A fault locating method and system in industrial scenarios under crowd-sourced sensing

Abstract

The application discloses a kind of industrial scene under crowd-sourcing perception fault location method and system.This method is: in the deployment of a plurality of sensing nodes in target industrial monitoring area, form distributed sensing network;Wherein, the running state data and environmental data of each selected industrial equipment in target industrial monitoring area are collected in real time by each sensing node corresponding to one or more sensing nodes;Each sensing node uploads to distributed collaborative platform;Distributed collaborative platform extracts the data features of collected data, and the data features corresponding to the same industrial equipment are weighted and fused as the fusion features of the industrial equipment;Each sensing node calculates the local fault location initial value according to its own position information and the fusion features of the industrial equipment, and sends it to other sensing nodes of the industrial equipment for cross-validation, corrects the positioning deviation, and finally determines the accurate position of the fault point.The positioning accuracy of the present application is significantly improved, and the real-time performance and response efficiency are greatly optimized.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of fault diagnosis and crowdsourced sensing technology, and relates to a fault location method and system based on crowdsourced sensing in industrial scenarios. Focusing on industrial scenarios, this invention is applicable to the rapid detection and accurate location of faults in various industrial systems (such as industrial production equipment, workshop conveyor lines, industrial power supply and distribution equipment, industrial pipelines, machine tools, etc.), and is particularly suitable for complex industrial environments such as industrial workshops, factories, and industrial plant areas. It can be widely applied to equipment operation and maintenance and fault handling in industries such as machinery manufacturing, chemical engineering, metallurgy, and building materials. Background Technology

[0002] With the rapid development of modern industrial technology, Internet of Things (IoT) technology, and communication technology, the scale of various systems is constantly expanding and their structures are becoming increasingly complex, significantly increasing the probability of failures. The efficiency and accuracy of fault location directly determine the system's operation and maintenance costs, operational stability, and security. Traditional fault location methods are mainly divided into two categories: one is a location method based on a single device or single-point monitoring, relying on professional sensors or manual inspections. This method suffers from problems such as limited sensing range, positioning lag, high labor costs, and poor adaptability to complex environments. Especially in confined environments and areas with signal attenuation, single-point monitoring cannot obtain comprehensive fault information, leading to large positioning errors. The other is a location method based on centralized sensing, which collects and analyzes data through a unified central platform. However, this method suffers from drawbacks such as high data transmission pressure, high response latency, weak anti-interference ability, and poor fault tolerance. Furthermore, when the central platform fails, the entire positioning system will be paralyzed.

[0003] Swarm intelligence sensing technology, as a novel technology integrating collective intelligence and distributed sensing, achieves comprehensive perception of the environment or system status through the collaborative work of multiple sensing nodes (such as smart terminals, sensors, and intelligent agents). It possesses self-organization, self-learning, and self-optimization characteristics and has shown promising application potential in fields such as sensing and localization, and fault diagnosis. However, current technologies applying swarm intelligence sensing to fault localization still have significant shortcomings: First, the data collected by sensing nodes suffers from multi-source heterogeneity, redundancy, and high noise levels, lacking effective data preprocessing and fusion mechanisms, leading to inaccurate fault feature extraction. Second, the collaboration between nodes is insufficient, often consisting of simple data aggregation, failing to fully leverage the advantages of collective intelligence and hindering the complementarity and verification of fault information. Third, the localization algorithm lacks specificity and has not been optimized to suit the distributed characteristics of swarm intelligence sensing, making it difficult to meet practical needs in terms of positioning accuracy and real-time performance in complex environments (such as electromagnetic interference, signal obstruction, and node failure). Fourth, there is a lack of effective abnormal node identification mechanisms; erroneous information sent by malicious or faulty nodes can severely affect the reliability of the localization results.

[0004] For example, in confined, enclosed areas of industrial workshops (such as inside machine tools, pipe cavities, and electrical distribution rooms), traditional positioning technologies cannot achieve accurate location due to equipment obstruction and electromagnetic interference. In industrial production lines, traditional centralized fault location methods suffer from high response delays, making it difficult to quickly pinpoint fault locations, leading to production line shutdowns, the expansion of fault areas, and significant economic losses. Therefore, there is an urgent need for a method that can fully leverage the advantages of collective sensing, overcome the aforementioned shortcomings of existing technologies, adapt to industrial scenarios, and achieve efficient, accurate, and reliable fault location. Summary of the Invention

[0005] In view of the problems of limited sensing range, low positioning accuracy, poor real-time performance, weak anti-interference ability and insufficient fault tolerance in existing fault location methods, as well as the problems of insufficient data fusion, poor node coordination and abnormal nodes affecting positioning reliability in the application of swarm intelligence sensing technology in industrial fault location, the purpose of this invention is to provide a fault location method and system under swarm intelligence sensing in industrial scenarios.

[0006] This invention provides an efficient data preprocessing and multi-source fusion mechanism to solve the problems of redundancy, noise and heterogeneity in the data collected by the swarm sensing nodes, accurately extract the fault characteristics of industrial equipment, and provide reliable data support for fault location.

[0007] This invention constructs a distributed collaborative sensing architecture, which fully leverages the advantages of collective intelligence to achieve information complementarity and collaborative verification among sensing nodes, thereby improving the comprehensiveness and accuracy of fault location.

[0008] This invention designs a targeted fault location algorithm, combines the distributed characteristics of crowd sensing, optimizes the location logic, improves the accuracy and real-time performance of fault location in complex industrial environments, and reduces location errors.

[0009] This invention adds an abnormal node identification and fault tolerance mechanism to filter erroneous information sent by malicious or faulty nodes, thereby improving the reliability and robustness of the positioning system.

[0010] This invention automates and intelligently locates faults, reduces manual intervention, lowers the cost of manual operation and maintenance in industrial scenarios, adapts to complex industrial environments such as industrial workshops, factories, and plant areas, and improves the operation and maintenance efficiency and operational stability of industrial systems.

[0011] The technical solution of this invention is as follows: A fault location method based on crowdsourced sensing in an industrial setting, comprising the following steps: Several sensing nodes are deployed in the target industrial monitoring area to form a distributed sensing network; each selected industrial device in the target industrial monitoring area corresponds to one or more sensing nodes. Each sensing node collects real-time operational status data and environmental data of the corresponding industrial equipment and uploads them to the distributed collaborative platform. The distributed collaborative platform extracts the data features of the collected data, and performs weighted fusion of the data features corresponding to the same industrial equipment to serve as the fusion feature of that industrial equipment. Each sensing node calculates an initial local fault location value based on its own location information and the fusion characteristics of the industrial equipment it is located in, and sends it to other sensing nodes of the industrial equipment for cross-verification to correct the positioning deviation and finally determine the precise location of the fault point.

[0012] The preferred method for cross-validation, correcting for positioning deviations, and ultimately determining the precise location of the fault is as follows: Each sensing node synchronizes its own calculated initial fault location value to neighboring sensing nodes and receives the initial fault location value from neighboring sensing nodes. Each sensing node performs a weighted fusion of its own calculated initial fault location value and the coordinate position in the received initial fault location value to obtain an intermediate cross-location value, which is then sent to other sensing nodes in the industrial equipment. Each sensing node iteratively calculates the coordinate correction amount based on the currently summarized initial fault location value and the intermediate value of cross-location. It then sends the information to other sensing nodes on the same industrial equipment for mutual verification until the positioning deviation of each sensing node on the same industrial equipment is less than the preset accuracy, thus finally determining the precise location of the fault point; among which, These are the correction values ​​for the x and y axis coordinates. The weights of the initial values ​​for the i-th fault location are ( , ) represents the coordinates of the intermediate value of the i-th intersection positioning, ) represents the fault location coordinates obtained in the current iteration, and n represents the number of initial fault location values.

[0013] Preferred, through Iteratively adjust the fault location coordinates; where, , ( ) represents the coordinates of the fault location obtained in the k-th iteration. The coordinates of the fault point obtained in the (k+1)th iteration are used to locate the fault point.

[0014] Preferably, the method for obtaining the cross-location intermediate value is as follows: calculate the joint weight of each fault location initial value; and perform a weighted average of the fault point coordinates in the currently summarized fault location initial values ​​to obtain the cross-location intermediate value.

[0015] Preferably, the sensing node filters the aggregated initial fault location values ​​and deletes those that exceed a preset threshold.

[0016] Preferably, the initial values ​​for fault location include fault point coordinates, confidence level, and feature matching degree.

[0017] Preferably, the precise location of the finally determined fault point is compared with the corresponding historical fault data. If the positioning error exceeds the preset threshold, the relevant parameters are adjusted and the initial value of the local fault positioning of each sensing node is reset to optimize the precise location of the finally determined fault point.

[0018] Preferably, the distributed collaborative platform employs an abnormal node detection model to monitor the uploaded data and operational status of each sensing node in real time, identify malicious nodes, faulty nodes, and data anomaly nodes, mark and isolate them, prohibit them from participating in data fusion and location calculations, and simultaneously activate a node redundancy replacement mechanism, with adjacent healthy sensing nodes replacing their sensing functions; the data distance between nodes and the data anomaly degree in the particle swarm optimization algorithm are combined as the fitness function of the particle swarm optimization algorithm to obtain an improved particle swarm optimization algorithm, and the abnormal node detection model uses the improved particle swarm optimization algorithm to identify malicious nodes, faulty nodes, and data anomaly nodes.

[0019] Preferably, the sensing node includes multiple types of sensors and communication modules; each sensing node adopts a heterogeneous deployment mode, and the sensing nodes are networked using a self-organizing communication protocol; the data features extracted from the collected data include vibration data features, current / voltage data features, image data features, and environmental data features.

[0020] A fault location system based on collective intelligent sensing in an industrial setting, characterized by comprising a sensing node deployment module, a data acquisition and preprocessing module, a multi-source data fusion module, and a distributed collaborative positioning module; The sensing node deployment module is used to deploy a number of sensing nodes in the target industrial monitoring area to form a distributed sensing network; wherein each selected industrial device in the target industrial monitoring area corresponds to one or more sensing nodes. The data acquisition and preprocessing module is used to collect the operating status data and environmental data of the corresponding industrial equipment and upload them to the distributed collaborative platform. The multi-source data fusion module is used to extract data features from the collected data and perform weighted fusion of data features corresponding to the same industrial equipment as the fusion features of the industrial equipment. The distributed collaborative positioning module is used to calculate the initial value of local fault location based on the location information of the sensing node and the fusion characteristics of the industrial equipment it is located on, and send it to other sensing nodes of the industrial equipment for cross-verification, correct the positioning deviation, and finally determine the precise location of the fault point.

[0021] Compared with the prior art, the present invention has the following significant advantages: 1) Positioning accuracy is significantly improved By using multi-source heterogeneous data fusion preprocessing technology, redundant data and interference noise in industrial scenarios (such as data deviations caused by machine tool vibration, workshop electromagnetic interference, and dust) are removed, and the fault characteristics of industrial equipment (such as machine tool vibration peaks, pipeline pressure fluctuations, and sudden changes in motor current) are accurately extracted. In the fault location of industrial power supply and distribution equipment, the positioning accuracy can be improved from the workshop level to the equipment point level, effectively avoiding the expansion of the fault range and reducing the risk of industrial production interruption.

[0022] 2) Real-time performance and response efficiency have been significantly improved. Employing a distributed collaborative architecture, each sensing node autonomously completes data acquisition, preprocessing, and initial local positioning calculations for industrial equipment. Nodes rapidly exchange information through self-organizing communication, eliminating reliance on a central platform for processing and significantly reducing data transmission pressure and response latency in industrial scenarios. Upon fault occurrence, data acquisition, fusion, and localization can be completed quickly. Compared to traditional centralized positioning methods, response time is shortened, enabling rapid location and handling of industrial equipment faults, reducing downtime and economic losses on industrial production lines, and adapting to the maintenance needs of continuous industrial production. It allows for rapid fault handling, preventing significant economic losses caused by prolonged production line downtime.

[0023] 3) Strong anti-interference ability and robustness By introducing swarm intelligence-driven anomaly node identification technology, abnormal nodes in industrial scenarios can be quickly identified and isolated (such as faulty nodes caused by workshop dust, equipment vibration, chemical corrosion, or nodes that maliciously report false data), filtering out erroneous data and avoiding interference from abnormal information on the positioning results. The sensing nodes adopt a self-organizing network mode, and can automatically reorganize the network after some nodes fail, ensuring the continuity of the sensing network. At the same time, multi-source data fusion technology can make up for the limitations of single data types and improve the system's adaptability in complex environments such as dense equipment, strong electromagnetic interference, and enclosed spaces in industrial workshops. Compared with traditional positioning methods, its robustness is significantly enhanced, adapting to the complex operation and maintenance needs of various industrial scenarios.

[0024] 4) High degree of automation and intelligence This system automates the entire process of industrial equipment fault location, from data acquisition, preprocessing, and fusion to anomaly identification, fault location, and result optimization, eliminating the need for manual intervention and reducing the cost of manual inspections and maintenance in industrial settings. Furthermore, the system possesses self-learning and self-optimization capabilities, continuously adjusting algorithm parameters based on historical fault data from industrial equipment to improve the accuracy and efficiency of subsequent location tracking, reducing reliance on specialized technical personnel. It is suitable for long-term monitoring and fault location of large-scale industrial plants and various types of industrial equipment, aligning with the trend of intelligent industrial operation and maintenance. In pipeline fault location within chemical plants, it can automatically complete data cleaning, fusion, and location, with the location results synchronously fed back to the plant's operation and maintenance platform, eliminating the need for manual troubleshooting and significantly reducing manual maintenance costs.

[0025] 5) Wide range of applicable scenarios and strong versatility By employing heterogeneous sensing node deployment and flexible algorithm configuration, the sensor type (such as vibration, pressure, temperature, and gas sensors), algorithm parameters, and deployment mode can be adjusted according to the needs of different industrial scenarios (such as machining, chemical industry, metallurgy, power supply and distribution, etc.) without requiring significant modifications to the core technology. It is adaptable to various complex industrial environments such as industrial workshops, factories, and plant areas, as well as the fault location needs of different types of industrial equipment (machine tools, pipelines, power supply and distribution equipment, etc.). It has strong versatility and broad application prospects.

[0026] 6) Low operation and maintenance costs and high practicality The sensing nodes adopt a low-power design, enabling long-term stable operation and reducing node replacement and maintenance costs in industrial scenarios. The distributed architecture eliminates the need for a complex central control platform, reducing system deployment costs in industrial plants. Meanwhile, rapid fault location and handling can reduce industrial equipment downtime and economic losses caused by faults. Compared with traditional fault location methods, it can reduce industrial operation and maintenance costs, meeting the operation and maintenance needs of industrial enterprises to reduce costs and increase efficiency, making it highly practical.

[0027] Therefore, this invention can effectively solve the defects of existing fault location methods in industrial scenarios, give full play to the advantages of swarm intelligence sensing technology, improve the efficiency, accuracy and reliability of fault location of industrial equipment, adapt to the actual needs of industrial operation and maintenance, and has broad industrial application value. Attached Figure Description

[0028] Figure 1 This is a flowchart of the method of the present invention.

[0029] Figure 2 This is a system diagram of the present invention. Detailed Implementation

[0030] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0031] The core technical solution of this invention is to construct a full-process fault location system consisting of "sensing node deployment - data acquisition and preprocessing - multi-source data fusion - abnormal node identification - distributed collaborative positioning - positioning result verification and optimization". By combining and optimizing with industrial scenarios, a core self-developed technology is formed to achieve rapid, accurate and reliable fault location.

[0032] The operating logic of this invention is as follows: Figure 1 As shown, the overall process is divided into 6 core steps, which are sequentially connected and optimized in a closed loop. The specific process is as follows: 1. Deployment of sensing nodes Based on the structural characteristics, equipment distribution, and production environment of the target industrial monitoring area, several sensing nodes are deployed to form a distributed sensing network. Each selected device within the target industrial monitoring area corresponds to one or more sensing nodes. The sensing nodes adopt a heterogeneous deployment mode, containing various types of sensors and communication modules. They can collect multi-dimensional industrial equipment and environmental data. The sensing nodes support self-organizing communication protocols, eliminating the need to rely on fixed communication infrastructure in industrial plants. This adapts to the deployment requirements of complex environments such as dense equipment, strong electromagnetic interference, and enclosed spaces in industrial workshops. The hybrid deployment and dynamic networking of heterogeneous nodes are core proprietary technologies.

[0033] 2. Data Acquisition and Preprocessing Each sensing node collects real-time data on the operating status of industrial equipment (such as machine tool vibration frequency, motor current and voltage changes, pipeline pressure, and equipment operating noise) and industrial environmental data (such as workshop temperature, humidity, dust concentration, and electromagnetic interference intensity). The collected raw data is then standardized in format (unifying data units, e.g., unifying all current data to A and pressure data to MPa) and deduplicated (using "node ID + collection timestamp + data type preprocessing, where: Standardize the format: unify the data units, for example, unify all current data to A and pressure data to MPa; Deduplication: Using "node ID + collection timestamp + data type" as a unique identifier, traverse all original data, delete redundant data with duplicate identifiers, and keep only one valid data. Missing value handling: For data with missing time ≤ 5 seconds, linear interpolation is used to calculate and fill in the missing value based on the valid data adjacent to the missing data to ensure data continuity; for data with missing time > 5 seconds, mean imputation is used to calculate the historical average value of the same type of data at that node (the average value of valid data in the last 10 minutes) and fill in the missing data. Noise filtering: Wavelet transform algorithm is used to decompose the original data at multiple scales, separate the signal from the noise, remove high-frequency noise components and retain low-frequency effective signals; at the same time, sliding window filtering method (window size set to 5-10 data points) is combined to smooth data fluctuations and further filter instantaneous noise. Outlier removal: Using the 3σ principle, the mean μ and standard deviation σ of each data type are calculated. Data exceeding the range [μ-3σ, μ+3σ] are identified as outliers and directly removed. For critical data of industrial equipment (such as motor current), a reasonable threshold is set based on the equipment's rated operating parameters (e.g., a current threshold of 0.5-1.5 times the equipment's rated current). Data exceeding this threshold are identified as outliers and removed. Normalization (min-max normalization algorithm): Preliminary feature extraction: 1) Vibration data characteristics: The FFT (Fast Fourier Transform) algorithm is used to convert the time-domain vibration data into frequency-domain data and extract preliminary features such as vibration frequency, peak value, and RMS value. 2) Current / voltage data characteristics: Extract peak values, valley values, abrupt changes, average values, and other characteristics of current / voltage to capture current / voltage fluctuations during equipment failures; 3) Image data features: Using the CNN (Convolutional Neural Network) algorithm, preliminary features such as surface defects, wear marks, and abnormal stains on the equipment in the image are extracted; 4) Environmental data (temperature, humidity, dust, etc.) characteristics: Extract abnormal fluctuation values ​​and durations of data that exceed the normal range to help determine the correlation between equipment failure and environmental factors.

[0034] After preprocessing, each node uploads the processed data, its own location information, and the collection timestamp to the distributed collaborative platform, while retaining local data backups to avoid data loss. The collaborative logic of preprocessing and data uploading has been independently developed and optimized to meet the integration needs of multiple devices and multiple data in industrial scenarios.

[0035] 3. Multi-source data fusion The distributed collaborative platform receives data uploaded by various sensing nodes and uses a combination of multimodal information fusion technology and weighted average algorithm to uniformly analyze and fuse heterogeneous data to extract fault feature parameters. During the fusion process, the weighted average algorithm is used to allocate weights according to the reliability of different types of data, so as to achieve complementarity of fault information, improve the accuracy of fault feature extraction, and solve the problem that a single data type cannot fully reflect the fault status of industrial equipment. The three-level processing flow of "cleaning-feature extraction-weighted fusion" and the weight allocation logic are the core self-developed technologies.

[0036] 4. Abnormal Node Identification An abnormal node detection model based on swarm intelligence algorithms and node operation status monitoring technology is adopted. This model collects the operation parameter data of each node, and generates standardized node operation data through time-series acquisition, format normalization, and noise reduction and redundancy removal. It then determines whether the current operation data is within a pre-set normal threshold range and outputs the determination result.

[0037] The abnormal node detection model monitors the uploaded data and operating status of each sensing node in real time, identifying malicious nodes, faulty nodes, and data anomaly nodes. The identified abnormal nodes are marked and isolated, prohibiting them from participating in data fusion and positioning calculations. At the same time, a node redundancy replacement mechanism is activated, with adjacent healthy nodes replacing their sensing functions to ensure that the data acquisition and positioning process is not affected. The abnormal node detection model combined with the particle swarm optimization algorithm and the node redundancy replacement logic are the core self-developed technologies.

[0038] 5. Distributed collaborative positioning Based on the fused fault feature parameters, a localization strategy combining the Weighted K-Nearest Neighbors (WKNN) algorithm and an image matching algorithm is adopted to achieve accurate fault location. Each sensing node calculates its local fault location initial value as a valid sample based on its own location information and the fused fault features. Then, the localization information is exchanged and cross-validated through distributed cooperative communication technology to correct the localization deviation and finally determine the precise location of the fault point. The localization process does not rely on centralized control of a central platform. The nodes cooperate autonomously, reducing data transmission pressure and improving the real-time performance of localization. The optimization of the WKNN algorithm, the fusion logic of the two algorithms, and the distributed cooperative verification mechanism are the core self-developed technologies.

[0039] The specific methods and processes of cross-validation: Cross-localization is based on a distributed collaborative architecture, employing a "multi-node pairing + multi-set data mutual verification" approach. It combines WKNN and image matching fusion algorithms to construct a localization network. Each sensing node acts as an independent unit, establishing temporary communication links with 3-5 neighboring sensing nodes. It synchronizes its calculated initial localization values ​​(including coordinates, confidence level, and feature matching degree) to these neighboring nodes, while simultaneously receiving and verifying data from them. The process is as follows: generating initial localization values ​​→ establishing communication links → exchanging localization data → cross-comparison to convergence range → outputting intermediate cross-localization values. The entire process involves autonomous node collaboration without the need for central platform control.

[0040] Generation of intermediate values ​​for cross-positioning: Based on the converged set of valid samples, the coordinates of all valid initial values ​​are weighted and fused using "comprehensive credibility + consistency score" as the joint weight, ultimately obtaining the cross-location median value. The consistency score is calculated using the following formula: when (Completely overlaps with all samples), score = 1 (full marks); when (The average Euclidean distance between the initial value of the i-th sample and the initial values ​​of all other samples) Exceeding the convergence threshold If the score is 0, it is marked as an invalid outlier and is directly removed.

[0041] 1. Calculate the joint weight of each valid initial value: Joint weight = Algorithm confidence Feature matching degree Consistency score 2. Calculate the weighted average of the x and y coordinates for each initial fault location value: in , The coordinates of the i-th valid initial value. The corresponding joint weights are given, and n represents the number of valid samples.

[0042] 3. The final output ( , This is the intermediate value for cross-positioning.

[0043] Deviation correction is based on the intermediate value of cross-location, and the process is as follows: Each node summarizes all valid cross-location data (i.e., the initial fault location value, including coordinates, confidence level, and feature matching degree) and the intermediate cross-location value; weights are assigned based on the confidence level and feature matching degree of the initial location value; error analysis is performed on the weighted location data through a self-developed fusion logic to remove abnormal data exceeding the preset threshold; the coordinate correction amount is calculated based on the valid data, and the location value is iteratively adjusted; multiple nodes synchronize the correction results and verify each other until the positioning deviation of each node is corrected. The accuracy is set to ≤ preset, ultimately outputting precise coordinates of the fault point to correct deviations and improve positioning reliability. The calculation formula is as follows: in These are the correction values ​​for the x and y axis coordinates. Let i be the weight of the i-th group of valid data, ( , ) represents the coordinates of the i-th group of valid cross-location data. ) represents the current iteration's positioning coordinates.

[0044] The iterative adjustment formula is: in , () represents the coordinates for the k-th iteration. The system locates the coordinates for the (k+1)th iteration and stops iterating until the deviation is less than or equal to the preset accuracy. Multiple nodes synchronously correct the results and verify each other until the positioning deviation of each node is less than or equal to the preset accuracy. Finally, the system outputs the accurate coordinates of the fault point, thereby correcting the deviation and improving the reliability of positioning.

[0045] 6. Verification and optimization of positioning results The final location result is compared with historical fault data to verify the location accuracy. The historical fault database contains both actual fault location information and inferred fault location information. The parameters for inferring the current fault location are continuously adjusted based on historical data. If the location error exceeds a preset threshold, the multi-source data fusion module and the distributed collaborative location module are invoked again to adjust algorithm parameters (such as weight allocation and matching strategy) and optimize the location result. At the same time, the current fault location data and optimized parameters are fed back to the system to achieve self-learning and self-optimization of the algorithm, improving the accuracy and efficiency of subsequent positioning. The closed-loop logic for location result feedback optimization is a self-developed technology.

[0046] This invention addresses the shortcomings of existing technologies and achieves the functional objectives of each stage through the following four core technologies, thus forming a complete fault location technology system: 1) Heterogeneous sensing node self-organizing deployment technology Technology integration: It integrates various industrial sensor technologies such as pressure sensors, temperature sensors, gas sensors, current sensors, and vibration sensors, as well as well-known networking technologies such as self-organizing communication protocols and node cluster head election technology, and optimizes and integrates them in combination with industrial scenarios.

[0047] Problem solved: It addresses the issues of poor adaptability of traditional single-sensor node deployment models in industrial scenarios, easy interruption of the sensing network, and inability to adapt to complex industrial environments (dense equipment, strong electromagnetic interference, and high dust levels).

[0048] Solution: Breaking away from the traditional single-node deployment approach, this solution adopts a hybrid deployment logic of heterogeneous nodes for industrial scenarios and a dynamic networking adaptation mechanism. First, based on the monitoring needs of different industrial scenarios such as machining, chemical processing, and power supply and distribution, corresponding sensor types are flexibly configured. Node deployment does not require pre-planned fixed locations and can be flexibly deployed according to the distribution of equipment and workshop layout within the industrial plant. Second, nodes use a well-known self-organizing communication protocol, automatically scanning surrounding nodes to complete network formation after power-on, and establishing data transmission paths through well-known node cluster head election technology. Finally, an automatic fault replacement mechanism is designed. When some nodes fail due to industrial environmental interference (corrosion, vibration, dust), the remaining nodes can automatically re-network to fill sensing blind spots, ensuring the continuity and stability of the sensing network.

[0049] Input: Target industrial scenario requirements (such as pipeline monitoring in chemical plant areas, machine tool monitoring in mechanical workshops), performance parameters of various known sensors, environmental topology information of the monitoring area (equipment distribution, spatial structure), self-organizing communication protocol parameters, and node cluster head election algorithm parameters.

[0050] Output: A heterogeneous distributed sensing network covering the target industrial area with self-healing capabilities, capable of stably collecting multi-dimensional industrial equipment and environmental data, ensuring the continuity and stability of the sensing network, and adapting to the monitoring needs of long-term continuous industrial production.

[0051] 2) Multi-source heterogeneous data fusion preprocessing technology Technology integration: Integrating well-known data preprocessing techniques such as outlier removal, noise filtering, and timestamp alignment; fault feature extraction algorithms such as Fast Fourier Transform (FFT) and wavelet transform; image processing algorithms such as Convolutional Neural Network (CNN); and well-known fusion technologies such as weighted average fusion algorithms, and combining them with industrial scenarios for optimization and integration.

[0052] Problem solved: This addresses the issues in industrial scenarios where data collected by swarm intelligence sensing nodes is redundant, noisy (e.g., machine tool vibration, workshop electromagnetic interference), and highly heterogeneous, making it impossible for a single data point to fully reflect the fault status of industrial equipment and resulting in inaccurate fault feature extraction.

[0053] Solution: An innovative three-stage preprocessing workflow of "cleaning - feature extraction - weighted fusion" is designed. Step 1, Data Cleaning: Known outlier removal and noise filtering techniques are used to process the collected raw data, removing noise from the industrial environment (such as machine tool vibration and electromagnetic interference) and outliers, and aligning the timestamps of different data types. Step 2, Feature Extraction: The known FFT algorithm is used to extract the frequency peak and frequency domain features of vibration signals, the known wavelet transform algorithm is used to extract fault features from pipeline pressure and current data, and the known CNN algorithm is used to extract equipment appearance defect features from image data. Step 3, Weight Allocation and Weighted Fusion: Combining the accuracy of industrial sensors and the data stability under workshop conditions, fusion weights are assigned to different types of fault features. A known weighted average algorithm is used to integrate multi-dimensional features to form a unified fault feature vector, achieving fault feature enhancement and redundant information compression.

[0054] Input: Industrial equipment operation data (such as machine tool vibration, motor current, pipeline pressure), industrial environmental data (such as temperature, humidity, electromagnetic interference intensity) and node metadata (node ​​ID, location, collection timestamp) collected by each heterogeneous sensing node.

[0055] Output: A unified fault feature vector with denoising, compressed redundant information, and enhanced fault characteristics, which can comprehensively and accurately reflect the fault status of industrial equipment. Compared with single data, the positioning error is reduced by about 50%, providing reliable data support for subsequent fault positioning.

[0056] 3) Swarm intelligence-driven anomaly node identification technology Technology integration: Integrating well-known particle swarm optimization (PSO) algorithm, node operation status monitoring technology (used to collect data such as node power, communication packet loss rate, etc.), and data consistency verification technology (used for abnormal node detection), and combining them with industrial scenarios for optimization and integration.

[0057] Problem solved: This addresses the issue that sensing nodes in industrial scenarios are susceptible to interference from dust, vibration, chemical corrosion, etc., which can cause malfunctions, or that malicious nodes may send false data, leading to erroneous data that interferes with fault location results and reduces the reliability of location.

[0058] Solution: An innovative multi-dimensional abnormal node detection index system is constructed, covering core indicators such as data integrity, data consistency, and node resistance to industrial interference. The fitness function of the well-known particle swarm optimization (PSO) algorithm is improved by combining the data distance between nodes with the data anomaly degree. By simulating swarm behavior, abnormal nodes (including faulty nodes caused by industrial environmental interference and nodes that maliciously send false data) are quickly identified. The identified abnormal nodes are marked and isolated. At the same time, an innovative node redundancy replacement mechanism is designed, in which adjacent healthy nodes replace their sensing functions to ensure that the data collection and positioning process is not affected.

[0059] Inputs: raw data uploaded by each sensing node, fused feature data, real-time operating status data of the node itself (such as power consumption, communication packet loss rate, anti-interference capability), and fitness function parameters of the improved PSO algorithm.

[0060] Output: A list of abnormal nodes (including node ID, abnormality type, and abnormality level), and a cleaned and reliable multi-source sensing dataset. This ensures the continuous and stable operation of the sensing network and improves the reliability and robustness of fault location in industrial scenarios.

[0061] 4) Improved distributed cooperative localization algorithm Technology integration: Integrating well-known weighted K-nearest neighbor (WKNN) algorithms, image matching algorithms (such as SIFT / SURF algorithms), distributed collaborative communication technology, and weighted average algorithms (for final accurate fault location calculation), and combining them with industrial scenarios for optimization and integration.

[0062] Problem solved: This addresses the issues of low positioning accuracy and poor real-time performance of traditional positioning algorithms in complex industrial environments (small, enclosed spaces, equipment obstructions, strong electromagnetic interference), as well as their reliance on centralized control via a central platform, which makes them unsuitable for the rapid fault handling needs of industrial equipment.

[0063] Solution: The traditional WKNN algorithm is innovatively optimized by introducing fault feature similarity as the second weight dimension, constructing a composite weight calculation method of "distance + feature similarity" to improve the accuracy of the initial positioning value; the optimized WKNN algorithm is innovatively integrated with a well-known image matching algorithm, and positioning deviations are corrected by comparing the image with standard images of industrial equipment; combined with well-known distributed collaborative communication technology, each sensing node autonomously exchanges positioning information and performs cross-validation without the need for centralized control of a central platform, eliminating abnormal positioning values, and finally obtaining the accurate fault location through a well-known weighted average algorithm.

[0064] Inputs: preprocessed fused fault feature vectors, the location coordinates of each sensing node, industrial equipment image data collected by each node, a preset standard image library of industrial equipment, improved WKNN algorithm parameters, image matching algorithm parameters, and distributed collaborative communication parameters.

[0065] Output: Precise geographic coordinates of industrial equipment fault locations (average positioning error controlled within 1.72m), and feed the positioning results back to the industrial operation and maintenance platform, greatly improving the accuracy and real-time performance of fault location, and adapting to the needs of rapid handling of industrial equipment faults.

[0066] like Figure 2 As shown, an embodiment of the present invention provides a fault location system under collective intelligent perception in an industrial scenario, characterized in that it includes a sensing node deployment module, a data acquisition and preprocessing module, a multi-source data fusion module, and a distributed collaborative positioning module; The sensing node deployment module is used to deploy a number of sensing nodes in the target industrial monitoring area to form a distributed sensing network; wherein each selected industrial device in the target industrial monitoring area corresponds to one or more sensing nodes. The data acquisition and preprocessing module is used to collect the operating status data and environmental data of the corresponding industrial equipment and upload them to the distributed collaborative platform. The multi-source data fusion module is used to extract data features from the collected data and perform weighted fusion of data features corresponding to the same industrial equipment as the fusion features of the industrial equipment. The distributed collaborative positioning module is used to calculate the initial value of local fault location based on the location information of the sensing node and the fusion characteristics of the industrial equipment it is located on, and send it to other sensing nodes of the industrial equipment for cross-verification, correct the positioning deviation, and finally determine the precise location of the fault point.

[0067] Specific application examples To further illustrate the practicality and feasibility of this invention, specific application examples are presented in conjunction with actual industrial application scenarios. All these examples are based on the technical solutions described above, fully demonstrating the application value of each core self-developed technology and meeting the actual needs of industrial equipment operation and maintenance, production line support, etc. Case Study: Fault Location in Industrial Power Supply and Distribution Equipment (Scenario: Power Supply and Distribution Workshop in a Large Chemical Plant) Application scenario: This invention is designed for power distribution workshops in large chemical plants. These workshops have dense distribution cabinets and cable lines, resulting in strong electromagnetic interference. Traditional manual inspections are difficult to quickly troubleshoot faults, and the fault location accuracy is low. Once a fault occurs, it can easily lead to the shutdown of the entire production line in the plant, causing significant economic losses. There is an urgent need for a precise and rapid fault location solution that is compatible with the heterogeneous node deployment technology and the abnormal node identification technology driven by swarm intelligence of this invention.

[0068] Application Implementation: Using the heterogeneous sensing node self-organizing deployment technology of this invention, known current, voltage, and temperature sensors are selected, and 12 sets of heterogeneous sensing nodes are deployed in the power supply and distribution workshop and surrounding cable trenches. The nodes do not need to be fixed in location and can be flexibly deployed according to the pipeline distribution. After the nodes are powered on, they automatically form a network through a self-organizing communication protocol, elect a cluster head node to establish a data transmission path, and when some nodes fail due to electromagnetic interference or dust, the remaining nodes automatically re-network to ensure the continuity of the sensing network.

[0069] Each node collects current and voltage signals from the distribution cabinet and temperature data from the cable trench in real time. Through multi-source heterogeneous data fusion preprocessing technology, the data is cleaned using known outlier removal and noise filtering techniques. Fault features are extracted using FFT and wavelet transform algorithms, and weighted fusion is performed by combining sensor accuracy with weighted fusion to form a unified fault feature vector.

[0070] The system uses swarm intelligence-driven abnormal node identification technology to collect node operating status data, construct multi-dimensional detection indicators, and use an improved PSO algorithm to calculate fitness values. It identifies abnormal nodes caused by electromagnetic interference and dust, marks and isolates them, and initiates a redundancy replacement mechanism, in which adjacent nodes increase the collection frequency and replace the functions of abnormal nodes.

[0071] Finally, an improved distributed collaborative positioning algorithm is used. Based on the fusion of fault feature vectors, the improved WKNN algorithm is used to calculate the initial positioning value. The deviation is corrected by combining the image matching algorithm. The nodes exchange positioning information and cross-validate through distributed collaborative communication technology to eliminate outliers and accurately locate fault points (such as loose cable joints or faulty components in the distribution cabinet). This improves the positioning accuracy from the traditional workshop level to the equipment location level.

[0072] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A fault location method based on crowdsourced sensing in an industrial setting, comprising the following steps: Several sensing nodes are deployed in the target industrial monitoring area to form a distributed sensing network; each selected industrial device in the target industrial monitoring area corresponds to one or more sensing nodes. Each sensing node collects real-time operational status data and environmental data of the corresponding industrial equipment and uploads them to the distributed collaborative platform. The distributed collaborative platform extracts the data features of the collected data, and performs weighted fusion of the data features corresponding to the same industrial equipment to serve as the fusion feature of that industrial equipment. Each sensing node calculates an initial local fault location value based on its own location information and the fusion characteristics of the industrial equipment it is located in, and sends it to other sensing nodes of the industrial equipment for cross-verification to correct the positioning deviation and finally determine the precise location of the fault point.

2. The method according to claim 1, characterized in that, The method for cross-validating, correcting for positioning errors, and finally determining the precise location of the fault is as follows: Each sensing node synchronizes its own calculated initial fault location value to neighboring sensing nodes and receives the initial fault location value from neighboring sensing nodes. Each sensing node performs a weighted fusion of its own calculated initial fault location value and the coordinate position in the received initial fault location value to obtain an intermediate cross-location value, which is then sent to other sensing nodes in the industrial equipment. Each sensing node iteratively calculates the coordinate correction amount based on the currently summarized initial fault location value and the intermediate value of cross-location. It then sends the information to other sensing nodes on the same industrial equipment for mutual verification until the positioning deviation of each sensing node on the same industrial equipment is less than the preset accuracy, thus finally determining the precise location of the fault point; among which, These are the correction values ​​for the x and y axis coordinates. The weights of the initial values ​​for the i-th fault location are ( , ) represents the coordinates of the intermediate value of the i-th intersection positioning, ) represents the fault location coordinates obtained in the current iteration, and n represents the number of initial fault location values.

3. The method according to claim 2, characterized in that, pass Iteratively adjust the fault location coordinates; where, , ( ) represents the coordinates of the fault location obtained in the k-th iteration. The coordinates of the fault point obtained in the (k+1)th iteration are used to locate the fault point.

4. The method according to claim 2, characterized in that, The method for obtaining the cross-location intermediate value is as follows: calculate the joint weight of each fault location initial value; and perform a weighted average of the fault point coordinates in the currently summarized fault location initial values ​​to obtain the cross-location intermediate value.

5. The method according to claim 2, characterized in that, The sensing nodes filter the aggregated initial fault location values ​​and delete those that exceed a preset threshold.

6. The method according to claim 1, characterized in that, The initial values ​​for fault location include the fault point coordinates, confidence level, and feature matching degree.

7. The method according to claim 1, characterized in that, The precise location of the finally determined fault point is compared with the corresponding historical fault data. If the positioning error exceeds the preset threshold, the relevant parameters are adjusted and the initial local fault positioning values ​​of each sensing node are reset to optimize the precise location of the finally determined fault point.

8. The method according to claim 1, characterized in that, The distributed collaborative platform employs an abnormal node detection model to monitor the uploaded data and operational status of each sensing node in real time. It identifies malicious nodes, faulty nodes, and data anomaly nodes, marks and isolates them, and prohibits them from participating in data fusion and location calculations. At the same time, it activates a node redundancy replacement mechanism, allowing adjacent healthy sensing nodes to replace their sensing functions. The platform combines the data distance between nodes and the data anomaly degree in the particle swarm optimization algorithm as the fitness function of the algorithm, resulting in an improved particle swarm optimization algorithm. The abnormal node detection model uses the improved particle swarm optimization algorithm to identify malicious nodes, faulty nodes, and data anomaly nodes.

9. The method according to claim 1, characterized in that, The sensing nodes include various types of sensors and communication modules; each sensing node adopts a heterogeneous deployment mode, and the sensing nodes are networked using a self-organizing communication protocol; the data features extracted from the collected data include vibration data features, current / voltage data features, image data features, and environmental data features.

10. A fault location system based on collective intelligent sensing in an industrial setting, characterized in that, It includes a sensing node deployment module, a data acquisition and preprocessing module, a multi-source data fusion module, and a distributed collaborative positioning module; The sensing node deployment module is used to deploy a number of sensing nodes in the target industrial monitoring area to form a distributed sensing network; wherein each selected industrial device in the target industrial monitoring area corresponds to one or more sensing nodes. The data acquisition and preprocessing module is used to collect the operating status data and environmental data of the corresponding industrial equipment and upload them to the distributed collaborative platform. The multi-source data fusion module is used to extract data features from the collected data and perform weighted fusion of data features corresponding to the same industrial equipment as the fusion features of the industrial equipment. The distributed collaborative positioning module is used to calculate the initial value of local fault location based on the location information of the sensing node and the fusion characteristics of the industrial equipment it is located on, and send it to other sensing nodes of the industrial equipment for cross-verification, correct the positioning deviation, and finally determine the precise location of the fault point.

Images