A multi-sensor fusion-based intelligent diagnosis and network connection method for support hangers

Abstract

This invention discloses an intelligent diagnostic and network-connected method for support and hanger status based on multi-sensor fusion. The method synchronously collects support and hanger status data using pressure, displacement, and angle sensors. After Kalman filtering for noise reduction, time-frequency domain feature extraction, and dynamic weighted fusion, the data is input into a pre-trained fault diagnosis model to identify the support and hanger's working status. The terminal generates local alarms and uploads abnormal data. The operation and maintenance management platform executes a graded repair strategy according to the fault level. If repair fails, a maintenance work order is generated and pushed to maintenance personnel. This invention achieves real-time monitoring, accurate diagnosis, and closed-loop operation and maintenance of support and hanger status, and can be widely applied to support and hanger operation and maintenance scenarios such as building electromechanical systems and power lines.

Description

Technical Field

[0001] This invention relates to a method for intelligent diagnosis and networking of support and hanger status based on multi-sensor fusion. Background Technology

[0002] Pipeline supports are critical components of pipeline and equipment support systems, used to bear the weight of pipelines and equipment, suppress operational vibrations, and limit displacement. Their structural stability directly affects the safe operation of the pipeline system. Loosening, deformation, or detachment of supports can lead to pipeline stress imbalance, connection point damage, and even secondary risks such as media leakage or equipment overturning. Currently, monitoring the condition of supports mainly relies on periodic manual inspections. In large buildings, industrial plants, and rail transit systems, supports are widely distributed and numerous, often installed in concealed locations such as high altitudes, ceilings, or tunnels. Manual inspections typically occur every 1 to 3 months, making real-time monitoring of condition changes impossible. Furthermore, inspection quality is constrained by personnel experience, responsibility, and site conditions, resulting in insufficient ability to identify early loosening or minor displacements, leading to a high risk of missed inspections. In addition, the long response chain from fault discovery to reporting, approval, and work assignment at each level results in significant delays in fault handling, easily missing the optimal intervention opportunity.

[0003] To replace manual inspections, existing technologies have developed intelligent support and hanger terminals that integrate sensors and communication functions. These terminals can automatically collect and remotely upload support and hanger status data, and are integrated with an operation and maintenance management platform for data aggregation. However, existing intelligent support and hanger terminals and their operation and maintenance methods still have significant limitations: Firstly, terminals often use a single type of sensor (such as only vibration sensors or only stress sensors), while the actual failure modes of supports and hangers are usually accompanied by coupled changes in multi-dimensional parameters such as pressure, displacement, and angle. A single sensor is difficult to fully reflect the true state of the structure, and existing terminals lack fusion diagnostic algorithms for multi-source data, which can easily lead to misjudgment or omission. Secondly, most existing IoT monitoring systems adopt a two-level forwarding architecture of "terminal-central platform-operation and maintenance platform". Data needs to be relayed through a third-party central platform, which not only increases deployment costs and transmission delays, but also makes it impossible to implement in remote factories or temporary sites without central platform coverage. Third, the existing operation and maintenance platform is limited to data display and threshold alarms, lacking a hierarchical automatic handling mechanism for anomalies. Fault repair and work order dispatch still rely on manual operation, and the operation and maintenance efficiency has not been substantially improved.

[0004] Therefore, existing support and hanger monitoring technologies cannot yet meet the actual engineering needs in terms of multi-sensor fusion diagnosis, deployment flexibility, and operation and maintenance automation. Summary of the Invention

[0005] The present invention provides a method for intelligent diagnosis and networking of support and hanger status based on multi-sensor fusion to solve the problems existing in the prior art.

[0006] The technical solutions adopted in this invention are as follows: A method for intelligent diagnosis and networking of support and hanger status based on multi-sensor fusion, applied to intelligent support and hanger terminals and operation and maintenance management platforms, includes the following steps: S1: The intelligent support terminal collects multi-dimensional data on the pressure load, displacement offset, and tilt angle of the support through pressure sensors, displacement sensors, and angle sensors. S2: The intelligent support terminal performs Kalman filtering noise reduction on the collected raw data, extracts the time domain features and frequency domain features of the data, and constructs a multi-source feature vector; S3: The intelligent support terminal performs dynamic weighted fusion calculation on the multi-source feature vector based on the real-time working status of the pressure sensor, displacement sensor and angle sensor; S4: The intelligent support terminal inputs the fused feature vector into the pre-trained support fault diagnosis model to identify the normal state, loose state, deformed state or detachment state of the support. S5: The intelligent support terminal compares the status recognition result with the preset threshold. If the threshold is exceeded, a local alarm is triggered and the abnormal data is uploaded to the operation and maintenance management platform. S6: The operation and maintenance management platform receives abnormal data and executes a graded automatic repair strategy according to the degree to which the abnormal data exceeds the threshold; if the fault is not eliminated after automatic repair, a maintenance work order is automatically generated and pushed to the maintenance personnel.

[0007] Furthermore, in S2, the time-domain features include mean, variance, peak value, and kurtosis, and the frequency-domain features include spectral centroid, root mean square, and power spectral density. The time-domain features and frequency-domain features are features extracted separately for each sensor.

[0008] Furthermore, in S3, the weight coefficient range of the dynamic weighted fusion calculation is: pressure sensor weight 0.35-0.45, displacement sensor weight 0.25-0.35, and angle sensor weight 0.25-0.35. The weight coefficients are normalized so that the sum of the pressure sensor weight, displacement sensor weight, and angle sensor weight is 1.

[0009] Furthermore, in S4, the pre-trained support and hanger fault diagnosis model is constructed using the random forest algorithm and trained with more than 10,000 sets of multi-condition samples.

[0010] Furthermore, in S5, the preset thresholds include: displacement offset exceeding 20% ​​of the support's displacement tolerance range, pressure load exceeding 15% of the support's pressure tolerance range, and tilt angle exceeding 20% ​​of the support's tilt tolerance range.

[0011] Furthermore, in S6, the fault level classification includes: Level 1 fault: Displacement deviation reaches 15% but less than 20% of the tolerable range, or pressure load reaches 12% but less than 15% of the tolerable range, or tilt angle reaches 15% but less than 20% of the tolerable range; Level 2 fault: Displacement deviation reaches 20% but less than 30% of the tolerable range, or pressure load reaches 15% but less than 20% of the tolerable range, or tilt angle reaches 20% but less than 30% of the tolerable range; Level 3 fault: The displacement deviation reaches or exceeds 30% of the acceptable range, or the pressure load reaches or exceeds 20% of the acceptable range, or the tilt angle reaches or exceeds 30% of the acceptable range, or it is identified as a detachment state.

[0012] Furthermore, the automatic repair strategy corresponding to the fault level includes: For Level 1 faults, the operation and maintenance management platform records fault information and pushes alerts to management personnel; For level 2 faults, the operation and maintenance management platform sends a restart command to the intelligent support and hanger terminal, which then performs a remote restart. In response to a Level 3 fault, the operation and maintenance management platform issues an emergency restart command to the intelligent support and hanger terminal and triggers a local audible and visual alarm on the intelligent support and hanger terminal.

[0013] Furthermore, in S6, the maintenance work order includes the fault level, fault type, terminal location information, historical monitoring data, and fault occurrence time; the maintenance work order can be pushed via SMS, APP, and WeChat notification.

[0014] Furthermore, in S1, the sampling frequency of the synchronous acquisition is 500ms / time, that is, the pressure sensor, displacement sensor and angle sensor synchronously acquire data at a frequency of 500ms / time.

[0015] Furthermore, in S5, if communication is interrupted during the process of uploading abnormal data to the operation and maintenance management platform, the intelligent support terminal will cache the abnormal data in the local storage and automatically resume uploading it to the operation and maintenance management platform after communication is restored.

[0016] The present invention has the following beneficial effects: (1) By synchronously collecting data from three types of sensors—pressure, displacement, and angle—and performing Kalman filtering for noise reduction, multi-feature extraction, and dynamic weighted fusion at the edge, it can comprehensively cross-validate multi-dimensional data, reduce the deviation caused by environmental interference or measurement blind spots of a single sensor, and make the fault identification results closer to the actual structural state of the support.

[0017] (2) The method supports direct IP communication between the intelligent support terminal and the operation and maintenance management platform without relying on a third-party central platform for forwarding. It is applicable to various scenarios with and without a central platform, reducing the requirements of the system deployment on the network infrastructure.

[0018] (3) The terminal completes data preprocessing and preliminary diagnosis at the edge, and only uploads abnormal results to avoid the delay caused by the remote transmission of full data; the operation and maintenance management platform executes a graded automatic repair strategy according to the degree of abnormality, and generates and pushes maintenance work orders in real time when automatic repair is ineffective, compressing the response process of traditional manual inspection and hierarchical reporting to the minute level.

[0019] (4) The automatic repair strategy reduces the need for manual intervention for minor and moderate faults, and the automatic allocation and tracking of intelligent work orders reduces scheduling and communication costs, enabling maintenance resources to be concentrated on high-risk fault nodes and optimizing human resource allocation.

[0020] (5) If communication is interrupted during abnormal data upload, the terminal can cache the data to the local storage and automatically resume the transmission after communication is restored, reducing the risk of loss of key abnormal information due to network interruption. Attached Figure Description

[0021] Figure 1 This is a flowchart of the present invention.

[0022] Figure 2 This is a flowchart of a multi-sensor fusion diagnostic algorithm.

[0023] Figure 3 A flowchart for fault level classification, automatic repair, and maintenance work order generation. Detailed Implementation

[0024] The invention will now be further described with reference to the accompanying drawings.

[0025] like Figure 1 As shown, this invention provides an intelligent diagnostic and networking method for the status of supports and hangers based on multi-sensor fusion. It is applied to intelligent support and hanger terminals and operation and maintenance management platforms, and can be widely used in various support and hanger operation and maintenance scenarios such as building electromechanical pipelines, power lines, chemical equipment, and rail transit tunnels.

[0026] The intelligent support and hanger terminal integrates data acquisition, edge computing, communication transmission, and local alarm functions. The operation and maintenance management platform has data reception, fault analysis, automatic repair, work order generation, and multi-channel push functions. This invention can eliminate the need to rely on a third-party central platform for data transfer. The intelligent support and hanger terminal can directly establish a TCP / IP communication connection with the operation and maintenance management platform through static IP configuration. It supports two modes: manual configuration of static IP and automatic IP acquisition via DHCP. It is suitable for conventional scenarios covered by a central platform as well as special scenarios such as remote factories and temporary construction sites without central platform coverage, which greatly reduces the requirements of system deployment on network infrastructure. The following is a detailed description of this invention.

[0027] After power-on initialization, the intelligent support terminal enters a stable operating state. The terminal synchronously collects core status data of the support using its built-in pressure sensor, displacement sensor, and angle sensor. Specifically, the pressure sensor collects data on the pressure load borne by the support, the displacement sensor collects data on the displacement offset, and the angle sensor collects data on the tilt angle. These three types of sensors are precisely matched to the force direction of the support and perform synchronous data acquisition at a fixed sampling frequency of 500ms / time, ensuring the real-time nature, synchronization, and accuracy of the collected data. The raw pressure load, displacement offset, and tilt angle data are transmitted in real-time to the edge computing unit of the intelligent support terminal, providing a foundation for subsequent data processing.

[0028] Taking the DN100 spring support commonly used in the field of building electromechanical systems as an example, the rated load capacity of this support is 10kN, the rated displacement stroke is 30mm, and the rated tilt safety angle is 15°. The three types of sensors are respectively matched with the rated parameter setting range of the support to ensure that the collected data is fully adapted to the actual operating conditions of the support.

[0029] After receiving the raw data collected by the three types of sensors, the edge computing unit of the intelligent support terminal first uses the Kalman filter algorithm to denoise the raw data. Through iterative calculation of prediction and update, it filters out various interference signals such as environmental vibration, on-site electromagnetic interference, and sensor noise, restores the true data of the support status, and avoids noise interference affecting the accuracy of subsequent diagnosis.

[0030] After data denoising, the edge computing unit extracts the corresponding time-domain and frequency-domain features from the data collected by the pressure sensor, displacement sensor, and angle sensor. The time-domain features include mean, variance, peak value, and kurtosis, while the frequency-domain features include spectral centroid, root mean square, and power spectral density. Based on the multi-dimensional features of a single sensor, a multi-source feature vector covering the three types of data (pressure, displacement, and angle) is constructed, providing a standardized data carrier for subsequent data fusion and fault diagnosis.

[0031] The edge computing unit of the intelligent support terminal performs dynamic weighted fusion calculation on the constructed multi-source feature vector based on the real-time operating status of pressure sensors, displacement sensors, and angle sensors. Combining the operational stability, environmental adaptability, and data reliability of the sensors, the weight coefficients are dynamically adjusted. The weight coefficient of the pressure sensor ranges from 0.35 to 0.45, the weight coefficient of the displacement sensor ranges from 0.25 to 0.35, and the weight coefficient of the angle sensor ranges from 0.25 to 0.35. All weight coefficients are normalized to ensure that the sum of the weight coefficients of the three types of sensors is 1. Through dynamic weighted fusion calculation, the data deviation and measurement blind spots of a single sensor are eliminated, and the effective information of multi-dimensional data is integrated, which greatly improves the accuracy of subsequent fault status identification.

[0032] After completing the dynamic weighted fusion of multi-source feature vectors, the intelligent support terminal inputs the fused feature vectors into the pre-trained support fault diagnosis model. This fault diagnosis model is built using the random forest algorithm and trained with more than 10,000 sets of multi-condition samples covering four states: normal, loose, deformed, and detached. The training samples cover support operation data with different load levels, different installation environments, and different fault degrees. Five-fold cross-validation is used to optimize the model accuracy. The model relies on edge computing units to complete fault identification locally without uploading the full amount of original data. It can quickly identify the current working state of the support, specifically including the normal state, loose state, deformed state, and detached state. The identification process is efficient and accurate, and can promptly capture early anomalies and serious faults of the support.

[0033] After the intelligent support terminal completes the fault status identification, it compares the identification result with the preset threshold. The preset threshold is set according to the design load and operation safety standards of the support. Specifically, it is 20% more than the displacement range that the support can withstand, 15% more than the pressure load that the support can withstand, and 20% more than the tilt angle that the support can withstand. If the parameter corresponding to the identification result exceeds the above preset threshold, it is determined that the support is in an abnormal operating state. The intelligent support terminal immediately triggers a local alarm and uploads the abnormal data, fault type, status identification result and other information to the operation and maintenance management platform.

[0034] If network fluctuations or communication interruptions occur during the abnormal data upload process, the intelligent support terminal will automatically cache the abnormal data to be uploaded to the local Flash memory. After the communication link is restored to normal, it will automatically and completely re-upload the cached abnormal data to the operation and maintenance management platform, preventing the loss of critical abnormal data.

[0035] After receiving abnormal data uploaded by the intelligent support terminal, the operation and maintenance management platform classifies the fault into three standardized levels based on the degree to which the abnormal data exceeds a preset threshold: Level 1, Level 2, and Level 3 faults. A Level 1 fault is defined as a displacement deviation that is 15% to less than 20% of the tolerable range, a pressure load that is 12% to less than 15% of the tolerable range, or a tilt angle that is 15% to less than 20% of the tolerable range. Level 2 faults are defined as displacement deviation reaching 20% ​​but less than 30% of the tolerable range, pressure load reaching 15% but less than 20% of the tolerable range, or tilt angle reaching 20% ​​but less than 30% of the tolerable range. A Level 3 fault occurs when the displacement deviation reaches or exceeds 30% of the acceptable range, or the pressure load reaches or exceeds 20% of the acceptable range, or the tilt angle reaches or exceeds 30% of the acceptable range, or the fault is identified as a detachment condition.

[0036] After the fault level classification is completed, the operation and maintenance management platform automatically executes the graded automatic repair strategy that matches the fault level.

[0037] For Level 1 faults, the operation and maintenance management platform only records the fault information and pushes the warning information to the management personnel, without issuing control commands to the intelligent support and hanger terminal; For level 2 faults, the operation and maintenance management platform sends a restart command to the corresponding intelligent support terminal. After receiving the command, the terminal performs a remote restart operation to attempt to restore normal data acquisition and operation status. In response to a Level 3 fault, the operation and maintenance management platform issues an emergency restart command to the corresponding intelligent support terminal and triggers the terminal's local audible and visual alarm to promptly remind on-site personnel to pay attention to potential safety hazards.

[0038] If the fault status of the support and hanger is still not eliminated after implementing the above-mentioned graded automatic repair strategy, the operation and maintenance management platform will automatically generate a standardized maintenance work order. The maintenance work order includes core information such as fault level, fault type, terminal location information, historical monitoring data and fault occurrence time. After the work order is generated, the operation and maintenance management platform will push the maintenance work order to the maintenance personnel through three methods: SMS push, APP push and WeChat notification, to ensure that the maintenance personnel can quickly obtain fault information and carry out on-site handling work.

[0039] Throughout the entire implementation process, the intelligent support terminal completes data acquisition, filtering and noise reduction, feature extraction, weighted fusion, and preliminary fault diagnosis at the edge. It only uploads abnormal data and key diagnostic results, effectively reducing data transmission volume and network bandwidth usage, and minimizing transmission latency. Simultaneously, the intelligent support terminal and the operation and maintenance management platform can directly establish a communication connection without the need for a third-party central platform. This adapts to various conventional scenarios and special scenarios without central platform coverage, enabling real-time monitoring, intelligent diagnosis, automatic repair, and intelligent operation and maintenance of supports, comprehensively improving the efficiency, accuracy, and safety of support operation and maintenance.

[0040] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent diagnosis and networking of support and hanger status based on multi-sensor fusion, applied to intelligent support and hanger terminals and operation and maintenance management platforms, characterized in that: Includes the following steps: S1: The intelligent support terminal collects multi-dimensional data on the pressure load, displacement offset, and tilt angle of the support through pressure sensors, displacement sensors, and angle sensors. S2: The intelligent support terminal performs Kalman filtering noise reduction on the collected raw data, extracts the time domain features and frequency domain features of the data, and constructs a multi-source feature vector; S3: The intelligent support terminal performs dynamic weighted fusion calculation on the multi-source feature vector based on the real-time working status of the pressure sensor, displacement sensor and angle sensor; S4: The intelligent support terminal inputs the fused feature vector into the pre-trained support fault diagnosis model to identify the normal state, loose state, deformed state or detachment state of the support. S5: The intelligent support terminal compares the status recognition result with the preset threshold. If the threshold is exceeded, a local alarm is triggered and the abnormal data is uploaded to the operation and maintenance management platform. S6: The operation and maintenance management platform receives abnormal data and executes a graded automatic repair strategy according to the degree to which the abnormal data exceeds the threshold; If the fault is not eliminated after automatic repair, a maintenance work order will be automatically generated and pushed to the maintenance personnel.

2. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S2, the time-domain features include mean, variance, peak value, and kurtosis, and the frequency-domain features include spectral centroid, root mean square, and power spectral density. The time-domain features and frequency-domain features are features extracted separately for each sensor.

3. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S3, the weight coefficient range of the dynamic weighted fusion calculation is: pressure sensor weight 0.35-0.45, displacement sensor weight 0.25-0.35, and angle sensor weight 0.25-0.

35. The weight coefficients are normalized so that the sum of the pressure sensor weight, displacement sensor weight, and angle sensor weight is 1.

4. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S4, the pre-trained support and hanger fault diagnosis model is constructed using the random forest algorithm and trained with more than 10,000 sets of multi-condition samples.

5. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S5, the preset thresholds include: displacement offset exceeding 20% ​​of the support's displacement tolerance range, pressure load exceeding 15% of the support's pressure tolerance range, and tilt angle exceeding 20% ​​of the support's tilt tolerance range.

6. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S6, the fault level classification includes: Level 1 fault: Displacement deviation reaches 15% but less than 20% of the tolerable range, or pressure load reaches 12% but less than 15% of the tolerable range, or tilt angle reaches 15% but less than 20% of the tolerable range; Level 2 fault: Displacement deviation reaches 20% but less than 30% of the tolerable range, or pressure load reaches 15% but less than 20% of the tolerable range, or tilt angle reaches 20% but less than 30% of the tolerable range; Level 3 fault: The displacement deviation reaches or exceeds 30% of the acceptable range, or the pressure load reaches or exceeds 20% of the acceptable range, or the tilt angle reaches or exceeds 30% of the acceptable range, or it is identified as a detachment state.

7. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 6, characterized in that: The automatic repair strategies corresponding to the fault levels include: For Level 1 faults, the operation and maintenance management platform records fault information and pushes alerts to management personnel; For level 2 faults, the operation and maintenance management platform sends a restart command to the intelligent support and hanger terminal, which then performs a remote restart. In response to a Level 3 fault, the operation and maintenance management platform issues an emergency restart command to the intelligent support and hanger terminal and triggers a local audible and visual alarm on the intelligent support and hanger terminal.

8. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1 or 7, characterized in that: In S6, the maintenance work order includes the fault level, fault type, terminal location information, historical monitoring data, and fault occurrence time; the maintenance work order is pushed out via SMS, APP, and WeChat notification.

9. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S1, the sampling frequency of the synchronous acquisition is 500ms / time, that is, the pressure sensor, displacement sensor and angle sensor synchronously acquire data at a frequency of 500ms / time.

10. The intelligent diagnosis and networking method for support and hanger status based on multi-sensor fusion as described in claim 1, characterized in that: In S5, if communication is interrupted during the process of uploading abnormal data to the operation and maintenance management platform, the intelligent support terminal will cache the abnormal data in the local storage and automatically resume uploading it to the operation and maintenance management platform after communication is restored.

Images