Intelligent sensor network optimization and self-calibration system and method for hoisting machinery
Through the intelligent sensor network system, combined with deep learning and distributed fault prediction algorithms, the measurement accuracy and fault diagnosis problems of the lifting mechanical sensor network in harsh environments are solved, and efficient and real-time sensor calibration and fault repair are achieved.
Patent Information
- Application Number
- CN202510666140.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-22
- Publication Date
- 2025-08-29
AI Technical Summary
The existing lifting machinery sensor networks have insufficient measurement accuracy in high temperature, high humidity and strong vibration environments, high calibration and maintenance costs, poor timeliness of fault diagnosis, low data reliability, and high computational complexity of traditional software calibration algorithms, which is difficult to meet the needs of real-time monitoring.
The intelligent sensor network system is adopted, including sensor nodes, data preprocessing modules, environmental data acquisition modules, self-calibration modules and adaptive adjustment modules, and calibration is used to use deep learning algorithms for calibration, and autonomous diagnosis and repair is achieved through distributed processing modules, combining federated learning and Markov decision-making processes for fault prediction.
It realizes high-precision calibration of the sensor network in harsh environments, reduces calibration costs, improves data reliability, predicts faults in advance and performs independent repairs, and improves the system's real-time monitoring capabilities.
Smart Images
Figure CN120558286A_ABST
Abstract
Description
Technical Field
[0001] This document relates to the field of sensor self-calibration technology, and in particular to an intelligent sensor network optimization and self-calibration system and method for lifting machinery. Background Art
[0002] As a critical piece of heavy equipment, real-time monitoring of the operating status of cranes is crucial to ensuring operational safety. Currently, this type of equipment primarily relies on traditional sensor networks for status monitoring, but existing technical solutions have the following technical drawbacks that need to be addressed:
[0003] First, there's insufficient environmental adaptability. Cranes often operate in harsh conditions of high temperature, high humidity, and strong vibration. Traditional sensors are significantly affected by these environmental factors, making it difficult to guarantee measurement accuracy. Experimental data shows that for every 10°C temperature fluctuation, the zero-point drift of the pressure sensor can reach 0.5% of its full-scale range, severely impacting the accuracy of monitoring data.
[0004] Secondly, the calibration and maintenance costs are high. Existing systems generally use regular manual calibration. According to statistics, the sensor network of large lifting machinery needs to be manually calibrated 2-3 times a month. Not only is the maintenance efficiency low, but the cost of a single calibration is as high as thousands of yuan, which brings a heavy burden to the enterprise.
[0005] Furthermore, fault diagnosis is time-sensitive, and traditional monitoring systems lack effective self-diagnosis capabilities. It takes an average of more than eight hours from the occurrence of a fault to its discovery. During this period, the equipment continues to operate with the fault, posing a major safety hazard.
[0006] In addition, data reliability is low. Under complex working conditions, the sensor false alarm rate is as high as 15%-20%. Frequent false alarms not only interfere with normal operations, but may also cause workers to respond slowly to real alarms.
[0007] While existing technologies attempt to use BP neural networks for software calibration, these technologies struggle to meet the demands of real-time monitoring of lifting machinery due to weak algorithm model generalization, high computational complexity, long training cycles (typically requiring over 48 hours), and dynamic response delays (response time > 500ms). Therefore, developing a sensor network system with environmental adaptability, intelligent self-calibration, and rapid fault diagnosis capabilities has significant engineering application value. Summary of the Invention
[0008] The present invention provides a system and method for optimizing and self-calibrating an intelligent sensor network for a lifting machinery, aiming to solve the above-mentioned problems.
[0009] According to an embodiment of the present invention, a system for optimizing and self-calibrating an intelligent sensor network for a lifting machine is provided, comprising:
[0010] The sensor node is configured to collect first operating status data of the lifting machinery in real time, wherein the first operating status data includes vibration, pressure, and tilt angle;
[0011] A data preprocessing module is used to perform lightweight preprocessing on the operating status data sensor data, extract key features and filter noise, and obtain second operating status data;
[0012] Environmental data acquisition module, used to collect environmental parameters in real time, including temperature, humidity and air pressure data;
[0013] a self-calibration module, calibrating the second operating state data based on a preset deep learning algorithm to obtain third operating state data;
[0014] an adaptive adjustment module, which obtains optimal parameters of the deep learning algorithm based on environmental parameters and recalibrates the second operating state data to obtain final operating state data;
[0015] The distributed processing module performs autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
[0016] According to an embodiment of the present invention, a method for optimizing and self-calibrating an intelligent sensor network for a lifting machinery is provided, comprising:
[0017] S1. The sensor node collects first operating status data of the lifting machinery in real time, where the first operating status data includes vibration, pressure, and tilt angle;
[0018] S2. The data preprocessing module performs lightweight preprocessing on the operating status data sensor data, extracts key features and filters noise, and obtains second operating status data;
[0019] S3. The self-calibration module calibrates the second operating state data based on a preset deep learning algorithm to obtain third operating state data;
[0020] S4. The environmental data acquisition module collects environmental parameters in real time, obtains optimal parameters of the deep learning algorithm based on the environmental parameters, and recalibrates the second operating state data to obtain final operating state data, wherein the environmental parameters include temperature and humidity data and air pressure data;
[0021] S5. Perform autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
[0022] The embodiments of the present invention have the following beneficial effects:
[0023] A deep learning-based calibration algorithm has been developed, significantly improving calibration accuracy compared to traditional methods. A multidimensional feature matrix of the environmental parameter-error relationship has been designed to enable intelligent matching of calibration parameters. Federated learning technology has been innovatively applied to distributed sensor networks, allowing each node to share learning results without exposing the original data. A fault prediction algorithm based on a Markov decision process has been developed, predicting sensor anomalies 8-12 hours in advance. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0025] Figure 1 A schematic diagram of an intelligent sensor network optimization and self-calibration system for a lifting machinery according to an embodiment of the present invention;
[0026] Figure 2 Flowchart of a method for optimizing and self-calibrating an intelligent sensor network for a lifting machinery according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this document.
[0028] System Example
[0029] According to an embodiment of the present invention, a smart sensor network optimization and self-calibration system for lifting machinery is provided. Figure 1 As shown, the intelligent sensor network optimization and self-calibration system for lifting machinery according to the embodiment of the present invention specifically includes:
[0030] The sensor node is used to collect first operating status data of the lifting machinery in real time, wherein the first operating status data includes vibration, pressure, and tilt angle; the sensor node includes a temperature sensor, a humidity sensor, an inclination sensor, a pressure sensor, and a vibration sensor.
[0031] A data preprocessing module is used to perform lightweight preprocessing on the operating status data sensor data, extract key features and filter noise, and obtain second operating status data;
[0032] Environmental data acquisition module, used to collect environmental parameters in real time, including temperature, humidity and air pressure data;
[0033] Sensor nodes distributed at key locations of the lifting machinery are responsible for collecting operating status data (such as vibration, pressure, tilt angle, etc.) and environmental parameters (such as temperature, humidity, air pressure, etc.) in real time. Each sensor node has a certain level of computing power and can perform preliminary processing on the collected data, such as lightweight preprocessing and feature extraction. An edge server is set up in each area to manage the sensor nodes in that area. The edge server receives processed data from the sensor nodes and communicates with other edge servers for federated learning. The edge server can also aggregate and analyze the data of local nodes, providing certain feedback and guidance to the local nodes. As the coordinator of federated learning, the central server is responsible for formulating learning strategies, distributing global model parameters, and collecting model update information uploaded by each edge server. The central server does not directly access the raw data of the sensor nodes and only processes the encrypted and aggregated model parameters.
[0034] a self-calibration module, calibrating the second operating state data based on a preset deep learning algorithm to obtain third operating state data;
[0035] The preset deep learning algorithm achieves dynamic deviation compensation by comparing the measurement data of adjacent sensor nodes. Specifically, the adjacent sensor nodes are grouped using a clustering algorithm, the mean and variance of the measurement data of the nodes in the group are calculated, the measurement data of each node is compared with the mean of the group, and the calibration parameters of the node are dynamically adjusted according to the deviation size.
[0036] Multidimensional characteristic matrix of environmental parameter-error relationship
[0037] an adaptive adjustment module, which obtains optimal parameters of the deep learning algorithm based on environmental parameters and recalibrates the second operating state data to obtain final operating state data;
[0038] The adaptive adjustment module includes generating a multidimensional feature matrix of the environmental parameter-error relationship, which is constructed based on historical environmental parameters and corresponding optimal algorithm parameter data. The optimal parameters of the deep learning algorithm under the current environmental parameters are obtained by looking up a mapping table. The adaptive adjustment module has online learning capabilities and can be dynamically updated according to new environmental parameters and calibration results.
[0039] The multidimensional feature matrix generation method in an embodiment of the present invention includes:
[0040] Utilize the environmental data acquisition module to continuously record environmental parameters such as temperature, humidity, and air pressure during the operation of the crane, and store historical environmental parameter data in time series. The time interval can be set according to actual needs, such as recording every minute or every five minutes.
[0041] While recording the environmental parameters, record the deep learning algorithm parameters (such as weights, biases, etc.) that achieve the best calibration effect after the self-calibration module calibrates the second operating state data under each environmental parameter. Through multiple experiments and actual operations, accumulate the optimal algorithm parameters corresponding to different environmental parameter combinations;
[0042] Check the collected historical environmental parameters and optimal algorithm parameter data to remove outliers and erroneous data. For example, if the temperature and humidity data show values outside the normal range (such as the temperature exceeding the limit temperature of the normal operating environment of the crane), it will be considered an outlier and removed. Normalize the environmental parameters (such as normalizing the temperature value to the [0,1] interval, the humidity value to the [0,1] interval, etc.) and algorithm parameters to make data of different magnitudes comparable, facilitating the subsequent construction of the matrix.
[0043] Environmental parameters such as temperature, humidity, and air pressure are used as different dimensions of the matrix, while also considering the dimensions of the algorithm parameters. For example, if there are three environmental parameters: temperature, humidity, and air pressure, and three key parameters of the deep learning algorithm (such as weights w1, w2, and bias b), a six-dimensional feature matrix is constructed.
[0044] Based on the historical environmental parameters and the corresponding optimal algorithm parameter data, each data point is mapped to the corresponding position in the matrix. For example, if the temperature at a certain moment is 25°C, the humidity is 60%, and the air pressure is 101kPa, the corresponding optimal algorithm parameters are w1=0.5, w2=0.3, and b=0.1. Then, find the position corresponding to these six parameter values in the matrix and record the data set at that position. If there are multiple data points at the same position, calculate their mean or use other aggregation methods to process them.
[0045] The adaptive adjustment module monitors new environmental parameters and calibration results in real time. When new environmental parameters and corresponding calibration results are generated, an appropriate update strategy is adopted based on the similarity between the new data and the existing data in the matrix. For example, if the new environmental parameters are similar to those at a certain location in the matrix, the algorithm parameters at that location are weighted and updated based on the difference in the calibration results, so that the matrix reflects the latest environmental parameter-error relationship.
[0046] The distributed processing module performs autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
[0047] The distributed processing module uses a distributed fault diagnosis algorithm to locate and diagnose sensor network faults through information exchange and collaborative computing between nodes. When a faulty node is detected, the distributed processing module automatically switches to a backup node, marks and records the faulty node, and simultaneously initiates a repair mechanism to perform repairs by adjusting sensor operating parameters or sending repair instructions. The specific steps are as follows:
[0048] Fault detection: Each node monitors its own operating status and data transmission in real time, and exchanges information with adjacent nodes. By comparing the data consistency and correlation between nodes, it determines whether there is a fault.
[0049] Fault localization: When a fault is detected, the location of the faulty node is determined through collaborative computing and information sharing between nodes. Distributed algorithms, such as those based on graph theory or distributed consensus algorithms, can be used to quickly and accurately locate the faulty node.
[0050] Backup node switching: After determining the faulty node, the distributed processing module automatically switches the data collection and processing tasks to the backup node to ensure the normal operation of the sensor network.
[0051] Fault marking and recording: Mark the faulty node and record the time, type and related data of the fault for subsequent analysis and processing.
[0052] Repair: Based on the type and severity of the fault, the corresponding repair mechanism is activated. For simple faults, the sensor's operating parameters can be adjusted to fix them. For more complex faults, repair instructions can be sent to the faulty node to guide it to perform repair operations.
[0053] This embodiment of the present invention introduces a fault prediction algorithm based on a Markov decision process (MDP) to predict sensor anomalies 8-12 hours in advance. The specific details are as follows:
[0054] (1) Define the elements of the MDP model:
[0055] State Set (S): A state is defined by comprehensively considering factors such as sensor measurement values, measurement value change rate, deviation from historical mean, and environmental parameters (temperature, humidity, and air pressure). For example, a vibration sensor measurement value within the normal range with a stable rate of change and minimal deviation from the historical mean, combined with suitable environmental parameters, is considered a normal state. A measurement value exceeding a certain threshold within the normal range, with an abnormal rate of change and adverse environmental parameters, is considered an abnormal state.
[0056] Action Set (A): Sets the actions that can be taken for sensors, including calibration, restart, replacement, and maintaining the status quo. Each action corresponds to different operations and resource consumption. Select the appropriate action in different states to optimize system performance.
[0057] Transition probability (P): This is estimated by analyzing large amounts of historical data and counting the frequency of transitions to various other states after executing a specific action in one state. For example, if a sensor is calibrating while its measurement values are fluctuating significantly, the probability of returning to a normal measurement state, remaining in a state of abnormal fluctuation, or transitioning to a more severe abnormal state is calculated.
[0058] Reward function (R): A reward value is set based on the impact of different state-action combinations on the system. If the action stabilizes the sensor state and reduces the risk of failure, a positive reward is given. Conversely, if the action does not improve the state or even worsens it, a negative reward is given. For example, a successful prediction of an impending sensor failure and prompt replacement, thus avoiding downtime and losses, will result in a high positive reward; an ineffective prediction of a failure, resulting in equipment downtime, will result in a negative reward.
[0059] Discount factor (γ): This factor balances current and future rewards, and ranges from 0 to 1. Given the immediacy and severity of crane failures, the discount factor can be appropriately smaller, such as 0.6-0.8, to prioritize current rewards over immediate failure risks.
[0060] (2) Data collection and preprocessing
[0061] Data Collection: Sensors are used to continuously collect data on the crane's operating status (vibration, pressure, tilt angle, etc.) and environmental parameters (temperature, humidity, air pressure, etc.). Historical sensor failure information and maintenance records are also recorded. This ensures the continuity and accuracy of data collection, providing rich data support for algorithm training.
[0062] Data cleaning: Remove outliers and duplicate values from the collected data and fill in missing values. For example, occasionally abnormal measurements are directly removed, and small amounts of missing measurement data are filled using methods such as averaging adjacent values and linear interpolation.
[0063] Feature engineering: Extract valuable features from raw data, such as calculating the mean, variance, standard deviation, and other statistical quantities of the measured values over a period of time, extracting frequency domain features through methods such as Fourier transform, and extracting the time-frequency characteristics of the signal using wavelet analysis, to enhance the data's ability to characterize the sensor status.
[0064] (3) Model training
[0065] Constructing a training dataset: Based on the defined state set, convert the preprocessed data into a state sequence, and determine the action performed in each state (if no actual action is taken, it is considered a status quo action) and the reward obtained to form a training dataset. Ensure that the training data covers all possible combinations of states, actions, and rewards to improve the generalization ability of the model;
[0066] Select algorithm: Use classic reinforcement learning algorithms such as Q-learning and policy gradient algorithm to train the model.
[0067] Iterative training: During training, the model is run through multiple trials with different initial states. Each trial starts from the initial state, selects an action based on the current policy, observes the new state transition and the reward obtained, and continuously updates the Q value or policy parameters until the model converges (e.g., the Q value no longer changes significantly and the policy becomes stable). Through a large number of trials and iterative training, the model learns the optimal state-action policy.
[0068] (4) Fault prediction and decision-making
[0069] Fault prediction: When new data arrives, the current state of the sensor is determined. Based on the trained model, the probability of the sensor transitioning to a faulty state within a certain period of time is predicted. If the predicted probability exceeds a set threshold (e.g., 0.5), the sensor is considered to be experiencing an impending fault.
[0070] Decision-making: Based on predictions and the optimal strategy learned by the model, the appropriate action is determined. For example, if a sensor is predicted to fail, actions such as replacement or proactive maintenance are executed according to the strategy to prevent the failure or minimize its impact. The cost and feasibility of the action are also considered to ensure the rationality of the decision.
[0071] (5) Model update and optimization
[0072] Online Learning: As new data is continuously generated, the model is trained online, updating model parameters in real time to adapt to changes in sensor characteristics or environmental changes. For example, if sensor aging causes changes in measurement characteristics, the model can adjust its prediction strategy in a timely manner.
[0073] Performance evaluation: Regularly evaluate model performance indicators such as fault prediction accuracy and decision effectiveness using new test datasets. If performance degrades, optimize the model by adjusting model parameters (such as learning rate and discount factor), re-engineering features, or changing the training algorithm.
[0074] Further including:
[0075] Data visualization interface, used to display sensor status, calibration parameters and fault alarm information.
[0076] By adopting the embodiments of the present invention, the following beneficial effects are achieved:
[0077] A deep learning-based calibration algorithm has been developed, significantly improving calibration accuracy compared to traditional methods. A multidimensional feature matrix of the environmental parameter-error relationship has been designed to enable intelligent matching of calibration parameters. Federated learning technology has been innovatively applied to distributed sensor networks, allowing each node to share learning results without exposing the original data. A fault prediction algorithm based on a Markov decision process has been developed, predicting sensor anomalies 8-12 hours in advance.
[0078] Method Example
[0079] According to an embodiment of the present invention, a method for optimizing and self-calibrating an intelligent sensor network for a lifting machinery is provided. Figure 2 This is a flow chart of the intelligent sensor network optimization and self-calibration method for lifting machinery according to an embodiment of the present invention. Figure 2 The intelligent sensor network optimization and self-calibration method for a lifting machinery according to the embodiment of the present invention specifically includes:
[0080] S1. The sensor node collects first operating status data of the lifting machinery in real time, where the first operating status data includes vibration, pressure, and tilt angle;
[0081] S2. The data preprocessing module performs lightweight preprocessing on the operating status data sensor data, extracts key features and filters noise, and obtains second operating status data;
[0082] S3. The self-calibration module calibrates the second operating state data based on a preset deep learning algorithm to obtain third operating state data;
[0083] S4. The environmental data acquisition module collects environmental parameters in real time, obtains optimal parameters of the deep learning algorithm based on the environmental parameters, and recalibrates the second operating state data to obtain final operating state data, wherein the environmental parameters include temperature and humidity data and air pressure data;
[0084] S5. Perform autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
[0085] The preset deep learning algorithm achieves dynamic deviation compensation by comparing the measurement data of adjacent sensor nodes. Specifically, the adjacent sensor nodes are grouped using a clustering algorithm, the mean and variance of the measurement data of the nodes in the group are calculated, the measurement data of each node is compared with the mean of the group, and the calibration parameters of the node are dynamically adjusted according to the deviation size.
[0086] The S4 specifically includes: generating a multidimensional feature matrix of the environmental parameter-error relationship, wherein the multidimensional feature matrix of the environmental parameter-error relationship is constructed based on historical environmental parameters and corresponding optimal algorithm parameter data, and obtaining the optimal parameters of the deep learning algorithm under the current environmental parameters by looking up the mapping table. The adaptive adjustment module has online learning capabilities and can be dynamically updated according to new environmental parameters and calibration results.
[0087] The S5 adopts a distributed fault diagnosis algorithm to realize fault location and diagnosis of the sensor network through information interaction and collaborative computing between nodes. When a faulty node is detected, the distributed processing module automatically switches to a backup node, marks and records the faulty node, and starts a repair mechanism to repair it by adjusting the working parameters of the sensor or sending repair instructions.
[0088] The method further comprises:
[0089] The sensor status, calibration parameters and fault alarm information are displayed through the data visualization interface.
[0090] By adopting the embodiments of the present invention, the following beneficial effects are achieved:
[0091] A deep learning-based calibration algorithm has been developed, significantly improving calibration accuracy compared to traditional methods. A multidimensional feature matrix of the environmental parameter-error relationship has been designed to enable intelligent matching of calibration parameters. Federated learning technology has been innovatively applied to distributed sensor networks, allowing each node to share learning results without exposing the original data. A fault prediction algorithm based on a Markov decision process has been developed, predicting sensor anomalies 8-12 hours in advance.
[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. An intelligent sensor network optimization and self-calibration system for lifting machinery, characterized in that include: The sensor node is configured to collect first operating status data of the lifting machinery in real time, wherein the first operating status data includes vibration, pressure, and tilt angle; A data preprocessing module is used to perform lightweight preprocessing on the operating status data sensor data, extract key features and filter noise, and obtain second operating status data; Environmental data acquisition module, used to collect environmental parameters in real time, including temperature, humidity and air pressure data; a self-calibration module, calibrating the second operating state data based on a preset deep learning algorithm to obtain third operating state data; an adaptive adjustment module, which obtains optimal parameters of the deep learning algorithm based on environmental parameters and recalibrates the second operating state data to obtain final operating state data; The distributed processing module performs autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
2. The method according to claim 1, characterized in that The preset deep learning algorithm achieves dynamic deviation compensation by comparing the measurement data of adjacent sensor nodes. Specifically, the adjacent sensor nodes are grouped using a clustering algorithm, the mean and variance of the measurement data of the nodes in the group are calculated, the measurement data of each node is compared with the mean of the group, and the calibration parameters of the node are dynamically adjusted according to the deviation size.
3. The system according to claim 1, wherein: The adaptive adjustment module includes generating a multidimensional feature matrix of the environmental parameter-error relationship, which is constructed based on historical environmental parameters and corresponding optimal algorithm parameter data. The optimal parameters of the deep learning algorithm under the current environmental parameters are obtained by looking up a mapping table. The adaptive adjustment module has online learning capabilities and can be dynamically updated according to new environmental parameters and calibration results.
4. The system according to claim 1, wherein: The distributed processing module adopts a distributed fault diagnosis algorithm to realize fault location and diagnosis of the sensor network through information interaction and collaborative computing between nodes. When a faulty node is detected, the distributed processing module automatically switches to a backup node, marks and records the faulty node, and starts a repair mechanism at the same time to perform repairs by adjusting the working parameters of the sensor or sending repair instructions.
5. The system according to claim 1, wherein: The system further comprises: Data visualization interface, used to display sensor status, calibration parameters and fault alarm information.
6. A method for optimizing and self-calibrating an intelligent sensor network for a lifting machine, characterized in that: include: S1. The sensor node collects first operating status data of the lifting machinery in real time, where the first operating status data includes vibration, pressure, and tilt angle; S2. The data preprocessing module performs lightweight preprocessing on the operating status data sensor data, extracts key features and filters noise, and obtains second operating status data; S3. The self-calibration module calibrates the second operating state data based on a preset deep learning algorithm to obtain third operating state data; S4. The environmental data acquisition module acquires environmental parameters in real time, obtains optimal parameters of the deep learning algorithm based on the environmental parameters, and recalibrates the second operating state data to obtain final operating state data, wherein the environmental parameters include temperature, humidity, and air pressure data; S5. Perform autonomous diagnosis and repair functions of the sensor network according to the final operating status data.
7. The method according to claim 6, characterized in that The preset deep learning algorithm achieves dynamic deviation compensation by comparing the measurement data of adjacent sensor nodes. Specifically, the adjacent sensor nodes are grouped using a clustering algorithm, the mean and variance of the measurement data of the nodes in the group are calculated, the measurement data of each node is compared with the mean of the group, and the calibration parameters of the node are dynamically adjusted according to the deviation size.
8. The method according to claim 6, characterized in that The S4 specifically includes: generating a multidimensional feature matrix of the environmental parameter-error relationship, wherein the multidimensional feature matrix of the environmental parameter-error relationship is constructed based on historical environmental parameters and corresponding optimal algorithm parameter data, and obtaining the optimal parameters of the deep learning algorithm under the current environmental parameters by looking up the mapping table. The adaptive adjustment module has online learning capabilities and can be dynamically updated according to new environmental parameters and calibration results.
9. The method according to claim 6, characterized in that The S5 adopts a distributed fault diagnosis algorithm to realize fault location and diagnosis of the sensor network through information interaction and collaborative computing between nodes. When a faulty node is detected, the distributed processing module automatically switches to a backup node, marks and records the faulty node, and starts a repair mechanism to repair it by adjusting the working parameters of the sensor or sending repair instructions.
10. The method according to claim 6, characterized in that The method further comprises: The sensor status, calibration parameters and fault alarm information are displayed through the data visualization interface.