Paperless remote inspection system for special equipment

Abstract

The invention relates to the technical field of knowledge maps, in particular to a special equipment paperless remote inspection system, which comprises an equipment map construction module used for generating a single equipment state feature node according to a received special equipment traction machine current value and a speed governor action speed value, and aggregating and generating a regional equipment association knowledge graph according to the brand model and the use unit information identified in the state feature node of the single equipment. According to the method, potential systematic defects which are caused by specific batches or manufacturers and are difficult to find through isolated detection are identified and quantified, finally, the quantified regional systematic risk entropy value is directly linked with the hierarchical control strategy, the inspection interval is adjusted, and resource configuration is optimized; therefore, the safety supervision of the special equipment is converted from a fixed and passive periodic inspection mode to a dynamic and predictive active management mode based on risks, and the scientificity of the safety supervision is improved.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and in particular to a paperless remote inspection system for special equipment. Background Technology

[0002] Knowledge graphs are large-scale semantic networks that describe concepts, entities, and relationships in the objective world in a structured form. Their basic building blocks are "entity-relationship-entity" triples, where entities act as nodes in the graph and relations as edges connecting the nodes, together forming a graph-like data structure designed to transform information into knowledge that can be understood and processed by computers.

[0003] Existing paperless remote inspection technology for special equipment treats each piece of equipment as an information silo, lacking the ability to uncover potential correlations between devices and making it difficult to identify systemic or regional risk patterns. For example, when a particular batch of traction machine brakes from a specific manufacturer has a widespread design flaw, only multiple isolated brake failure alarms occurring at different locations and times can be recorded. It takes a long time and a lot of manual data comparison and statistical analysis to discover potential safety hazards. During this period, other equipment in use from the same batch continues to operate with risks. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a paperless remote inspection system for special equipment.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a paperless remote inspection system for special equipment includes: The equipment graph construction module is used to generate individual equipment status feature nodes based on the received special equipment traction machine current value and speed limiter action speed value, and aggregate and generate a regional equipment association knowledge graph based on the brand model and user unit information identified in the individual equipment status feature nodes. The dynamic cycle extrapolation module is used to aggregate the operating status feature values ​​of the target device node and its neighboring nodes based on the regional device association knowledge graph, calculate and generate the failure probability of the time window, map the failure probability of the time window to the risk cycle correspondence table, and find and match the safety margin index corresponding to the failure probability. The regional entropy calculation module is used to group all equipment in the region according to the manufacturer and batch label based on the dynamic safety margin index, generate a regional risk feature propagation matrix, calculate the probability distribution items in the regional risk feature propagation matrix, and obtain the regional systemic risk entropy value. The inspection reporting and scheduling planning module is used to select control strategies to trigger special rectification, suspension of registration or emergency response based on the regional systemic risk entropy value, generate graded risk circuit breaker instructions, adjust the inspection reporting interval of the equipment to be inspected according to the graded risk circuit breaker instructions, and reallocate inspection personnel and testing instrument resources to generate a dynamic inspection scheduling plan.

[0006] 2. The paperless remote inspection system for special equipment according to claim 1, characterized in that the system further comprises: The registration and login module is used for identity registration and system login. It supports one-click login via WeChat for unit users and binds the information of the unit using special equipment. The equipment management module is used to query, browse, and maintain information on all special equipment in use by the unit. The inspection application management module is used to realize online management of the entire inspection application process. It includes providing video tutorials for special equipment inspection, providing a catalog of special equipment fee standards, allowing the applicant to select the special equipment to be inspected and submit the inspection application materials, and allowing the accepting end to view the inspection application information and accept the application in real time. It also receives inspection application acceptance message reminders, and can view the acceptance status and payment notices in real time.

[0007] 3. The special equipment paperless remote inspection system according to claim 1, characterized in that the step of obtaining the regional equipment association knowledge graph is as follows: Based on the current value of the traction machine and the speed limiter action speed value of the special equipment, the interval index is located by comparing each item with the preset inspection and testing standard threshold range. The field is mapped according to the interval index and the status code and spliced ​​in a fixed order to generate the status feature node of the individual equipment. Based on the status feature nodes of the individual devices, the brand model and user unit information are parsed, the set of status feature nodes of individual devices in the same region is retrieved and compared according to the brand model, and the attribution relationship is checked by the consistency of user unit information to generate connection edges. Based on the connecting edges, merge the state feature nodes of the same single device that are repeated, establish an adjacency table according to the start and end points of the connecting edges, generate connected groups based on the adjacency table and attach node attributes, and generate a regional device association knowledge graph.

[0008] Preferably, the step of obtaining the failure probability of the time window is as follows: Based on the knowledge graph of the region's equipment association, the target equipment node is selected and the operating status feature values ​​of neighboring nodes are retrieved simultaneously. The traction machine current timing and the speed limiter action speed timing are extracted, aligned and completed according to a unified timestamp reference, and a sequence of operating status feature values ​​is generated. The failure probability of the time window is calculated based on the sequence of operating status feature values.

[0009] Preferably, the step of obtaining the safety margin index is as follows: Based on the failure probability within the time window, find the entry in the risk cycle table that matches the failure probability range within the time window, extract the corresponding safety margin index field, associate it with the target device node, and obtain the safety margin index.

[0010] Preferably, the steps for obtaining the regional risk feature propagation matrix are as follows: Based on the dynamic safety margin index, the equipment in the area is grouped according to the manufacturer label and the batch label. The number of dynamic safety margin indices below the threshold and the total number of equipment in the group are counted for each group. The manufacturer label and the batch label are associated with the corresponding quantity to generate a grouped statistical list. Based on the grouped statistical list, the ratio of the number of devices below the threshold under each combination of manufacturer label and batch label to the total number of risky devices in the region is calculated as the actual risk probability distribution item. At the same time, the ratio of the total number of devices to the total number of devices in the region is calculated as the reference probability distribution item, and a regional risk feature propagation matrix is ​​generated.

[0011] Preferably, the steps for obtaining the regional systemic risk entropy value are as follows: Calculate the regional systemic risk entropy value based on the regional risk characteristic propagation matrix.

[0012] Preferably, the step of obtaining the tiered risk circuit breaker instruction is as follows: Based on the regional systemic risk entropy value, the control strategies for special rectification, suspension of registration, or emergency response are determined by comparing each of the first-level, second-level, and third-level warning lines and determining the range in which the risk falls. The control strategies, applicable equipment lists, and execution start and end times are then packaged to generate graded risk circuit breaker instructions.

[0013] Preferably, the steps for obtaining the dynamic inspection scheduling plan are as follows: Based on the aforementioned risk-based circuit breaker instructions, the equipment to be inspected is classified into special rectification, suspension of registration and emergency response. For each category, a shortening, freezing or expedited mark is set for the inspection interval. At the same time, an assignment constraint list is established for inspection personnel and testing instrument resources, and an inspection interval and resource allocation plan is generated. Based on the inspection interval duration and resource allocation scheme, the equipment to be inspected is sorted according to the expiration time of the inspection interval duration. The start and end times of the execution of the graded risk circuit breaker instruction are combined to match the resources of inspection personnel and testing instruments, resolve conflicts in the same time period, output the executable task sequence, and generate a dynamic inspection scheduling plan.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, real-time operational data of special equipment, such as traction machine current and speed limiter operating speed, is transformed into structured state feature nodes. Relationships between these nodes are established based on inherent attributes such as brand, model, and user unit, constructing a knowledge graph that dynamically reflects the health status and internal connections of regional equipment. This graph is transformed into a living entity capable of real-time sensing of the operational status of a group of equipment. Based on this, the operational status characteristics of the target equipment and its neighboring nodes are aggregated to calculate the failure probability within future time windows and generate a quantified safety margin index. This achieves forward-looking and refined prediction of risks for individual equipment, further expanding the analytical dimensions from... By elevating individual cases to the regional system level, and grouping equipment by manufacturer and batch of equipment put into use, a systemic risk entropy value representing the degree of risk concentration and abnormal distribution is calculated. This approach can effectively identify and quantify potential systemic defects caused by specific batches or manufacturers that are difficult to detect through isolated testing. Ultimately, the quantified regional systemic risk entropy value is directly linked to the hierarchical control strategy, adjusting the inspection interval and optimizing resource allocation. This transforms the safety supervision of special equipment from a fixed, passive, periodic inspection model to a dynamic, predictive, and risk-based proactive management model, thereby improving the scientific nature of safety supervision. Attached Figure Description

[0015] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] Please see Figure 1 This invention provides a technical solution: a paperless remote inspection system for special equipment, comprising: The equipment graph construction module is used to generate individual equipment status feature nodes based on the received special equipment traction machine current value and speed limiter action speed value. Based on the brand, model and user unit information identified in the individual equipment status feature nodes, it aggregates and generates a regional equipment association knowledge graph. The dynamic cycle extrapolation module is used to aggregate the operating status feature values ​​of the target device node and its neighboring nodes based on the regional device association knowledge graph, calculate and generate the failure probability of the time window, map the failure probability of the time window to the risk cycle correspondence table, and find and match the safety margin index corresponding to the failure probability. The regional entropy calculation module is used to group all equipment in the region according to the manufacturer and batch label based on the dynamic safety margin index, generate a regional risk feature propagation matrix, calculate the probability distribution items in the regional risk feature propagation matrix, and obtain the regional systemic risk entropy value. The inspection reporting and scheduling planning module is used to select control strategies that trigger special rectification, suspension of registration, or emergency response based on the regional systemic risk entropy value, generate graded risk circuit breaker instructions, adjust the inspection reporting interval of the equipment to be inspected according to the graded risk circuit breaker instructions, and reallocate inspection personnel and testing instrument resources to generate a dynamic inspection scheduling plan.

[0018] The steps for obtaining the knowledge graph of regional device associations are as follows: Based on the current value of the traction machine and the speed limiter action speed value of the special equipment, the interval index is located by comparing each item with the preset inspection and testing standard threshold range. The field is mapped according to the interval index and the status code and spliced ​​in a fixed order to generate the status feature node of the individual equipment. Based on the status feature nodes of individual devices, the brand, model and user unit information are parsed, the set of status feature nodes of individual devices in the same area is retrieved and compared according to the brand and model, and the attribution relationship is checked by the consistency of user unit information to generate connection edges. Based on the connecting edges, merge the state feature nodes of the individual devices that point to the same node. Establish an adjacency table according to the start and end points of the connecting edges. Generate connected groups and attach node attributes based on the adjacency table to generate a regional device association knowledge graph.

[0019] Specifically, based on the traction machine current value and the speed limiter operating speed value of the special equipment, the equipment rated parameter table stored in the database is retrieved to extract the rated current value of the corresponding model. With rated speed value For example, setting , Based on this, a multi-level state range is constructed, with the normal current range set as... The current overload warning range is and the dangerous current range is Similarly, the normal fluctuation range of speed is set as follows: The received real-time sampling data is substituted into the above interval for comparison. If the real-time current value is 11A, it is determined that it falls into the current overload warning interval, and the corresponding interval index value is obtained, such as "Idx_C2". If the real-time speed value is 1.0m / s, the interval index value "Idx_S1" is obtained. The preset status code mapping table is read. This mapping table uses a hash structure to store the correspondence between the index and the binary feature code. For example, "Idx_C2" is mapped to "1010" and "Idx_S1" is mapped to "0001". According to the predetermined splicing protocol, the device ID field, timestamp field, current feature code and speed feature code are combined in bit order. For example, a unique feature string is generated using the format "ID (32bit) + Time (64bit) + CodeC (4bit) + CodeS (4bit)". A memory address is allocated for this string to build a data structure. This structure is marked as a single device status feature node.

[0020] Based on the status feature node of a single device, read the reserved metadata fields in the node's data structure, and extract the brand model string and the unit code converted from the device registration code. For example, extract the brand field as "Brand_X_Series_5" and the unit code as "Unit_ID_7782". Set the search radius to 5 kilometers, traverse all other nodes in the database whose geographical coordinates are within this radius, establish a candidate node set, and start a double loop comparison program. For each candidate node in the set, first compare whether the ASCII codes of the brand model string are completely consistent. If they are consistent, mark it as "same type device". Then compare the unit code. If the codes are the same, mark it as "device with the same ownership". Calculate the association weight based on the comparison results and set the basic weight. If the brand and model are the same, the weight increases by 0.5; if the unit of use is the same, the weight increases by 0.3. Only when the total weight value after calculation is greater than or equal to the preset minimum association threshold of 0.5 is the association relationship confirmed. An edge object is instantiated, the current node is set as the source node, the candidate node is set as the target node, the calculated weight value is written into the attribute field of the edge object, and the edge object is stored in the temporary edge list. This establishes the logical connection relationship between the nodes and generates the connecting edge.

[0021] Based on the connecting edges, the temporary edge list is cleaned. The source node ID and target node ID of all edges are checked. If two edges have the same start and end points, and the timestamp difference is less than a preset jitter tolerance (e.g., 100 milliseconds), they are considered duplicate or redundant data. A merging operation is performed, the average weight of the two edges is calculated as the new weight, and redundant edge objects are discarded. Then, an empty adjacency list structure is initialized. The processed edge list is traversed, and the target node index of each edge is appended to the adjacency list of the source node, completing the topology construction of the graph data. Based on this adjacency list, a breadth-first search algorithm (BFS) is applied. Connectivity analysis is performed, starting from the first node in the list and traversing to find all reachable nodes to form a connected subgraph, which is marked as "Group_01". The number of nodes and average weight in this group are recorded. If the number of nodes exceeds the preset lower limit of the group size (e.g., 5), the group is retained; otherwise, it is marked as discrete noise and removed. Finally, all retained connected groups are traversed, and the group ID is written as a meta-attribute into the data structure of each node. At the same time, the centrality index of the group is calculated. The graph data object containing the complete topology, node attributes and group information is serialized and stored to generate a regional device association knowledge graph.

[0022] The steps to obtain the probability of failure within a time window are as follows: Based on the regional equipment association knowledge graph, the target equipment node is selected and the operating status feature values ​​of neighboring nodes are retrieved simultaneously. The traction machine current timing and the speed limiter action speed timing are extracted, aligned and completed according to a unified timestamp reference, and a sequence of operating status feature values ​​is generated. Based on the sequence of operational status characteristic values, the failure probability within the time window is calculated using the following formula: ; in, The probability of failure within the time window. This represents the number of target device nodes and their neighboring nodes in the regional device association knowledge graph. To predict the length of the time window, For the first Each device in time The instantaneous value of the traction machine current. For the first Each device in time The instantaneous value of the speed limiter's operating speed. For the first The historical average traction machine current value of each device under normal operating conditions. For the first The standard deviation of the traction machine current value of each device For the first The historical average speed limiter activation speed value of each device under normal operating conditions. For the first The standard deviation of the operating speed value of the speed limiter of each device The risk contribution weighting coefficient for current characteristics The risk contribution weighting coefficient for speed characteristics.

[0023] Specifically, based on the regional device association knowledge graph, the unique identifier ID of the target device node, such as "Elevator_A_001", is read using the graph query language. The adjacency edge table of this node is then traversed to obtain a list of neighbor node IDs with a weight value greater than 0.5. Using the current time as the baseline, a 24-hour sampling window is established by backtracking. A batch query command is sent to the time-series database to retrieve the raw traction machine current reading packets and speed limiter action speed reading packets of the target device and all neighbor devices within this window. The obtained raw data packets are decompressed and cleaned to remove null values ​​or garbled characters (e.g., NaN or negative values) caused by sensor malfunctions. A unified timeline is then established, using millisecond-level UNIX timestamps as its scale. The step size is set to the reciprocal of the sampling frequency (e.g., if the sampling rate is 50Hz, the step size is 20ms). The raw data stream of each device is traversed and mapped onto a unified time axis. For data gaps caused by network latency, if the gap duration is less than 500ms, cubic spline interpolation is used to fit the missing value based on three adjacent valid data points. If the gap duration is greater than 500ms, it is filled with the average baseline value of the device under historical normal operating conditions. After aligning and completing the timestamps of all devices, the current and speed values ​​at each time point are flattened and concatenated according to device order to construct a multidimensional matrix structure. Rows represent time steps, and columns represent the feature dimensions of different devices. Finally, this multidimensional matrix is ​​serialized and stored to generate a sequence of operating state feature values. The formula for calculating the failure probability within a time window incorporates the concepts of an exponential decay model and the Mahalanobis distance. It integrates the energy of current and speed deviations from multiple devices within a region in a spatiotemporal manner to construct a probabilistic model reflecting the regional collaborative risk density. Compared to single-machine threshold alarms, this formula utilizes… By weighted summing of the synchronization state deviations of the associated devices, the system can accurately detect the precursors of common-mode faults caused by power grid fluctuations or regional environmental factors. The introduction of the square term amplifies the risk signal that deviates significantly from the normal value, while the integral and average operations smooth out the influence of instantaneous random noise, thereby quantifying the probability of system failure within the entire time window.

[0024] This parameter represents the total number of target device nodes and their neighboring nodes in the regional device association knowledge graph. The steps to obtain this parameter are: locating the target node in the graph database; using a breadth-first search (BFS) algorithm to find neighboring nodes within a one-hop range whose connection weights meet the requirements; and counting the total number of nodes including the target node. For example, in a certain calculation, if the target elevator is associated with two adjacent elevators under the same transformer, then... The value obtained is 3.

[0025] This parameter represents the prediction time window length in seconds. The steps to obtain this parameter are as follows: It is set based on the frequency characteristics of the equipment's operating cycle. Through spectral analysis of historical fault cases, it was found that abnormal current fluctuations typically last 5 to 15 seconds. To cover the complete abnormal fluctuation cycle while also considering real-time calculations, 10 seconds is selected as a standard time slice. Therefore, it is set... .

[0026] Indicates the first Each device in time The instantaneous value of the traction machine current, in amperes (A), is obtained by real-time acquisition using a Hall current sensor installed at the output of the frequency converter in the elevator control cabinet. The sampling frequency is set to 50Hz, i.e., once every 0.02 seconds. The time window is extracted during the calculation process. The discrete sampling point values ​​within the range, for example, the first device in The current reading at 1 second was collected as follows A.

[0027] Indicates the first Each device in time The instantaneous value of the speed limiter's operating speed, in meters per second (m / s), is obtained by monitoring the rotational speed of the speed limiter's disc using a rotary encoder, calculating the linear velocity using a pulse conversion coefficient, and also acquiring the data at a frequency of 50Hz. For example, the first device... The speed readings were collected in seconds. m / s.

[0028] Indicates the first The historical average traction machine current value of each device under normal operating conditions, in amperes (A), is obtained by selecting the operating data of the device over the past 30 days when there were no fault reports and the load was 50% of the rated load. The average current is calculated using the averaging method. For example, by statistically analyzing the historical data of the first device, the average current under normal load can be obtained. A.

[0029] Indicates the first The standard deviation of the traction machine current value for each device, in amperes (A), is obtained by calculating the dispersion of the current value relative to the average value based on the selected historical data segment of normal operation. For example, the standard deviation of the current for the first device is calculated. A.

[0030] Indicates the first The historical average speed limiter activation speed value of each device under normal operating conditions, in meters per second (m / s). The steps to obtain this parameter are as follows: select historical data from the device's steady-speed operation phase for statistical analysis, excluding acceleration and deceleration phases, and calculate the average value. For example, if the rated operating speed of the first device remains stable, the statistical analysis yields... m / s.

[0031] Indicates the first The standard deviation of the speed limiter's operating speed value for each device, in meters per second (m / s), is obtained by calculating the root mean square deviation of speed fluctuations during normal steady-speed operation, reflecting the stability of the device's speed control. For example, the standard deviation of the speed of the first device can be calculated. m / s.

[0032] The risk contribution weight coefficient representing the current characteristics ranges from 0 to 1. The steps to obtain this parameter are as follows: Based on the Analytic Hierarchy Process (AHP), a fault influencing factor judgment matrix is ​​constructed. Then, based on the proportion of shutdowns caused by current anomalies (e.g., 60%) and the proportion of shutdowns caused by speed anomalies (e.g., 40%) in historical fault records, normalization is performed, and the result is calculated. .

[0033] The risk contribution weight coefficient representing speed characteristics ranges from 0 to 1. The steps for obtaining this parameter are as follows: Based on the aforementioned analysis of fault influencing factors, since speed limiter malfunctions typically directly trigger the safety brake, resulting in serious consequences but occurring less frequently, the weight is determined by combining expert ratings and historical data statistics, and this determination satisfies... Calculations yielded .

[0034] Calculations based on parameters: The area has Devices, time window seconds, using discretized approximate integral ( Seconds, that is, only counting The state at this point in time.

[0035] The parameters are set as follows: , , , .

[0036] Reference parameters for device 1: , , , .

[0037] exist Instantaneous values ​​collected at each moment: (Current too high) (Slightly fast).

[0038] The first step is to calculate the squared standardized deviation of the current characteristic: ; ; The second step is to calculate the squared standardized deviation of the velocity characteristic: ; ; The third step is to calculate the weighted risk value (the value within the integral kernel): ; Step 4: Integrate execution time and average device performance (here) ): ; Step 5: Calculate the failure probability : ; Look up mathematical constants

[0039] ; The results indicate that within the current time window, due to a significant deviation of two standard deviations in the current value and some fluctuations in speed, the calculated failure probability within the time window is as high as 93.92%. This means that the device is highly likely to be in an abnormal or imminent failure risk range under the current conditions, and the system should respond immediately to this high probability value.

[0040] The steps to obtain the safety margin index are as follows: Based on the failure probability within the time window, find the entry in the risk cycle table that matches the failure probability range within the time window, extract the corresponding safety margin index field, associate it with the target device node, and obtain the safety margin index.

[0041] Specifically, based on the failure probability within a time window, a pre-built risk cycle correspondence table stored in an in-memory database is invoked. The table's construction process is as follows: Historical operating data and fault logs of special equipment within the region over the past five years are collected. The failure probability value calculated for each piece of equipment one hour before a failure occurs is statistically analyzed to form a failure precursor probability distribution set. The risk grading threshold is determined using quantile statistics. For example, the failure probabilities corresponding to historical failure samples are sorted from smallest to largest. The value of 0.3 at the 25th quantile is set as the upper limit for low risk, and the value of 0.7 at the 75th quantile is set as the lower limit for high risk, thus dividing the probability intervals into [0, 0.3), [0.3, 0.7), and [0.7, 0.7]. [1.0] There are three levels, corresponding to "Safe Period", "Observation Period" and "Risk Period" respectively, and a corresponding safety margin index is set for each interval. For example, "Safe Period" corresponds to an index range of 90-100, "Observation Period" corresponds to 60-90, and "Risk Period" corresponds to 0-60. The specific mapping relationship is stored in key-value pair form. When the calculated failure probability of the time window is received (e.g., 0.9392), the interval definitions in the risk cycle correspondence table are traversed. It is determined that 0.9392 falls into the interval [0.7, 1.0]. The basic safety margin index corresponding to this interval (e.g., 40) is read, and linear interpolation is performed to fine-tune it according to the relative position of the probability value within the interval. The calculation formula is as follows: The specific index value (approximately 12.16) is calculated. Finally, this value is encapsulated into a status update package and sent to the map management module. The module locates the attribute field of the target device node, updates the original security status value to the calculated 12.16, completes the association marking, and obtains the security margin index.

[0042] The steps for obtaining the regional risk characteristic propagation matrix are as follows: Based on the dynamic safety margin index, the equipment in the area is grouped according to the manufacturer label and the batch label. The number of dynamic safety margin indices below the threshold and the total number of equipment in the group are counted for each group. The manufacturer label and the batch label are associated with the corresponding quantity to generate a grouped statistical list. Based on the grouped statistical list, the ratio of the number of devices below the threshold under each manufacturer's label and batch label combination to the total number of risky devices in the region is calculated as the actual risk probability distribution item. At the same time, the ratio of the total number of devices to the total number of devices in the region is calculated as the reference probability distribution item, and a regional risk characteristic propagation matrix is ​​generated.

[0043] Specifically, based on the dynamic safety margin index, a numerical standard for determining whether equipment poses a potential risk is first established. This standard is based on in-depth analysis of big data from special equipment safety monitoring in the region over the past five years. By analyzing the correspondence between over 5,000 equipment failure reports and the recorded safety margin index at that time, a scatter plot of the failure rate and index value was plotted. It was found that when the safety margin index is below 60, the probability of unplanned equipment downtime or component damage increases exponentially (e.g., from an average of 0.5% to over 5%). Therefore, based on this statistical law, the risk threshold for the safety margin index is set to 60 as a screening criterion. Subsequently, a multi-level hash mapping table structure is initialized to store the grouped statistical results and traverse the region. For each node in the domain device association knowledge graph, read the "Manufacturer" tag (e.g., "Manufacturer_A") and "Batch" tag (e.g., "Batch_2022_Q3") stored in the node's attributes. Simultaneously, extract the dynamic safety margin index calculated in real time for that node. Compare the extracted index value with the set risk threshold of 60. If the index value is less than 60, the device is determined to be a "low-margin risk device". In the hash table, under the key value of the corresponding manufacturer and batch, increment the "risk device count" field by one. Regardless of the index value, increment the "total number of devices in the group" field under the corresponding key value. After completing the traversal and statistics of all nodes, export and serialize the data in the hash table to generate a grouped statistical list.

[0044] Based on the grouped statistics list, a global summary is performed on all grouped data recorded in the list. Each entry in the list is traversed, and the "risk equipment count" of each group is accumulated to obtain the total number of equipment in a risky state within the region (denoted as ). At the same time, the total number of devices within each group is summed to obtain the total number of all registered devices in the area (denoted as ). Based on these two global baseline values, a normalized ratio is calculated for each grouped item. For the item in the list... The manufacturer and the first The combination of each batch is used to read the number of risky devices recorded in the statistics. The proportion of risk in the total risk equipment is calculated by division, which is the actual risk proportion. This ratio reflects the degree of risk concentration in a specific batch, while also considering the total number of devices reading that group. Calculate its proportion in the total number of devices in the region, i.e., the reference distribution proportion. This ratio reflects the natural market share of this batch of equipment, calculated as follows. and These two feature values ​​are paired and filled into a two-dimensional data structure with the manufacturer as the row index and the batch as the column index. For batch combinations that do not exist, a minimum value is filled in to avoid computational singularity. Finally, a structured data body containing the topological information of the risk distribution of equipment in the whole region is constructed, and a regional risk feature propagation matrix is ​​generated.

[0045] The steps for obtaining the regional systemic risk entropy value are as follows: Based on the regional risk characteristic propagation matrix, the regional systemic risk entropy value is calculated using the following formula: ; in, This represents the entropy value of regional systemic risk. For the first The manufacturer's label and the first The actual risk probability distribution under each batch label combination. , For the first The manufacturer's label and the first The number of dynamic safety margin indices below the threshold for each batch of labels used. For the first The manufacturer's label and the first Reference probability distribution under each batch label combination in use , For the first The manufacturer's label and the first The total number of devices under each batch label combination in operation. This represents the total number of labels from the manufacturers. For the first The number of batch labels used under each manufacturer's label. For the first The number of batches of labels used under each manufacturer's label. For the first The manufacturer's label and the first The number of dynamic safety margin indices below the threshold for each batch of labels used. For the first The manufacturer's label and the first The total number of devices under each batch label combination.

[0046] Specifically, the formula for calculating the entropy value of regional systemic risk introduces the concept of relative entropy (Kullback-Leibler divergence) from information theory to measure the probability of the actually observed risk distribution. Reference probability based on the natural distribution of the number of devices The difference between them is calculated using this formula. This is not the traditional disorder entropy, but rather the entropy of systematic risk deviation. If the risk is randomly and uniformly distributed across all devices, then... and When the results are consistent and close to 0, and when the results are significantly positive, it indicates that the risk is highly concentrated in certain manufacturers or batches, revealing potential systemic quality defects or batch-related hazards.

[0047] This represents the total number of manufacturer labels. The steps to obtain this parameter are: scanning the row dimensions of the regional risk feature propagation matrix and counting the number of different manufacturer identifiers contained within it. For example, if the region mainly involves two suppliers, "Manufacturer A" and "Manufacturer B", then... .

[0048] Indicates the first The number of batch labels used under a manufacturer's label is obtained by the following steps: For the first batch label in the matrix... The system performs a search and counts the number of batches with valid device records under that manufacturer. For example, manufacturer A has 2 major batches and manufacturer B has 1 major batch.

[0049] Indicates the first The manufacturer's label and the first The total number of devices under each batch label combination is obtained by reading it directly from the previously generated group statistics list. For example, manufacturer A has 100 devices in batch 1 (A-1) and 100 devices in batch 2 (A-2), and manufacturer B has 200 devices in batch 1 (B-1). The total number of devices in the region is... .

[0050] Indicates the first The manufacturer's label and the first The number of dynamic safety margin indices below a threshold under a given batch label combination is determined by retrieving the device count below the threshold (e.g., 60) from the grouped statistics list. For example: 30 low-margin devices were found in batch A-1 (potentially batch defects), 5 in batch A-2, and 15 in batch B-1. The total number of risky devices in the region is [not specified]. .

[0051] The actual risk probability distribution term is represented by the following formula: The steps to obtain this parameter are as follows: calculate using the above statistical values, for example: , , .

[0052] The reference probability distribution term is calculated using the following formula: The steps to obtain this parameter are as follows: calculate using the total number of devices, for example: , , .

[0053] Calculations based on parameters: Based on the example data set above, the relative entropy component is calculated for each grouping item: First item (Manufacturer A, Batch 1): ; Look up the natural logarithm table.

[0054] ; Item 2 (Manufacturer A, Batch 2): ; ; ; Item 3 (Manufacturer B, Batch 1): ; ; ; Summation calculation : ; The results indicate that the regional systemic risk entropy value is 0.2805. This positive value (significantly greater than 0) quantifies the imbalance in risk distribution. The first term contributes the largest positive value (0.5253), indicating that the risk proportion (60%) of the equipment in the first batch of Manufacturer A is much higher than its equipment quantity proportion (25%), showing a strong systemic batch risk characteristic. Other batches perform better (negative contribution indicates that the risk is below average). Therefore, the calculation result of 0.2805 directly indicates that a special rectification strategy needs to be launched for the first batch of Manufacturer A.

[0055] The steps for obtaining a tiered risk circuit breaker instruction are as follows: Based on the regional systemic risk entropy value, the control strategies for special rectification, suspension of registration, or emergency response are determined by comparing each of the first-level, second-level, and third-level warning lines and determining the range in which the risk falls. The control strategies, applicable equipment lists, and execution start and end times are then packaged to generate graded risk circuit breaker instructions.

[0056] Specifically, based on the regional systemic risk entropy value, historical monthly risk entropy value records from the past thirty-six months are retrieved from the regional special equipment safety supervision database to construct a time series sample set containing over one thousand historical data points. This sample set is then cleaned to remove extreme noise caused by data transmission errors, and statistical methods are used to calculate the average value of the sample set. (For example, set to 0.12) and standard deviation (For example, set to 0.04), based on the statistical laws of the normal distribution, set three-level dividing lines, and... Adding one standard deviation, or 0.16, as the first-level warning line, will... Adding twice the standard deviation, i.e., 0.20, as the secondary warning line, will... Adding three standard deviations, i.e., 0.24, as the third-level warning line, and receiving the currently calculated regional systemic risk entropy value of 0.2805, this value is compared with the above three warning lines. If the value is greater than the third-level warning line of 0.24, the current region is determined to be in an "extremely high risk" state. According to the preset risk response strategy matrix, this state corresponds to the highest priority "suspend registration" and "emergency response" composite strategy, that is, suspend the use registration of relevant batches of equipment and complete a comprehensive investigation within a specified time. Then, extract the combination key values ​​with the highest entropy contribution in the risk feature propagation matrix (e.g., "manufacturer A - batch 1"), lock the applicable equipment list, and set the start and end times for strategy execution, where the execution duration is... The calculation is based on the degree of overflow of the risk entropy value relative to the three-level warning line, with a basic investigation period set at 7 days. The calculation formula is as follows: Substituting the numerical values, we obtain The system sets the number of days to 9, rounded up, with the start time set to the current time and the end time set to 9 days after the current time. It then encapsulates the determined control strategy type, the list of unique identifiers for restricted devices, the information of the responsible manufacturer, and the specific time span into a standardized instruction data packet to generate a graded risk circuit breaker instruction.

[0057] The steps to obtain the dynamic inspection scheduling plan are as follows: Based on the risk-based circuit breaker instructions, the equipment to be inspected is classified into special rectification, suspension of registration and emergency response. For each category, a shortening, freezing or expedited mark is set for the inspection interval. At the same time, an assignment constraint list is established for inspection personnel and testing instrument resources, and an inspection interval and resource allocation plan is generated. Based on the inspection interval and resource allocation plan, the equipment to be inspected is sorted according to the expiration time of the inspection interval. The start and end times of the execution of the graded risk circuit breaker command are combined to match the resources of inspection personnel and testing instruments, resolve conflicts in the same time period, output the executable task sequence, and generate a dynamic inspection scheduling plan.

[0058] Specifically, based on the tiered risk circuit breaker instruction, the applicable equipment list and risk level in the instruction data packet are analyzed. All registered special equipment within the region is divided into three control levels. Equipment explicitly listed as "Manufacturer A - Batch 1" is placed in the "Core Control Area." Equipment from the same manufacturer but different batches as equipment in the Core Control Area is placed in the "Associated Impact Area." The remaining equipment is placed in the "Regular Operation Area." For equipment in the "Core Control Area," its inspection status is directly set to "Frozen," meaning all regular inspection processes are suspended and the registration status is locked. For equipment in the "Associated Impact Area," the inspection interval is shortened, and its original inspection cycle is retrieved. (For example, 30 days), the compression coefficient is calculated based on the regional systemic risk entropy value. (Right now ), calculate the new interval duration The calculated result is 21.585 days, rounded down to 21 days. For equipment in the "routine operation zone," the original inspection cycle remains unchanged. The qualification database of inspection personnel and the ledger of testing instruments are loaded simultaneously. Resource allocation constraints are established for different control levels. For tasks in the "core control zone," the constraints are set as "personnel qualification level greater than or equal to senior engineer" and "instrument accuracy level better than 0.5." For tasks in the "related influence zone," the constraints are set as "personnel qualification level greater than or equal to intermediate engineer." The resource database is scanned, and the certificate validity period of each inspector and the last calibration date of each instrument are verified. A list of compliant resources that meet the above hard conditions is selected. The selected personnel ID set and instrument ID set are associated and bound with the corresponding equipment tasks to generate the inspection interval and resource allocation plan.

[0059] Based on the inspection interval and resource allocation scheme, a priority queue of tasks to be processed is constructed. The equipment to be inspected is initially sorted according to risk control levels, with the priority order being "core control area" > "related impact area" > "routine operation area". Within the same level, a secondary sort is performed based on the order of the inspection interval expiration time, generating a linear sequence of tasks to be executed. A two-dimensional scheduling matrix is ​​initialized, where the rows correspond to the selected compliant inspection personnel and instrument combinations, and the columns correspond to the available time slices within the next 9-day execution window (with 2 hours as the minimum scheduling unit). The sorted task sequence is traversed, and the execution start and end times specified in the tiered risk circuit breaker instructions are read. For the highest priority task in the sequence, the row of inspection personnel that meets its resource constraints is searched in the scheduling matrix, and the earliest consecutive idle time slice within the execution window is found, for example, locking the time slot from 9:00 AM to 11:00 AM on day 1. If multiple inspection personnel meet the conditions, the personnel with the fewest currently assigned tasks are selected to achieve load balancing. If resource contention occurs within the same time slice, i.e., multiple high-risk tasks simultaneously request the same scarce instrument, the urgency score of the task is calculated. (in (The remaining hours until the deadline) prioritize tasks with higher scores, and postpone tasks with lower scores to the next available time slice. After confirming the time slice, mark it as "occupied" in the matrix and deduct the corresponding instrument inventory. Finally, convert the completed scheduling matrix into structured scheduling data containing task ID, assigned personnel, instruments carried, inspection location and specific execution time, and generate a dynamic inspection scheduling plan table.

Claims

1. A paperless remote reporting system for special equipment, characterized in that, The system comprises: An equipment graph construction module, configured to generate a single-equipment state feature node according to the received special equipment tractor current value and speed limiter action speed value, and aggregate a regional equipment associated knowledge graph according to the brand model and use unit information identified in the single-equipment state feature node; A dynamic cycle deduction module, configured to aggregate the running state feature values of a target equipment node and neighbor nodes according to the regional equipment associated knowledge graph, calculate a time window failure probability, map the time window failure probability to a risk cycle corresponding table, and find and match a safety margin index corresponding to the failure probability; A regional entropy value calculation module, configured to group all equipment in a region according to manufacturers and use batch labels according to the dynamic safety margin index, statistically generate a regional risk feature propagation matrix, calculate the probability distribution items in the regional risk feature propagation matrix, and obtain a regional systematic risk entropy value; An inspection scheduling planning module, configured to select a control strategy of triggering a special rectification, suspending registration or emergency response according to the regional systematic risk entropy value, generate a hierarchical risk fuse instruction, adjust the inspection interval length of an equipment to be inspected according to the hierarchical risk fuse instruction, and re-allocate inspection personnel and detection instrument resources to generate a dynamic inspection scheduling plan table.

2. The paperless remote reporting system for special equipment according to claim 1, wherein, The system further comprises: A registration and login module, configured to support one-key login through a unit WeChat user and bind special equipment use unit information; An equipment management module, configured to query, browse and maintain all in-use special equipment information of the unit; An inspection management module, configured to realize online management of the whole inspection process, including providing a special equipment inspection video tutorial, providing a special equipment charging standard directory, selecting a special equipment to be inspected on an inspection terminal and submitting inspection data, and allowing a receiving terminal to view inspection information in real time and accept the inspection, receive inspection acceptance message reminders, view acceptance status in real time and receive payment notices.

3. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition steps of the regional equipment associated knowledge graph are: According to the special equipment tractor current value and the speed limiter action speed value, the interval index is located by item-by-item comparison with the preset inspection detection standard threshold interval, the field mapping is performed according to the interval index and the state code, and the single-equipment state feature node is generated by splicing in a fixed order; According to the single-equipment state feature node, the brand model and the use unit information are analyzed, the single-equipment state feature node set in the same region is compared according to the brand model, and the attribution relationship is checked for consistency of the use unit information to generate a connection edge; According to the connection edge, the single-equipment state feature nodes pointing to the same are merged, the adjacency relationship table is established according to the start point and the end point of the connection edge, the connected groups are generated according to the adjacency relationship table and the node attributes are added, and the regional equipment associated knowledge graph is generated.

4. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition steps of the time window failure probability are: According to the regional device association knowledge graph, a target device node is selected and the running state characteristic values of neighbor nodes are retrieved, traction machine current time series and speed limiter action speed time series are extracted, aligned and completed according to a unified timestamp reference, and a running state characteristic value sequence is generated; According to the running state characteristic value sequence, a time window failure probability is calculated.

5. The paperless remote reporting system for special equipment according to claim 1, wherein, The safety margin index acquisition step is: According to the time window failure probability, an entry matching the time window failure probability interval in the risk period corresponding table is searched, a corresponding safety margin index field is extracted, and a safety margin index associated with the target device node is obtained.

6. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition step of the regional risk feature propagation matrix is: According to the dynamic safety margin index, the devices in the region are grouped according to the manufacturer label and the commissioning batch label, the number of dynamic safety margin indexes below the threshold and the total number of devices in each group are counted, the manufacturer label and the commissioning batch label are associated with the corresponding number, and a grouping statistics list is generated; According to the grouping statistics list, the ratio of the number of devices below the threshold to the total number of risk devices in the region under each combination of manufacturer label and commissioning batch label is calculated as the actual risk probability distribution item, and the ratio of the total number of devices to the total number of devices in the region is calculated as the reference probability distribution item, and a regional risk feature propagation matrix is generated.

7. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition step of the regional systematic risk entropy value is: According to the regional risk feature propagation matrix, the regional systematic risk entropy value is calculated.

8. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition step of the hierarchical risk fuse instruction is: According to the regional systematic risk entropy value, the first, second and third warning lines are compared one by one, the control strategy of special rectification, suspension of registration or emergency response is determined according to the interval, the control strategy, the applicable device list and the execution start and end time are packaged, and the hierarchical risk fuse instruction is generated.

9. The paperless remote reporting system for special equipment according to claim 1, wherein, The acquisition step of the dynamic inspection scheduling plan table is: According to the hierarchical risk fuse instruction, the devices to be inspected are divided into special rectification, suspension of registration and emergency response, the shortening, freezing or urgent label of the inspection interval length is set for each type, the inspection personnel and detection instrument resources are simultaneously established for assignment constraints, and the inspection interval length and resource allocation scheme are generated; According to the inspection interval length and resource allocation scheme, the devices to be inspected are sorted according to the expiration of the inspection interval length, the inspection personnel and detection instrument resources are matched according to the execution start and end time in the hierarchical risk fuse instruction, the conflicts in the same period are eliminated, and the executable task sequence is output, and the dynamic inspection scheduling plan table is generated.

Images