hierarchical scheduling mechanism, the hierarchical scheduling mechanism including three levels of system-level global resource allocation and load balancing, cluster-level intra-regional node cooperation and task sharing and node-level fine regulation of internal functional modules, to form a self-repairing closed loop of health degree evaluation-failure prediction-topology adjustment-resource optimization.

2. A method for foundation pit monitoring of a self-repairing wireless sensor network, using the self-repairing wireless sensor network system for foundation pit monitoring according to claim 1, characterized in that, comprising the following steps: collecting displacement data, environmental parameters, and structural stress data within the foundation pit monitoring area; preprocessing, anomaly detection, and health degree evaluation of the displacement data, environmental parameters, and structural stress data through a multi-dimensional health index weight matrix; predicting equipment failure based on the health degree evaluation results through a time series prediction network of a Transformer architecture, and generating a data reconstruction strategy through spatial correlation modeling and a multivariate spatio-temporal tensor decomposition algorithm; dynamically adjusting the network topology structure according to the equipment failure prediction results and a link quality evaluation matrix; and optimizing system resource allocation based on a deep reinforcement learning algorithm, performing node task allocation, communication power control, and sampling frequency adjustment through a hierarchical scheduling mechanism according to the network topology adjustment and the data reconstruction strategy, forming a self-repairing closed loop.

Images