A dynamic performance evaluation and navigation system for safety production management

Abstract

The application relates to the technical field of safety management, and particularly discloses a safety production management dynamic duty performance evaluation and navigation system, which collects multi-source space-time behavior data and work environment data of duty performance personnel, fuses to generate a fusion matching degree, constructs a dynamic duty performance benchmark grid, forms a duty performance task queue containing a forced duty performance pulse signal, calculates a duty performance space guidance link according to the duty performance task queue and a real-time space coordinate chain, generates a space-time navigation signaling package carrying a risk level and a standard action guide, pushes the signaling package to a terminal, and according to superposition comparison of an actual behavior track and the guidance link, outputs a duty performance correction signal and a duty performance score report; the application realizes process monitoring, real-time space guidance and closed-loop feedback of safety duty performance of the duty performance personnel, and improves objectivity and timeliness of duty performance management.

Description

Technical Field

[0001] This invention relates to the field of safety management technology, specifically to a dynamic performance evaluation and navigation system for safety production management. Background Technology

[0002] In enterprise safety production management, the safety performance of personnel is a crucial link in ensuring on-site operational safety. Enterprises primarily evaluate the safety performance of personnel (such as deputy general managers, chief engineers, and safety directors) based on quarterly or annual record reviews, on-site spot checks, and manually filled-out performance records. Specifically, enterprises develop performance checklists based on safety production responsibility standards, requiring personnel to enter specific safety-critical areas (such as control rooms, water treatment stations, coal yards, and booster stations) at prescribed frequencies for inspection, monitoring, or participation in safety activities. Safety management personnel then compile attendance and work records for scoring.

[0003] In safety production management, how to match the static safety performance list of personnel with the dynamic temporary maintenance instructions and emergency response status in real time, and transform it into a quantifiable performance task queue and spatial guidance link, so as to overcome the limitations of traditional back-end account evaluation in capturing deviations in the performance process, the inability of temporary tasks to automatically reach the site, and the difficulty of verifying "whether they have truly entered the critical area and stayed for a sufficient period of time" using a single identity recognition method. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic performance evaluation and navigation system for safety production management to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: A dynamic performance evaluation and navigation system for safety production management includes: The data acquisition module is used to collect multi-source spatiotemporal behavioral data and work environment data of the personnel performing their duties. The multi-source spatiotemporal behavioral data includes facial recognition access records, radio frequency identification positioning signals and access control sequence data. The work environment data includes planned maintenance work tickets, temporary operation tickets and environmental risk warning signals extracted from the production management system. The data fusion module integrates multi-source spatiotemporal behavioral data with operational environment data into a dynamic data package for job performance. Based on the dynamic data package, it extracts the dwell time and arrival frequency of personnel in each safety-critical area, and generates a fusion matching degree for each personnel's job based on the job performance benchmark library, covering the on-site area, job performance frequency, and standard actions. The task queue generation module constructs a dynamic performance benchmark grid for each employee based on the fusion matching degree and the real-time acquired temporary maintenance instructions or emergency response status. It also marks non-compliance items and excessive task items in the dynamic performance benchmark grid, forming a performance task queue containing mandatory performance pulse signals. The navigation signaling generation module calculates the performance space guidance link corresponding to each unfulfilled task based on the performance task queue and the real-time spatial coordinate chain of the personnel performing the tasks, and generates a spatiotemporal navigation signaling package carrying the regional risk level and standard action guidance based on the performance space guidance link. The comparison and feedback module pushes the spatiotemporal navigation signaling packet to the terminal carried by the personnel performing their duties. Based on the superposition and comparison of the actual behavior trajectory with the performance space guidance link and standard action guidance in the spatiotemporal navigation signaling packet, it generates a performance correction signal and performance score report for the next collection cycle.

[0006] As a further aspect of the present invention: the output process of the fusion matching degree is as follows: The facial recognition data at each time point is recorded and aligned with the radio frequency identification positioning signal by timestamp, and a spatiotemporal coordinate string carrying directional identifiers is synthesized based on the entry and exit direction marks in the access control time sequence data. The spatiotemporal coordinate string is compared with the predefined electronic fence of the safety critical area. The difference between the start and end time of each collision is extracted as the dwell time, and the number of collisions within a unit period is counted as the arrival frequency. The duration of stay and the frequency of arrival are multiplied by the standard weight of the corresponding region in the job performance benchmark database, the product results are accumulated and normalized, and then the fusion matching degree is output.

[0007] As a further aspect of the present invention: the synthesized spatiotemporal coordinate string carrying direction identifier specifically includes: Extract the trigger time, personnel identification, and entry / exit direction markers of each access control event from the access control access sequence data, and arrange them in ascending order of time to form an access control event stream; The facial recognition system matches the recorded facial identifiers with the personnel identifiers in the access control event stream one by one. The trigger time of the successfully matched access control event is used as the anchor point, and the radio frequency identification positioning signal sequence within a preset time window before and after the anchor point is extracted. For each location point in the RFID location signal sequence, calculate the time difference between its timestamp and the trigger time of the most recent access control event. Based on the sign of the time difference and the entry / exit direction mark of the corresponding access control event, assign a direction identifier to the location point and output the spatiotemporal coordinate string carrying the direction identifier in chronological order.

[0008] As a further aspect of the present invention: the formation of the task queue containing the mandatory performance pulse signal specifically includes: A blank grid is constructed with the job identifier of each employee as the vertical axis and the safety critical area identifier as the horizontal axis. The fusion matching degree is filled into the corresponding cell as the benchmark performance value. The area identifier and urgency level carried in the temporary maintenance order or emergency response status are analyzed. The urgency level is converted into a pulse intensity value according to a preset mapping relationship. The pulse intensity value is superimposed on the baseline performance value of the corresponding area cell in the blank grid to obtain a dynamic performance baseline grid. Traverse each cell in the dynamic performance benchmark grid, compare the superimposed current value with the standard threshold in the job performance benchmark library, mark cells below the lower threshold as substandard items, and mark cells above the upper threshold as excessive task items; Sort all non-compliant items by cell value from low to high, and sort the excess task items by cell value from high to low. Extract the top three non-compliant items and the top two excess task items after sorting, attach the highest level pulse identifier to each, and output the task queue containing the mandatory performance pulse signal in chronological order.

[0009] As a further aspect of the present invention: the conversion of the urgency level into a pulse intensity value according to a preset mapping relationship specifically includes: Extract the area code segment and urgency code segment from the preset field location from the temporary maintenance order or emergency response status, match the area code segment with the pre-stored safety critical area identifier library, and output the area identifier; Based on the urgency coding segment, look up the preset three-level pulse mapping table and output the basic pulse intensity value corresponding to the urgency level; Obtain the current duty load value corresponding to the region identifier in the dynamic duty load benchmark grid, multiply the base pulse intensity value by the normalized complement of the current duty load value, and output the final pulse intensity value.

[0010] As a further aspect of the present invention: the generation of the spatiotemporal navigation signaling packet carrying the regional risk level and standard action guidelines specifically includes: Extract the target area identifier corresponding to each unfulfilled task from the task queue, obtain the coordinate point with the latest timestamp in the real-time spatial coordinate chain of the personnel performing the task as the starting coordinate, and map the target area identifier to the ending coordinate; Using Manhattan distance and regional connectivity constraints, the polyline path between the starting and ending coordinates along the preset accessible grid boundary is calculated, and the polyline path is used as the guiding link for the performance space. Based on the target area identifier, the pre-stored risk level database is retrieved, the corresponding risk level label is extracted, and the corresponding standard action guidance statement is extracted from the action knowledge base according to the standard action code of the unfulfilled duties. The sequence of broken line nodes carried by the duty performance space guidance link is packaged with risk level labels and standard action guidance statements in timestamp order, and the pulse intensity parsed from the mandatory duty performance pulse signal is added as a signaling priority identifier to generate spatiotemporal navigation signaling.

[0011] As a further aspect of the present invention: the process of outputting the polyline segment path is as follows: Using the grid cell containing the starting coordinates as the current cell, extract the set of passable grid boundaries in the four neighboring directions of the current cell; The Manhattan distance from the current cell to the termination coordinate is decomposed into horizontal and vertical differences. Prioritize moving one cell along the direction with the larger absolute value of the difference to the accessible grid boundary, and update the current coordinates. Repeat the direction selection action until the current coordinates coincide with the termination coordinates, record the sequence of boundary turning points each time the grid cell is crossed, and connect the turning points in sequence to output the polyline path.

[0012] As a further aspect of the present invention: the comparison and superposition of the actual behavioral trajectory with the performance space guidance link and standard action guidance in the spatiotemporal navigation signaling packet specifically includes: The actual behavior trajectory is discretized into a sequence of trajectory points according to the timestamp, and the broken line node sequence of the performance space guidance link and the expected time window of the standard action guidance are extracted from the spatiotemporal navigation signaling packet. Calculate the spatial projection distance from each trajectory point in the trajectory point sequence to the nearest polyline segment, record the deviation points whose projection distance exceeds the preset tolerance threshold and their timestamps, and accumulate the number of deviation points; Align the sequence of deviations with the expected time window of the standard action guide on the timeline, determine whether there are deviations within the expected time window, and mark the standard action as not executed if there are deviations. Based on the number of deviations and the number of standard actions not performed, a performance correction signal is generated to reduce the data collection frequency in the next collection cycle, and a performance score report is generated to deduct the corresponding score.

[0013] The beneficial effects of this invention are: (1) This invention collects multi-source spatiotemporal behavioral data, such as facial recognition access records, radio frequency identification positioning signals, and access control time sequence data, and combines them with planned maintenance work tickets, temporary operation tickets, and environmental risk early warning signals to transform the on-site performance behavior of personnel into quantifiable dwell time, arrival frequency, and fusion matching degree. Based on this, a dynamic performance benchmark grid and performance task queue are constructed. Compared with the traditional evaluation method that relies on quarterly or annual ledger review, this invention realizes in-process monitoring and automatic scoring of the safe performance process of personnel, reduces the deviation caused by manual statistics and subjective judgment, and makes the performance records traceable and verifiable.

[0014] (2) This invention generates a spatiotemporal navigation signaling packet carrying a broken line node sequence, risk level label, and standard action guidance statement based on the task queue, and pushes it to the terminal carried by the personnel performing the duties. At the same time, by superimposing and comparing the actual behavior trajectory with the spatial guidance link of the duty performance, deviation points are detected and it is determined whether the standard action has not been executed. Based on this, a duty performance correction signal and a duty performance score report are generated to correct subsequent collection cycles. This mechanism provides real-time spatial location guidance and action reminders for the personnel performing the duties, reducing the possibility of missing the duty performance due to temporary maintenance or emergency response. Through deviation feedback and score deduction, a closed loop is formed, which improves the responsiveness and binding force of duty performance management. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the output process of the fusion matching degree in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 As shown, this invention is a dynamic performance evaluation and navigation system for safety production management, comprising: The data acquisition module is used to collect multi-source spatiotemporal behavioral data and work environment data of the personnel performing their duties. The multi-source spatiotemporal behavioral data includes facial recognition access records, radio frequency identification positioning signals and access control sequence data. The work environment data includes planned maintenance work tickets, temporary operation tickets and environmental risk warning signals extracted from the production management system. The data fusion module integrates multi-source spatiotemporal behavioral data with operational environment data into a dynamic data package for job performance. Based on the dynamic data package, it extracts the dwell time and arrival frequency of personnel in each safety-critical area, and generates a fusion matching degree for each personnel's job based on the job performance benchmark library, covering the on-site area, job performance frequency, and standard actions. The task queue generation module constructs a dynamic performance benchmark grid for each employee based on the fusion matching degree and the real-time acquired temporary maintenance instructions or emergency response status. It also marks non-compliance items and excessive task items in the dynamic performance benchmark grid, forming a performance task queue containing mandatory performance pulse signals. The navigation signaling generation module calculates the performance space guidance link corresponding to each unfulfilled task based on the performance task queue and the real-time spatial coordinate chain of the personnel performing the tasks, and generates a spatiotemporal navigation signaling package carrying the regional risk level and standard action guidance based on the performance space guidance link. The comparison and feedback module pushes the spatiotemporal navigation signaling packet to the terminal carried by the personnel performing their duties. Based on the superposition and comparison of the actual behavior trajectory with the performance space guidance link and standard action guidance in the spatiotemporal navigation signaling packet, it generates a performance correction signal and performance score report for the next collection cycle.

[0019] The data acquisition module is used to collect multi-source spatiotemporal behavioral data and work environment data of personnel performing their duties. The multi-source spatiotemporal behavioral data includes facial recognition access records, RFID positioning signals, and access control sequence data. The work environment data includes planned maintenance work orders, temporary operation orders, and environmental risk warning signals extracted from the production management system. Specifically, it includes: The data acquisition process relies on the industrial security facilities and production management information system already deployed on-site. The "facial recognition access record" in the multi-source spatiotemporal behavioral data is acquired through network high-definition cameras installed at various entrances and exits, key passages in production areas, and access control points in the central control room. Facial recognition algorithms are embedded in the front end of the cameras; when personnel pass by, their facial images are captured and compared with a pre-stored facial database. If a match is successful, a access record is generated containing the personnel's name, job identifier, access time, and camera location number. The "RFID positioning signal" is collected through passive RFID tags installed inside the safety helmets worn by personnel, in conjunction with RFID readers deployed on the tops of safety-critical areas (including the water treatment station, coal yard, turbine platform, and booster station). The readers scan the tags within their coverage area every 2 seconds, generating a positioning signal containing the tag's unique code, the reader's number, and the receiving timestamp. "Access control access sequence data" is collected through card swiping or facial recognition access control gates at each entrance and exit. The gate controller records the trigger time, personnel identification and entry / exit direction markings (the entry direction is marked as "entry" and the exit direction is marked as "departure") when each access event occurs.

[0020] The "Planned Maintenance Work Orders and Temporary Operation Orders" in the work environment data are extracted from the production management system's database. The production management system automatically records the order number, work area, work content, planned start time, planned end time, and issuer information when the work order or operation order is generated. This invention synchronizes these fields daily via a data interface. "Environmental risk warning signals" are collected from sensors deployed throughout the plant area, such as gas detectors, anemometers, and dust concentration monitors. Each sensor has a set threshold; when the monitored value exceeds the threshold, a risk warning signal is generated. This signal includes the sensor's location identifier, the risk type (e.g., excessive methane, excessive wind speed), and the trigger time. All of the above data is stored in a local time-series database using a unified timestamp format.

[0021] Please see Figure 2 As shown, in the data fusion module, multi-source spatiotemporal behavioral data and operational environment data are fused into a dynamic data package for job performance. Based on the dynamic data package, the dwell time and arrival frequency of personnel in each safety-critical area are extracted. Then, based on the job performance benchmark library, a fusion matching degree covering the on-site area, job performance frequency, and standard actions is generated for each personnel's job, specifically including: The access control event sequence data is extracted by taking the trigger time, personnel identifier, and entry / exit direction marker for each access control event and arranging them in ascending order of time to form an access control event stream. In practice, access control event sequence data from the past 24 hours is read from the local time-series database. Each record contains the trigger time (accurate to the second), personnel identifier (corresponding to a unique number in the personnel's face database), and entry / exit direction marker (valued as "enter" or "departure"). These records are arranged into a sequence, called the access control event stream, according to their trigger times from earliest to latest.

[0022] Facial recognition matches recorded facial identifiers with personnel identifiers in the access control event stream one by one. Using the trigger time of a successfully matched access control event as an anchor point, the RFID positioning signal sequence within a preset time window before and after this anchor point is extracted. The preset time window length is set to 120 seconds, i.e., 120 seconds before and 120 seconds after the anchor point. Specifically, during matching, the most recently successfully recognized facial identifier from the facial recognition record is selected. Access control events with the same personnel identifier and trigger times within 5 seconds of the facial recognition time are searched in the access control event stream. The trigger time of this access control event is used as the anchor point. Then, all positioning points whose timestamps fall within a 120-second range before and after the anchor point are selected from the RFID positioning signals and arranged in ascending order of timestamps to form a positioning signal sequence.

[0023] For each location point in the RFID positioning signal sequence, the time difference between its timestamp and the trigger time of the most recent access control event is calculated. Based on the sign of this time difference and the entry / exit direction marker of the corresponding access control event, a direction identifier is assigned to the location point, and a spatiotemporal coordinate string carrying the direction identifier is output in chronological order. Specifically, if the difference between the location point's timestamp and the access control event trigger time is positive, the location point is determined to have occurred after the access control event. In this case, if the entry / exit direction marker of the access control event is "entry," the direction identifier of the location point is assigned the value "moving inward." If the time difference is negative, the location point is determined to have occurred before the access control event. In this case, if the entry / exit direction marker is "departure," the direction identifier is assigned the value "moving outward." If the direction identifier cannot be uniquely determined according to the above rules (e.g., the time difference is zero), it is assigned the value "boundary point." Each spatiotemporal coordinate string contains the three-dimensional coordinates of the location point (represented by the horizontal and vertical coordinates of the company's floor plan and the floor number), the timestamp, and the direction identifier.

[0024] The spatiotemporal coordinate string is compared with a predefined electronic fence for safety-critical areas. The difference between the start and end times of each collision is extracted as the dwell time, and the number of collisions within a unit period is counted as the arrival frequency. The electronic fence for safety-critical areas is pre-drawn on a digital map, with each area defined by a set of closed polygon vertex coordinates. During collision comparison, each location point in the spatiotemporal coordinate string is checked to see if it falls inside any of the electronic fence polygons. When multiple consecutive location points fall within the same area, the timestamp of the first location point is taken as the entry time, and the timestamp of the last location point as the exit time; the difference between the two (in minutes) is the dwell time. The unit period is set to one calendar day, and the number of collisions (i.e., entry times) occurring in the same area within that period is counted as the arrival frequency.

[0025] The dwell time and arrival frequency are each multiplied by the standard weight of the corresponding area in the job performance benchmark database. The products are then summed, normalized, and the resulting fusion matching degree is output. The job performance benchmark database is a pre-established table based on the company's safety production responsibility standards. For each job (e.g., deputy general manager, chief engineer) and each safety-critical area (e.g., water treatment plant, coal yard, control room), standard weights for dwell time and arrival frequency are set. For example, the deputy general manager's dwell time weight in the control room is 0.6, and the arrival frequency weight is 0.4. The measured dwell time (in minutes) is multiplied by the dwell time weight, and the measured arrival frequency (in times) is multiplied by the arrival frequency weight. The two products are then added together to obtain the original accumulated value.

[0026] Normalization is performed using a maximum-minimum normalization method. The theoretical maximum and minimum values ​​of the original accumulated value for each position in each region are pre-determined based on historical data; for example, the theoretical maximum is 100 points and the theoretical minimum is 0 points. The difference between the current original accumulated value and the minimum is calculated, then divided by the difference between the maximum and minimum values. The quotient is multiplied by 100 and rounded to the nearest integer, yielding a fusion matching degree ranging from 0 to 100. A higher fusion matching degree indicates a higher degree of conformity between the employee's performance in that region and the job baseline requirements. Finally, the fusion matching degree for each employee in each safety-critical region is output as key-value pairs.

[0027] In the task queue generation module, a dynamic performance benchmark grid is constructed for each employee based on the fusion matching degree and the real-time acquired temporary maintenance instructions or emergency response status. Unmet requirements and excessive tasks are marked in the dynamic performance benchmark grid, forming a performance task queue containing mandatory performance pulse signals, specifically including: A blank grid is constructed using each employee's job title as the vertical axis and the safety-critical area identifier as the horizontal axis. The fusion matching degree is then filled into the corresponding cell as the baseline performance value. In practice, a list of job titles (including positions such as Deputy General Manager, Chief Engineer, and Safety Director) and a list of safety-critical area identifiers (including areas such as water treatment station, coal yard, central control room, booster station, and steam turbine platform) are pre-established in the enterprise's safety management database. A blank grid is generated with the number of rows equal to the number of job titles and the number of columns equal to the number of areas, using the job title as the row index along the vertical axis and the area identifier as the column index along the horizontal axis. Each cell is initially empty. Then, the fusion matching degree (integers ranging from 0 to 100) for each employee in each safety-critical area, output from the previous steps, is iterated through. Based on the employee's job title and the corresponding area identifier, a unique cell in the grid is located, and the fusion matching degree value is filled into that cell as the baseline performance value for that job title in that area. If a job title has no performance requirements in a certain area, the corresponding cell remains empty and is not included in subsequent calculations.

[0028] The system parses the area identifier and urgency level carried in temporary maintenance instructions or emergency response statuses, converting the urgency level into pulse intensity values ​​according to a preset mapping relationship. Temporary maintenance instructions or emergency response statuses are generated by the production management system when unplanned maintenance or emergencies occur. Their data format includes a fixed-length 16-byte field, where bytes 1 to 8 are the area code segment, and bytes 9 to 10 are the urgency code segment. The area code segment uses numerical encoding; for example, "AREA_01" represents the central control room, "AREA_02" represents the water treatment station, etc. By matching against a pre-stored safety-critical area identifier library, the corresponding area identifier is output (consistent with the area identifier on the grid's horizontal axis). The urgency code segment takes values ​​of 01, 02, or 03, corresponding to Level 1 (low), Level 2 (medium), and Level 3 (high) urgency levels, respectively. The preset mapping relationship uses a linear formula to calculate the base pulse intensity value: ; in, This represents the basic pulse intensity value, in minutes. This indicates the urgency level (1, 2, or 3). Therefore, the base pulse intensity value for Level 1 urgency is 5 points, Level 2 is 10 points, and Level 3 is 15 points.

[0029] Obtain the current performance load value corresponding to the region identifier in the dynamic performance baseline grid. Multiply the base pulse intensity value by the normalized complement of the current performance load value to output the final pulse intensity value. The current performance load value refers to the baseline performance value (i.e., fusion matching degree) already filled in the corresponding cell of this region. This value reflects the current performance completion level of the position in this region, and its value ranges from 0 to 100. The normalized complement is defined as 1 minus the quotient obtained by dividing the current performance load value by 100. The formula for calculating the final pulse intensity value is: ; in, This represents the final pulse intensity value, in minutes. The base pulse intensity value; This indicates the current performance load value (i.e., the baseline performance value in this cell). For example, a deputy general manager's baseline performance value in the control room area is 80 points. If a level 2 emergency order is received from the control room area (…), If the baseline performance score is 10, then the normalized complement is 1 minus 80 divided by 100, which equals 0.2. The final pulse intensity value is 10 multiplied by 0.2, which equals 2 points. If the baseline performance score is low (e.g., 20 points), then the normalized complement is 0.8, and the final pulse intensity value is 10 multiplied by 0.8, which equals 8 points. This calculation ensures that regions with good performance receive lower pulse intensity when receiving temporary instructions, avoiding over-intervention; while regions with insufficient performance receive higher pulse intensity, enhancing the urgency of task delivery.

[0030] The dynamic performance baseline grid is obtained by overlaying pulse intensity values ​​onto the baseline performance values ​​of corresponding cells in the blank grid. Specifically, the final pulse intensity value is added to the original baseline performance value of the cell to obtain a new cell value. If the same cell receives multiple temporary maintenance instructions or emergency response statuses within a short period, the final pulse intensity values ​​calculated each time are accumulated sequentially. The overlaid value may exceed 100; in this case, the actual accumulated value is retained without truncation to reflect the overload status. For cells that have not received any temporary instructions, their values ​​remain unchanged. After completing all overlay operations, the resulting grid is called the dynamic performance baseline grid.

[0031] The system iterates through each cell in the dynamic performance benchmark grid, comparing the superimposed current value with the standard thresholds in the job performance benchmark library. Cells below the lower threshold are marked as substandard, and cells above the upper threshold are marked as overloaded tasks. The job performance benchmark library predefines two thresholds for each job in each safety-critical area: a lower threshold and an upper threshold. For example, for the deputy general manager in the control room area, the lower threshold is set to 60 points, and the upper threshold is set to 95 points. The lower threshold represents the minimum performance requirement; values ​​below this are considered substandard. The upper threshold represents the normal performance limit; values ​​above this are considered overloaded tasks (indicating that the job has received too many temporary tasks or has exceeded its performance quota in this area). During iteration, empty cells (i.e., job-area combinations with no performance requirements) are ignored. The current value of each cell is compared with the corresponding lower and upper thresholds: if the current value is less than the lower threshold, the cell is marked as "not meeting the standard" and the difference between the current value and the lower threshold is recorded; if the current value is greater than the upper threshold, the cell is marked as "excessive task" and the difference between the current value and the upper threshold is recorded; otherwise, the cell is marked as "normal".

[0032] Sort all non-compliant items by cell value from low to high, and sort the excess task items by cell value from high to low. Extract the top three non-compliant items and the top two excess task items after sorting, attach the highest-level pulse flag to each, and output the task queue containing mandatory performance pulse signals in chronological order. During sorting, non-compliant items are based on their current cell value (i.e., the summed values); the lower the value, the greater the performance gap and the higher the priority, so the top three are selected. Excess task items are based on their current cell value; the higher the value, the more severe the task overload, so the top two are selected. For each extracted non-compliant item and excess task item, attach a highest-level pulse flag (represented by one byte, with a value of 0xFF). This flag is used to increase the signaling priority of the corresponding task in subsequent steps. Then, organize the output according to the following chronological order: first output all non-compliant items (in sorted order), then output all excess task items (in sorted order). Each output record contains a job identifier, a region identifier, the current cell value, a flag type (unfulfilled or excessive tasks), and a highest-level pulse identifier. These records together constitute a task queue containing mandatory performance pulse signals.

[0033] In the navigation signaling generation module, based on the task queue and the real-time spatial coordinate chain of the personnel performing their duties, the spatial guidance link corresponding to each unperformed task is calculated. Then, based on the spatial guidance link, a spatiotemporal navigation signaling package carrying regional risk levels and standard action guidelines is generated, specifically including: Extract the target area identifier corresponding to each unfulfilled task from the task queue. Use the latest timestamped coordinate point in the real-time spatial coordinate chain of the personnel performing the task as the starting coordinate, and map the target area identifier to the ending coordinate. Each unfulfilled task record in the task queue includes a job identifier, area identifier, current cell value, marker type, and highest-level pulse identifier. Extract the area identifier as the target area identifier, such as "central control room" or "water treatment station". The real-time spatial coordinate chain of the personnel performing the task is formed by arranging the continuously collected RFID positioning signals in ascending order of timestamps. Take the latest timestamped coordinate point in this sequence (including the horizontal and vertical coordinates and floor number) as the starting coordinate. A pre-generated area coordinate mapping table is created, where each safety-critical area identifier corresponds to an ending coordinate, which is selected as the center point or main entrance / exit location of the area (e.g., the ending coordinate of the central control room is set to the center point in front of the control panel). By looking up this mapping table, the target area identifier is converted into the ending coordinate.

[0034] Using Manhattan distance and regional connectivity constraints, a polyline path along a predefined traversable grid boundary is calculated between the starting and ending coordinates. This polyline path serves as the guiding link for the work space. The predefined traversable grid boundary divides the factory area map into square grid cells with sides of 1 meter. Each grid cell is marked as "traversable" (e.g., corridor, passageway, open area) or "impassable" (e.g., equipment base, fence, wall). Manhattan distance is defined as the sum of the absolute values ​​of the lateral and longitudinal differences between the starting and ending coordinates on the grid. Regional connectivity constraints require that the path can only move along adjacent traversable grid cells (up, down, left, and right directions), and oblique movement or crossing impassable cells is not allowed.

[0035] Using the grid cell containing the starting coordinates as the current cell, extract the set of walkable grid boundaries in the four directions (north, south, west, and east) of the current cell. Specifically, check the "walkable" marker of the cell adjacent to the north of the current cell; if it is walkable, add that direction to the set; similarly check the south, west, and east sides. If a cell adjacent to a certain direction exceeds the map boundary or is marked as impassable, it is not added to the set. This set is the candidate set of directions for movement allowed by the current cell.

[0036] The Manhattan distance from the current cell to the termination coordinate is decomposed into a horizontal difference and a vertical difference. The horizontal difference is calculated by subtracting the horizontal coordinate of the termination coordinate from the horizontal coordinate of the current cell and taking the absolute value. The vertical difference is calculated by subtracting the vertical coordinate of the termination coordinate from the vertical coordinate of the current cell and taking the absolute value. The horizontal and vertical differences are compared: if the horizontal difference is greater than the vertical difference, a horizontal movement direction (moving one cell to the left or right) is prioritized; if the vertical difference is greater than the horizontal difference, a vertical movement direction (moving one cell up or down) is prioritized; if they are equal, a horizontal movement direction is prioritized. After determining the preferred direction, its availability is checked from the set of accessible mesh boundaries for the current cell. If available, the cell is moved one cell along that direction, and the coordinates of the current cell are updated; if unavailable, another direction is selected from the set (first trying a direction perpendicular to the preferred direction, then trying the opposite direction) for movement.

[0037] Repeat the direction selection and movement steps, recalculating the horizontal and vertical differences between the current cell and the termination coordinates after each movement, until the coordinates of the current cell and the termination coordinates completely coincide. Each time a grid cell is crossed (i.e., each time a new cell is moved to a position different from the previous cell), the coordinates of the boundary point of the new cell (taking the coordinates of the center point of the grid cell) are recorded as the boundary turning point. All boundary turning points are connected sequentially according to the movement order to form a polyline path, which is the service space guidance link. If, during the movement, no passable direction is encountered (i.e., all four directions are blocked) and the termination coordinates have not yet been reached, the calculation is terminated, an empty link is output, and an "unreachable" message is generated.

[0038] The system retrieves the corresponding risk level label from the pre-stored risk level database based on the target area identifier, and extracts the corresponding standard action guidance statement from the action knowledge base based on the standard action code for the unfulfilled duty item. The risk level database pre-assigns a risk level label to each safety-critical area; for example, the central control room is "low risk," the water treatment station is "high risk," and the coal yard is "medium risk." The action knowledge base stores the mapping relationship between the standard action code and guidance statement for each unfulfilled duty item; for example, the code "ACTION_01" corresponds to "check the pressure gauge value upon arrival and take a photo and upload it." The sequence of polygonal nodes (coordinates of each boundary turning point arranged in timestamp order) carried by the duty performance guidance link is sequentially packaged with the risk level label and standard action guidance statement into a data packet. Simultaneously, the pulse intensity (i.e., the calculated final pulse intensity value) is parsed from the mandatory duty performance pulse signal, and this pulse intensity is appended to the data packet header as a signaling priority identifier (higher pulse intensity, higher priority). The final generated data packet is called a spatiotemporal navigation signaling packet, which is pushed to the mobile terminal carried by the personnel performing the duties for subsequent duty performance guidance and comparison feedback.

[0039] In the comparison and feedback module, after the spatiotemporal navigation signaling packet is pushed to the terminal carried by the personnel, the module compares the actual behavioral trajectory with the performance space guidance link and standard action guidance in the spatiotemporal navigation signaling packet to generate a performance correction signal and performance score report for the next data collection cycle. Specifically, this includes: The actual behavioral trajectory is discretized into a sequence of trajectory points based on timestamps. The polyline node sequence of the duty space guidance link and the expected time window for standard action guidance are extracted from the spatiotemporal navigation signaling packet. The actual behavioral trajectory is reported by the terminal device carried by the personnel (integrating RFID tags and location reporting functions) every 2 seconds, with each location point containing a timestamp, x-coordinate, y-coordinate, and floor number. All reported location points are arranged in ascending order by timestamp to obtain the trajectory point sequence. Two parts are parsed from the generated spatiotemporal navigation signaling packet: the polyline node sequence corresponding to the duty space guidance link (composed of a series of boundary turning point coordinates in sequence), and the expected time window corresponding to the standard action guidance. The expected time window is predefined by the standard action knowledge base; for example, the expected time window for the action "check the pressure gauge value upon arrival" is a time interval of 120 seconds, 60 seconds before and after the expected arrival time.

[0040] The spatial projection distance from each trajectory point in the trajectory point sequence to the nearest polyline segment is calculated sequentially. For each trajectory point, the nearest polyline segment (the line segment between two adjacent nodes) in the polyline node sequence is first found, and then the perpendicular projection distance from the trajectory point to the line segment is calculated. Specifically, the trajectory point is connected to the two endpoints of the line segment, forming two vectors. The dot product of these vectors is used to determine whether the projected point falls within the line segment's range. If the projected point lies on the line segment, the perpendicular distance is the distance from the point to the line; if the projected point is on the extension of the line segment, the shorter of the two endpoints is taken as the projection distance. A preset tolerance threshold of 1 meter is set, meaning trajectory points with a distance less than or equal to 1 meter are considered spatially undevised. Trajectory points with a projection distance exceeding 1 meter are marked as deviation points, their original timestamps are recorded, and the total number of deviation points is accumulated.

[0041] Align the sequence of deviation points with the expected time window of the standard action guideline on the timeline. Specifically, take the timestamp of each deviation point and determine if it falls within the expected time window range of the standard action guideline. For example, if the expected time window is from 9:00:00 AM to 9:02:00 AM, and the timestamp of a deviation point is 9:01:30 AM, then the deviation point is considered to be within the expected time window. If at least one deviation point exists within the expected time window, the standard action is marked as "not executed"; if no deviation points exist within the expected time window, it is marked as "executed". For multiple different standard action guidelines, each is judged independently according to its respective expected time window.

[0042] Based on the number of deviation points and the number of unexecuted standard actions, a performance correction signal is generated to reduce the data acquisition frequency of the next acquisition cycle. A preset threshold of 5 deviation points is set. If the cumulative number of deviation points exceeds 5, a correction signal is generated, indicating that the data acquisition frequency of the next acquisition cycle will be reduced from once every 2 seconds to once every 5 seconds to reduce the generation of invalid data. If the number of deviation points does not exceed 5, the acquisition frequency remains unchanged. Simultaneously, for each additional unexecuted standard action, a flag is added to the correction signal, indicating that the data acquisition frequency for the subsequent three consecutive acquisition cycles will be further reduced to once every 10 seconds.

[0043] Generate a performance score report with corresponding deductions. An initial performance score of 100 points is preset. 0.5 points are deducted for each deviation, and 5 points are deducted for each unexecuted standard action. Specifically, the deductions are calculated as follows: first, the total deduction for deviations is calculated (total number of deviations multiplied by 0.5); then, the total deduction for unexecuted actions is calculated (number of unexecuted standard actions multiplied by 5). These two calculations are added together to obtain the total deduction. The initial performance score is subtracted from the total deduction to obtain the current performance score. If the result is less than 0, it is set to 0. This performance score report is output in text format, including the employee's job title, the timeframe of the current evaluation period, the number of deviations, the number of unexecuted standard actions, and the final score.

[0044] The generated performance correction signal and performance score report are sent to the data acquisition layer and evaluation management layer, respectively. The performance correction signal is sent to the acquisition controller responsible for collecting RFID positioning signals. The controller adjusts the time interval of subsequent acquisition cycles based on the frequency indication in the signal. The performance score report is stored in the local database and simultaneously pushed to the employee and their supervisor via the company's instant messaging tool for performance improvement. After all the above processes are completed, the current evaluation cycle ends, and the next cycle will start collecting data again using the corrected acquisition frequency.

[0045] The working principle of this invention is as follows: Multi-source spatiotemporal behavioral data of personnel performing duties (including facial recognition access records, RFID positioning signals, and access control sequence data) and work environment data (including planned maintenance work orders, temporary operation orders, and environmental risk warning signals) are collected. This multi-source spatiotemporal behavioral data and work environment data are fused into a dynamic duty performance data package. The duration of stay and frequency of arrival of personnel in each safety-critical area are extracted from this data. Based on the job performance benchmark library, a fusion matching degree for each personnel's job is generated. According to the fusion matching degree and real-time acquired temporary maintenance instructions or emergency response status, a dynamic duty performance benchmark grid is constructed with job identifiers as the vertical axis and area identifiers as the horizontal axis. A pulse intensity value calculated from the urgency level and current duty load value is superimposed on the benchmark duty performance value of the grid. Cells below the lower threshold are marked as non-compliant items, and cells above the upper threshold are marked as overloaded tasks. The first three non-compliant items and the first two overloaded tasks are extracted and marked with the highest-level pulse identifier, forming a duty performance task queue containing mandatory duty performance pulse signals. From the duty performance task queue... The target area identifier for unfulfilled duties is extracted from the column. Using the latest coordinate point in the real-time spatial coordinate chain of the personnel performing duties as the starting coordinate, the target area identifier is mapped to the ending coordinate. The Manhattan distance and regional connectivity constraints are used to calculate the polyline path along the traversable grid boundary as the performance space guidance link. Combined with the risk level label and standard action guidance statement of the target area, a spatiotemporal navigation signaling packet carrying the polyline node sequence, risk level, standard action guidance and pulse priority is generated and pushed to the terminal carried by the personnel performing duties. Finally, the actual behavior trajectory reported by the terminal is discretized into a trajectory point sequence. The spatial projection distance of each trajectory point to the nearest polyline segment of the performance space guidance link is calculated. Deviation points exceeding the tolerance threshold and their timestamps are recorded. The deviation point sequence is aligned with the expected time window of the standard action guidance to determine whether the action has not been executed. Based on the number of deviation points and the number of unexecuted actions, a performance correction signal is generated to reduce the data collection frequency of the next collection cycle. At the same time, the corresponding score is deducted to obtain the performance score report, completing the entire dynamic evaluation and navigation loop of performance.

[0046] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A dynamic performance evaluation and navigation system for safety production management, characterized in that, include: The data acquisition module is used to collect multi-source spatiotemporal behavioral data and work environment data of the personnel performing their duties. The multi-source spatiotemporal behavioral data includes facial recognition access records, radio frequency identification positioning signals and access control sequence data. The work environment data includes planned maintenance work tickets, temporary operation tickets and environmental risk warning signals extracted from the production management system. The data fusion module integrates multi-source spatiotemporal behavioral data with operational environment data into a dynamic data package for job performance. Based on the dynamic data package, it extracts the dwell time and arrival frequency of personnel in each safety-critical area, and generates a fusion matching degree for each personnel's job based on the job performance benchmark library, covering the on-site area, job performance frequency, and standard actions. The task queue generation module constructs a dynamic performance benchmark grid for each employee based on the fusion matching degree and the real-time acquired temporary maintenance instructions or emergency response status. It also marks non-compliance items and excessive task items in the dynamic performance benchmark grid, forming a performance task queue containing mandatory performance pulse signals. The navigation signaling generation module calculates the performance space guidance link corresponding to each unfulfilled task based on the performance task queue and the real-time spatial coordinate chain of the personnel performing the tasks, and generates a spatiotemporal navigation signaling package carrying the regional risk level and standard action guidance based on the performance space guidance link. The comparison and feedback module pushes the spatiotemporal navigation signaling packet to the terminal carried by the personnel performing their duties. Based on the superposition and comparison of the actual behavior trajectory with the performance space guidance link and standard action guidance in the spatiotemporal navigation signaling packet, it generates a performance correction signal and performance score report for the next collection cycle.

2. The dynamic performance evaluation and navigation system for safety production management according to claim 1, characterized in that, The process of outputting the fusion matching degree is as follows: The facial recognition data at each time point is recorded and aligned with the radio frequency identification positioning signal by timestamp, and a spatiotemporal coordinate string carrying directional identifiers is synthesized based on the entry and exit direction marks in the access control time sequence data. The spatiotemporal coordinate string is compared with the predefined electronic fence of the safety critical area. The difference between the start and end time of each collision is extracted as the dwell time, and the number of collisions within a unit period is counted as the arrival frequency. The duration of stay and the frequency of arrival are multiplied by the standard weight of the corresponding region in the job performance benchmark database, the product results are accumulated and normalized, and then the fusion matching degree is output.

3. The dynamic performance evaluation and navigation system for safety production management according to claim 2, characterized in that, The synthesized spatiotemporal coordinate string carrying direction identifiers specifically includes: Extract the trigger time, personnel identification, and entry / exit direction markers of each access control event from the access control access sequence data, and arrange them in ascending order of time to form an access control event stream; The facial recognition system matches the recorded facial identifiers with the personnel identifiers in the access control event stream one by one. The trigger time of the successfully matched access control event is used as the anchor point, and the radio frequency identification positioning signal sequence within a preset time window before and after the anchor point is extracted. For each location point in the RFID location signal sequence, calculate the time difference between its timestamp and the trigger time of the most recent access control event. Based on the sign of the time difference and the entry / exit direction mark of the corresponding access control event, assign a direction identifier to the location point and output the spatiotemporal coordinate string carrying the direction identifier in chronological order.

4. The dynamic performance evaluation and navigation system for safety production management according to claim 1, characterized in that, The formation of the task queue containing mandatory performance pulse signals specifically includes: A blank grid is constructed with the job identifier of each employee as the vertical axis and the safety critical area identifier as the horizontal axis. The fusion matching degree is filled into the corresponding cell as the benchmark performance value. The area identifier and urgency level carried in the temporary maintenance order or emergency response status are analyzed. The urgency level is converted into a pulse intensity value according to a preset mapping relationship. The pulse intensity value is superimposed on the baseline performance value of the corresponding area cell in the blank grid to obtain a dynamic performance baseline grid. Traverse each cell in the dynamic performance benchmark grid, compare the superimposed current value with the standard threshold in the job performance benchmark library, mark cells below the lower threshold as substandard items, and mark cells above the upper threshold as excessive task items; Sort all non-compliant items by cell value from low to high, and sort the excess task items by cell value from high to low. Extract the top three non-compliant items and the top two excess task items after sorting, attach the highest level pulse identifier to each, and output the task queue containing the mandatory performance pulse signal in chronological order.

5. The dynamic performance evaluation and navigation system for safety production management according to claim 4, characterized in that, The process of converting the urgency level into a pulse intensity value according to a preset mapping relationship specifically includes: Extract the area code segment and urgency code segment from the preset field location from the temporary maintenance order or emergency response status, match the area code segment with the pre-stored safety critical area identifier library, and output the area identifier; Based on the urgency coding segment, look up the preset three-level pulse mapping table and output the basic pulse intensity value corresponding to the urgency level; Obtain the current duty load value corresponding to the region identifier in the dynamic duty load benchmark grid, multiply the base pulse intensity value by the normalized complement of the current duty load value, and output the final pulse intensity value.

6. The dynamic performance evaluation and navigation system for safety production management according to claim 1, characterized in that, The generation of the spatiotemporal navigation signaling packet carrying the regional risk level and standard action guidelines specifically includes: Extract the target area identifier corresponding to each unfulfilled task from the task queue, obtain the coordinate point with the latest timestamp in the real-time spatial coordinate chain of the personnel performing the task as the starting coordinate, and map the target area identifier to the ending coordinate; Using Manhattan distance and regional connectivity constraints, the polyline path between the starting and ending coordinates along the preset accessible grid boundary is calculated, and the polyline path is used as the guiding link for the performance space. Based on the target area identifier, the pre-stored risk level database is retrieved, the corresponding risk level label is extracted, and the corresponding standard action guidance statement is extracted from the action knowledge base according to the standard action code of the unfulfilled duties. The spatiotemporal navigation signaling packet is generated by packaging the broken line node sequence carried by the duty performance space guidance link with risk level labels and standard action guidance statements in timestamp order, and adding the pulse intensity parsed from the mandatory duty performance pulse signal as the signaling priority identifier.

7. A dynamic performance evaluation and navigation system for safety production management according to claim 6, characterized in that, The process of outputting the polyline segment path is as follows: Using the grid cell containing the starting coordinates as the current cell, extract the set of passable grid boundaries in the four neighboring directions of the current cell; The Manhattan distance from the current cell to the termination coordinate is decomposed into horizontal and vertical differences. Prioritize moving one cell along the direction with the larger absolute value of the difference to the accessible grid boundary, and update the current coordinates. Repeat the direction selection action until the current coordinates coincide with the termination coordinates, record the sequence of boundary turning points each time the grid cell is crossed, and connect the turning points in sequence to output the polyline path.

8. A dynamic performance evaluation and navigation system for safety production management according to claim 1, characterized in that, The comparison and integration of the actual behavioral trajectory with the performance space guidance link and standard action instructions in the spatiotemporal navigation signaling packet specifically includes: The actual behavior trajectory is discretized into a sequence of trajectory points according to the timestamp, and the broken line node sequence of the performance space guidance link and the expected time window of the standard action guidance are extracted from the spatiotemporal navigation signaling packet. Calculate the spatial projection distance from each trajectory point in the trajectory point sequence to the nearest polyline segment, record the deviation points whose projection distance exceeds the preset tolerance threshold and their timestamps, and accumulate the number of deviation points; Align the sequence of deviations with the expected time window of the standard action guide on the timeline, determine whether there are deviations within the expected time window, and mark the standard action as not executed if there are deviations. Based on the number of deviations and the number of standard actions not performed, a performance correction signal is generated to reduce the data collection frequency in the next collection cycle, and a performance score report is generated to deduct the corresponding score.

Images