Access control equipment fault prediction system and method based on multi-source data
By integrating multi-source data and building a performance degradation prediction model, the problems of low efficiency and insufficient prediction accuracy in the fault management of traditional access control equipment are solved. This achieves efficient fault prediction and optimized allocation of maintenance resources, thereby improving equipment stability and access experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-09
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional access control equipment fault management methods are inefficient, struggle to capture early signs of faults, have limited data collection dimensions, lack accurate predictions, and lack intelligent analysis and dynamic adaptation capabilities. This results in biased and inaccurate fault predictions, making it prone to misjudgments and omissions.
By integrating multi-source data from building management, security, and environmental monitoring systems, an equipment aging sample database is constructed, a performance degradation prediction model is trained, the historical operating pressure index and comprehensive failure risk score of the equipment are calculated, a preventive maintenance priority queue is generated, and the allocation of maintenance resources is optimized.
It enables efficient fault prediction of access control equipment, avoids missed faults in old equipment, optimizes the allocation of maintenance resources, reduces the probability of sudden faults, and improves the stability of equipment operation and the experience of personnel passage.
Smart Images
Figure CN121836031A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of access control equipment fault prediction technology, specifically a fault prediction system and method for access control equipment based on multi-source data. Background Technology
[0002] With the increasing demands for access security and efficiency in building security and park management, access control equipment has been widely used in various scenarios such as office buildings and residential communities, becoming a core component of the security system. Its stable operation is directly related to regional security control and personnel access experience. Therefore, it is of great practical significance to predict the failure of access control equipment and carry out preventive maintenance.
[0003] However, traditional access control equipment fault management methods often face the following problems when dealing with the fault prevention needs of multiple devices and complex scenarios: First, they rely on manual inspections, which are inefficient and prone to missed diagnoses. Traditional access control maintenance often adopts a periodic manual inspection mode. Faced with a large number of access control devices in the target area, this not only consumes a lot of manpower and resources, but also makes it difficult for humans to capture early signs of equipment failure, often missing the best maintenance time, resulting in sudden failures that affect access. Second, the data collection dimensions are limited, and the prediction accuracy is insufficient. Existing related methods mostly rely only on the basic operating data of the access control devices themselves, without integrating multi-source related data such as traffic flow and equipment aging status. This fails to comprehensively reflect the actual operating load and wear and tear of the equipment, resulting in strong bias and large deviations in fault prediction. In addition, they lack intelligent analysis and dynamic adaptation capabilities. Existing prediction methods mostly use fixed thresholds to judge anomalies, without considering the differences in different scenarios and equipment aging stages. This easily leads to problems such as misjudgment during peak periods and missed detection of old equipment. Furthermore, they do not combine the contextual characteristics of historical data to match similar operating conditions, making it difficult to identify potential high-load risks. Summary of the Invention
[0004] The purpose of this invention is to provide a fault prediction system and method for access control equipment based on multi-source data, so as to solve the problems raised in the prior art.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting faults in access control equipment based on multi-source data, the method comprising the following steps: Obtain the historical operation dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. Calculate the historical operating pressure index of the access control device within the corresponding time slice, compare it with the preset dynamic health baseline threshold, mark the time slices that exceed the threshold, and summarize them to form a set of historical high load periods of the access control device; Determine the starting time and predicted duration for the assessment, obtain the historical high-load period set of the target access control device, and define the context features of any time slice; determine the corresponding context features for each future time slice, filter out the set of historical time slices that match the features from the historical data based on the features, calculate the proportion of time slices in the set that belong to the historical high-load period, and determine whether the future time slice is in a high-load risk state based on the preset proportion threshold. The number of time slots in which access control equipment is judged to be in a high-load risk state within the predicted time period is statistically analyzed. The comprehensive failure risk score of the access control equipment within the corresponding predicted time period is calculated. Based on the comprehensive risk score of each device, a preventive maintenance priority queue is generated, and maintenance personnel are arranged to carry out maintenance work according to the queue.
[0006] Obtain the historical operational dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. The specific steps include: After establishing connections and achieving data sharing with the building management platform, security platform, and environmental monitoring system, the historical operation dataset H of all access control devices within the target area is obtained, where H = {H1, H2, ..., H...}. x H X}, where H x Let x represent the set of multi-source data information of the x-th access control device within the historical observation period, where x = 1, 2, ..., X, and X represents the total number of access control devices in the target area; H is a set of historical data information of any access control device. a The historical data information includes the frequency E of abnormal events occurring in access control device a during the t-th historical monitoring time slice. at Average response delay D at , where D at Let t represent the average time delay from receiving the instruction to the completion of the door lock mechanism action of access control device a in time slice t, where t = 1, 2, ..., T, and T represents the total number of time slices divided within the historical observation period, and a ∈ {1, 2, ..., X}.
[0007] The specific steps for calculating the frequency of abnormal events occurring in access control device a during the t-th historical monitoring time slice are as follows: Obtain the set of abnormal event types S = {s1, s2, ..., s} for access control device a. n}, where s1, s2, ..., s nThese represent the 1st, 2nd, ..., nth types of abnormal events of access control device a, where n represents the number of abnormal event types of access control device a. Based on the obtained set of abnormal event types of access control device a, and the abnormal events s of access control device a within time slice t. i The frequency of occurrence is calculated by determining the frequency of abnormal events occurring in access control device a within the t-th historical monitoring time slice, as defined below: E at =Σ i=1 n (w) i ×f a,t,i ); where w i Indicates abnormal event s i The corresponding weighting coefficient, f a,t,i Indicates abnormal event s i The number of times this occurred within the t-th monitoring time slice in history.
[0008] Based on the ratio of abnormal event frequency to total number of operations, and the ratio of actual average response delay to maximum allowable response delay, the historical operating pressure index of the access control device within the corresponding time slice is calculated in conjunction with the device's basic performance coefficient. This index is then compared with a preset dynamic health baseline threshold, and time slices exceeding this threshold are marked. These are then compiled into a set of historical high-load periods for the access control device. The specific steps include: According to formula R at =[b1×(E at / N at ) + b2 × (D at / D ma )] / B a Among them, R at N represents the historical operating pressure index of access control device a within time slice t. at D represents the total number of operations performed by access control device a within time slice t. ma B represents the maximum allowable response delay threshold for access control device A in its design. a This represents the basic performance coefficient of access control device 'a'. Historical operating stress index greater than the dynamic health baseline threshold α at The time slices are marked, and the dynamic health baseline threshold α at The set P represents the historical high-load periods of access control device a, calculated by adding one standard deviation to the mean of the pressure index for the same historical period. a P a ={p a1 p a2 , ..., p ak , ..., p aK}, p akLet k represent the time slice in which access control device a is marked as historically high-load, where k = 1, 2, ..., K, and K represents the number of times access control device a is marked as high-load within the historical period.
[0009] The methods for determining the basic performance coefficients of access control device 'a' include: Construct an equipment aging sample database: Obtain historical maintenance records for Q access control devices of the same product model, where Q represents the total number of access control devices of the same product model as the target device. Each record corresponds to the status of one device at a specific point in time and includes feature data and tag data. The feature data includes cumulative usage time, cumulative number of operations, preset total number of times the device is used within its lifespan, and average daily usage. The tag data is the actual performance degradation ratio R, defined as R=D. j0 / P0; where D j0 P0 represents the average response time of device sample j during a recent period of low load, while P0 represents the rated response time of this model. Using the feature data as input and the label data R as the prediction target, a supervised learning algorithm is used to train the regression model M; the supervised learning algorithm is either the gradient boosting decision tree algorithm or the random forest algorithm. The current feature data X of acquisition device a a Input it into the trained performance degradation prediction model M to obtain the predicted degradation factor β output by the model. a ; Calculate the basic performance coefficient B a =β a .
[0010] The process involves determining the start time and predicted duration for the assessment, obtaining a set of historical high-load periods for the target access control equipment, defining the contextual features of any time slice including weekday attributes, time period attributes, and pedestrian flow level attributes, and classifying different pedestrian flow levels based on the quantiles of historical pedestrian flow data. For each future time slice, a corresponding contextual feature is determined, and a set of historical time slices that perfectly match this feature is selected from historical data. The proportion of historical time slices belonging to historical high-load periods in the selected set is calculated, and a preset proportion threshold is used to determine whether the future time slice is in a high-load risk state. Specific steps include: Determine the starting time point t0 and the predicted duration S to be evaluated. The predicted duration corresponds to S future time slices. For access control device a, obtain its historical high-load period set P. a ={p a1 p a2 , ..., p ak , ..., p aK}, define the context feature triple C = (W, HT, L) of the device for any time slice. W is the weekday identifier, taking values from the set {weekday, weekend, holiday}. HT is the time period identifier, which is the hour number in a day. L is the pedestrian flow level identifier, divided into three levels {low, medium, high} based on historical data. The division method is as follows: Obtain the historical pedestrian flow data sequence of all time slices of the access control device a within the historical observation period, sort the historical pedestrian flow data sequence, calculate its first quartile Q1 (i.e., the 33% quantile) and second quartile Q2 (i.e., the 66% quantile). Based on the quantiles, establish the pedestrian flow level determination rule: Low level: The real-time pedestrian flow l ≤ Q1; Medium level: Q1 < l ≤ Q2; High level: l > Q2; For each future time slice s, s = 1, 2,..., S, calculate its absolute time point as t s = t0 + s×Δt, where Δt represents the length of a single time slice, and determine the context feature vector C of the access control device a within this time slice as = (W s , HT s , L s ), where W s and HT s are determined by the calendar and hour of t s , and L s is mapped to a level by querying the average pedestrian flow data of the access control device a in the historical same period according to a preset threshold. From the historical data, filter out all historical time slice sets V as that satisfy the context similarity condition with C s . The similarity condition is: the same W value, the same HT value, and the same L value; if there is no match with the same L value in history, relax the condition to that the L value difference does not exceed one level; <00**********>For each future time slice t s , calculate the proportion of the time slices belonging to the historical high-load period set P s in its corresponding similar historical time slice set V a , denoted as ρ s . The calculation formula is as follows: ρ s = N / U, where N is the number of time slices t that satisfy t ∈ V s and t ∈ P a , and U is the total number of time slices in the set V s . If V s is an empty set, then define ρ s = 0. If ρ s > δ, then predict that the access control device a is in a high-load risk state at the time slice ts, where δ is a preset proportion threshold.
[0011] The number of time slots during which access control equipment is identified as being in a high-risk state within the predicted duration is statistically analyzed, and the comprehensive fault risk score of the access control equipment within the corresponding predicted duration is calculated. Specific steps include: The number of time slices within a statistical prediction period S in which access control device a is predicted to be in a high-load risk state is denoted as K. a According to the formula Rs a =K a / S; Calculate the comprehensive failure risk score of device a within the predicted future time.
[0012] Generate a preventative maintenance priority queue, and schedule maintenance personnel to carry out maintenance work based on this queue. Specific steps include: The system iterates through all access control devices within the target area to obtain a comprehensive fault risk score set for each device. Based on this, all devices are sorted in descending order to generate a preventive maintenance priority queue. Maintenance personnel are then assigned to perform maintenance according to the preventive maintenance priority queue.
[0013] A fault prediction system for access control equipment based on multi-source data includes: a data acquisition module, a parameter calculation module, a model training module, a risk prediction module, a score generation module, and a maintenance scheduling module. The data acquisition module acquires historical operational datasets of access control equipment in a target area, selecting any one device's historical data set. The parameter calculation module calculates the historical operational pressure index for the corresponding time slice of the device, compares it with a preset dynamic health baseline threshold, marks time slices exceeding the threshold, and aggregates them into a historical high-load period set. The model training module constructs a device aging sample database and trains a performance degradation prediction model using a supervised learning algorithm. The risk prediction module determines the starting time point and prediction duration for evaluation, defines time slice context features, matches each future time slice with a corresponding set of historical time slices, and judges future high-load risk by calculating the proportion of high-load periods in the set. The score generation module counts the number of high-load risk time slices within the prediction duration and calculates the comprehensive fault risk score for the equipment. The maintenance scheduling module generates a preventative maintenance priority queue based on the comprehensive fault risk score of each device and schedules maintenance personnel to perform maintenance according to the queue.
[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. By integrating multi-source data from building management, security and environmental monitoring systems, and training a performance degradation prediction model using an equipment aging sample database, a basic performance coefficient suitable for the aging state of the equipment is obtained, and the historical operating pressure index calculation results are corrected. Unlike the fixed evaluation method in the existing technology that ignores the differences in equipment aging, this invention can match the actual operating state of new and old equipment and avoid missing faults in old equipment. 2. By calculating the comprehensive fault risk score by statistically analyzing the number of high-load risk time slices, a preventive maintenance priority queue is generated. Unlike the indiscriminate maintenance scheduling method in the prior art, this invention can optimize the allocation of maintenance resources, prioritize the handling of high-risk equipment, and reduce the probability of sudden failures. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating a method for predicting faults in access control equipment based on multi-source data, as proposed in this invention. Detailed Implementation
[0016] 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.
[0017] like Figure 1 As shown, the present invention provides a technical solution, a method for predicting faults in access control equipment based on multi-source data, the method comprising the following steps: Obtain the historical operation dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. Calculate the historical operating pressure index of the access control device within the corresponding time slice, compare it with the preset dynamic health baseline threshold, mark the time slices that exceed the threshold, and summarize them to form a set of historical high load periods of the access control device; Determine the starting time and predicted duration for the assessment, obtain the historical high-load period set of the target access control device, and define the context features of any time slice; determine the corresponding context features for each future time slice, filter out the set of historical time slices that match the features from the historical data based on the features, calculate the proportion of time slices in the set that belong to the historical high-load period, and determine whether the future time slice is in a high-load risk state based on the preset proportion threshold. The number of time slots in which access control equipment is judged to be in a high-load risk state within the predicted time period is statistically analyzed. The comprehensive failure risk score of the access control equipment within the corresponding predicted time period is calculated. Based on the comprehensive risk score of each device, a preventive maintenance priority queue is generated, and maintenance personnel are arranged to carry out maintenance work according to the queue.
[0018] Obtain the historical operational dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. The specific steps include: After establishing connections and achieving data sharing with the building management platform, security platform, and environmental monitoring system, the historical operation dataset H of all access control devices within the target area is obtained, where H = {H1, H2, ..., H...}. x H X}, where H x Let x represent the set of multi-source data information of the x-th access control device within the historical observation period, where x = 1, 2, ..., X, and X represents the total number of access control devices in the target area; H is a set of historical data information of any access control device. a The historical data information includes the frequency E of abnormal events occurring in access control device a during the t-th historical monitoring time slice. at Average response delay D at , where D at Let t represent the average time delay from receiving the instruction to the completion of the door lock mechanism action of access control device a in time slice t, where t = 1, 2, ..., T, and T represents the total number of time slices divided within the historical observation period, and a ∈ {1, 2, ..., X}.
[0019] The specific steps for calculating the frequency of abnormal events occurring in access control device a during the t-th historical monitoring time slice are as follows: Obtain the set of abnormal event types S = {s1, s2, ..., s} for access control device a. n}, where s1, s2, ..., s n These represent the 1st, 2nd, ..., nth types of abnormal events of access control device a, where n represents the number of abnormal event types of access control device a. Based on the obtained set of abnormal event types of access control device a, and the abnormal events s of access control device a within time slice t. i The frequency of occurrence is calculated by determining the frequency of abnormal events occurring in access control device a within the t-th historical monitoring time slice, as defined below: E at =Σ i=1 n (w) i ×f a,t,i ); where w i Indicates abnormal event s i The corresponding weighting coefficient, f a,t,i Indicates abnormal event s i The number of times this occurred within the t-th monitoring time slice in history.
[0020] Based on the ratio of abnormal event frequency to total number of operations, and the ratio of actual average response delay to maximum allowable response delay, the historical operating pressure index of the access control device within the corresponding time slice is calculated in conjunction with the device's basic performance coefficient. This index is then compared with a preset dynamic health baseline threshold, and time slices exceeding this threshold are marked. These are then compiled into a set of historical high-load periods for the access control device. The specific steps include: According to formula R at =[b1×(E at / N at ) + b2 × (D at / D ma )] / B a Among them, R at N represents the historical operating pressure index of access control device a within time slice t. at D represents the total number of operations performed by access control device a within time slice t. ma B represents the maximum allowable response delay threshold for access control device A in its design. a This represents the basic performance coefficient of access control device 'a'. Historical operating stress index greater than the dynamic health baseline threshold α at The time slices are marked, and the dynamic health baseline threshold α at The set P represents the historical high-load periods of access control device a, calculated by adding one standard deviation to the mean of the pressure index for the same historical period. a P a ={p a1 p a2 , ..., p ak , ..., p aK}, p ak Let k represent the time slice in which access control device a is marked as historically high-load, where k = 1, 2, ..., K, and K represents the number of times access control device a is marked as high-load within the historical period.
[0021] The methods for determining the basic performance coefficients of access control device 'a' include: Construct an equipment aging sample database: Obtain historical maintenance records for Q access control devices of the same product model, where Q represents the total number of access control devices of the same product model as the target device. Each record corresponds to the status of one device at a specific point in time and includes feature data and tag data. The feature data includes cumulative usage time, cumulative number of operations, preset total number of times the device is used within its lifespan, and average daily usage. The tag data is the actual performance degradation ratio R, defined as R=D. j0 / P0; where D j0denotes the average response time of device sample j during the most recent low-load period, and P0 denotes the rated response time of this model; Taking the feature data as input and the label data R as the prediction target, a regression model M is trained using a supervised learning algorithm; the supervised learning algorithm uses a gradient boosting decision tree algorithm or a random forest algorithm; Collect the current feature data X of device a a and input it into the trained performance degradation prediction model M to obtain the predicted degradation multiple β output by the model a ; Calculate the basic performance coefficient B a = β a .
[0022] Determine the starting time point and prediction duration to be evaluated, obtain the set of historical high-load periods of the target access control device, define the context features of any time slice to include weekday attribute, time period attribute, and pedestrian flow level attribute, and divide different pedestrian flow levels based on the quantiles of historical pedestrian flow data; determine the corresponding context features for each future time slice, screen out the set of historical time slices that exactly match this feature from historical data, calculate the proportion of historical time slices belonging to historical high-load periods in the screened set of historical time slices, and judge whether this future time slice is in a high-load risk state according to a preset proportion threshold. The specific steps include: Determine the starting time point t0 and prediction duration S to be evaluated. The prediction duration corresponds to the next S time slices. For access control device a, obtain its set of historical high-load periods P a ={p a1 , p a2 ,..., p ak ,..., p aK}, define the context feature triple C = (W, HT, L) of the device at any time slice, where W is the weekday identifier, taking values from the set {weekday, weekend, holiday}, HT is the time period identifier, which is the hourly time period number of a day, and L is the pedestrian flow level identifier, divided into three levels {low, medium, high} based on historical data. The division method is as follows: Obtain the historical pedestrian flow data sequence of all time slices of access control device a during the historical observation period, sort the historical pedestrian flow data sequence, calculate its first quartile Q1 (i.e., the 33% quantile) and second quartile Q2 (i.e., the 66% quantile), and establish a pedestrian flow level determination rule based on the quantiles: low level: real-time pedestrian flow l ≤ Q1; medium level: Q1 < l ≤ Q2; high level: l > Q2; For each future time slice s, s = 1, 2,..., S, calculate its absolute time point as t s = t0 + s × Δt, where Δt represents the length of a single time slice, and determine the context feature vector C of access control device a within this time sliceas =(W s HT s L s ), where W s and HT s By t s The calendar and time are determined, L s By querying the average pedestrian traffic data of access control device A during the same historical period, and mapping it to a level according to a preset threshold, all devices with the same traffic level as C are filtered out from the historical data. as The set of historical time slices V that satisfy the condition of context similarity s The similarity conditions are: the same W value, the same HT value, and the same L value; if there are no matches with the same L value in the history, the condition is relaxed to the point that the L values differ by no more than one level. For each future time slice t s Calculate the corresponding set of similar historical time slices V s In the middle, it belongs to the set P of historically high load periods. a The proportion of time slices, denoted as ρ s The calculation formula is as follows: ρ s =N / U, where N is the number of integers satisfying t∈V. s And t∈P a The number of time slices t, where U is the set V. s The total number of time slices, if V s If it is an empty set, then define ρ. s =0, if ρ s If the value is greater than δ, then the access control device a is predicted to be in a high-load risk state in time slice ts, where δ is a preset proportional threshold.
[0023] The number of time slots during which access control equipment is identified as being in a high-risk state within the predicted duration is statistically analyzed, and the comprehensive fault risk score of the access control equipment within the corresponding predicted duration is calculated. Specific steps include: The number of time slices within a statistical prediction period S in which access control device a is predicted to be in a high-load risk state is denoted as K. a According to the formula Rs a =K a / S; Calculate the comprehensive failure risk score of device a within the predicted future time.
[0024] Generate a preventative maintenance priority queue, and schedule maintenance personnel to carry out maintenance work based on this queue. Specific steps include: The system iterates through all access control devices within the target area to obtain a comprehensive fault risk score set for each device. Based on this, all devices are sorted in descending order to generate a preventive maintenance priority queue. Maintenance personnel are then assigned to perform maintenance according to the preventive maintenance priority queue.
[0025] In Example 1: A stable data sharing connection is established with the building management platform, security monitoring platform and environmental monitoring system to comprehensively collect historical operating data of all access control devices. These data cover various operating information of each device during long-term use. For any selected access control device, its corresponding historical data set is extracted, and the relevant data of each time slice in the historical observation period are filtered out, including the occurrence of different types of abnormal events and the average response delay from the device receiving the access command to the door lock mechanism completing the opening and closing action. The historical operating pressure index of the equipment is calculated, and various types of abnormal events that may occur are identified. Different weights are assigned based on the impact of each abnormal event on equipment operation. The total frequency of abnormal events in each time slice is calculated by combining the frequency of occurrence of each type of abnormal event within the corresponding time slice. Then, the total number of operations, the maximum allowable response delay, and the basic performance coefficient of the equipment are obtained for each time slice. The basic performance coefficient is obtained by constructing an aging sample database of the same model of equipment, collecting the cumulative usage characteristics and actual performance degradation of the sample equipment, training a prediction model using a suitable supervised learning algorithm, and inputting the current feature data of the target equipment. The historical operating pressure index for each time slice is calculated by integrating the ratio of abnormal event frequency to total number of operations, the ratio of actual response delay to maximum allowable response delay, and the basic performance coefficient. This pressure index is compared with a preset dynamic health baseline threshold, which is determined based on the statistical characteristics of the equipment's historical pressure index for the same period. Time slices exceeding the threshold are marked, and these are compiled into a set of historical high-load periods for the equipment. To predict future high-load risks, the starting time and prediction period for assessment are determined, the future time range covered by the prediction is clarified, and the contextual characteristics of time slices are defined, including the attribute identifiers of weekdays, weekends, or holidays, the specific time period identifiers within a day, and the traffic flow level identifiers. The traffic flow level is divided into different levels based on the distribution characteristics of the equipment's historical traffic flow data. For each future time slice, the weekday attribute and time period attribute are determined based on its corresponding calendar information and hour information. The traffic flow level is determined by combining the traffic flow data of the same period in history, forming a contextual feature vector for the future time slice. A set of historical time slices that completely match the feature vector is selected from the historical data. If there is no complete match, the matching conditions for the traffic flow level are appropriately relaxed. The proportion of historical high-load periods in the selected set of historical time slices is calculated and compared with a preset proportion threshold to determine whether the future time slice is in a high-load risk state. The number of time slices in which equipment is identified as being in a high-load risk state within the statistical prediction period is counted. Combined with the total prediction duration, the comprehensive failure risk score of the equipment is calculated. The above steps are repeated for all access control equipment to obtain the comprehensive failure risk score of each equipment. The equipment is then sorted according to the score to generate a preventive maintenance priority queue. Maintenance management personnel use this queue to prioritize the scheduling of resources to carry out maintenance operations on high-risk equipment.
[0026] A fault prediction system for access control equipment based on multi-source data includes: a data acquisition module, a parameter calculation module, a model training module, a risk prediction module, a score generation module, and a maintenance scheduling module. The data acquisition module acquires historical operational datasets of access control equipment in a target area, selecting any one device's historical data set. The parameter calculation module calculates the historical operational pressure index for the corresponding time slice of the device, compares it with a preset dynamic health baseline threshold, marks time slices exceeding the threshold, and aggregates them into a historical high-load period set. The model training module constructs a device aging sample database and trains a performance degradation prediction model using a supervised learning algorithm. The risk prediction module determines the starting time point and prediction duration for evaluation, defines time slice context features, matches each future time slice with a corresponding set of historical time slices, and judges future high-load risk by calculating the proportion of high-load periods in the set. The score generation module counts the number of high-load risk time slices within the prediction duration and calculates the comprehensive fault risk score for the equipment. The maintenance scheduling module generates a preventative maintenance priority queue based on the comprehensive fault risk score of each device and schedules maintenance personnel to perform maintenance according to the queue.
[0027] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
Claims
1. A method for predicting faults in access control equipment based on multi-source data, characterized in that: The method includes the following steps: Obtain the historical operation dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. Calculate the historical operating pressure index of the access control device within the corresponding time slice, compare it with the preset dynamic health baseline threshold, mark the time slices that exceed the threshold, and summarize them to form a set of historical high load periods of the access control device; Determine the starting time and predicted duration for the assessment, obtain the historical high-load period set of the target access control device, and define the context features of any time slice; determine the corresponding context features for each future time slice, filter out the set of historical time slices that match the features from the historical data based on the features, calculate the proportion of time slices in the set that belong to the historical high-load period, and determine whether the future time slice is in a high-load risk state based on the preset proportion threshold. The number of time slots in which access control equipment is judged to be in a high-load risk state within the predicted time period is statistically analyzed. The comprehensive failure risk score of the access control equipment within the corresponding predicted time period is calculated. Based on the comprehensive risk score of each device, a preventive maintenance priority queue is generated, and maintenance personnel are arranged to carry out maintenance work according to the queue.
2. The method for predicting access control equipment faults based on multi-source data according to claim 1, characterized in that: Obtain the historical operational dataset of access control devices within the target area, and arbitrarily select the historical data information set of one access control device. The historical data information includes the frequency of abnormal events occurring in each time slice of the historical observation period of the target access control device and the average response delay from receiving the instruction to the completion of the door lock mechanism action. The specific steps include: After establishing connections and achieving data sharing with the building management platform, security platform, and environmental monitoring system, the historical operation dataset H of all access control devices within the target area is obtained, where H = {H1, H2, ..., H...}. x H X }, where H x Let x represent the set of multi-source data information of the x-th access control device within the historical observation period, where x = 1, 2, ..., X, and X represents the total number of access control devices in the target area; H is a set of historical data information of any access control device. a The historical data information includes the frequency E of abnormal events occurring in access control device a during the t-th historical monitoring time slice. at Average response delay D at , where D at Let t represent the average time delay from receiving the instruction to the completion of the door lock mechanism action of access control device a in time slice t, where t = 1, 2, ..., T, and T represents the total number of time slices divided within the historical observation period, and a ∈ {1, 2, ..., X}.
3. The method for predicting access control equipment faults based on multi-source data according to claim 2, characterized in that: The specific steps for calculating the frequency of abnormal events occurring in access control device a during the t-th historical monitoring time slice are as follows: Obtain the set of abnormal event types S = {s1, s2, ..., s} for access control device a. n }, where s1, s2, ..., s n These represent the 1st, 2nd, ..., nth types of abnormal events of access control device a, where n represents the number of abnormal event types of access control device a. Based on the obtained set of abnormal event types of access control device a, and the abnormal events s of access control device a within time slice t. i The frequency of occurrence is calculated by determining the frequency of abnormal events occurring in access control device a within the t-th historical monitoring time slice, as defined below: E at =Σ i=1 n (w) i ×f a,t,i ); where w i Indicates abnormal event s i The corresponding weighting coefficient, f a,t,i Indicates abnormal event s i The number of times this occurred within the t-th monitoring time slice in history.
4. The method for predicting access control equipment faults based on multi-source data according to claim 3, characterized in that: Based on the ratio of the frequency of abnormal events to the total number of operations, the ratio of the actual average response delay to the maximum allowable response delay, and combined with the equipment's basic performance coefficient, the historical operating pressure index of the gate access control equipment within the corresponding time slice is calculated. The data is compared with a preset dynamic health baseline threshold, time slices exceeding the threshold are marked, and these are aggregated to form a historical high-load period set for the access control device. Specific steps include: According to formula R at =[b1×(E at / N at ) + b2 × (D at / D ma )] / B a Among them, R at N represents the historical operating pressure index of access control device a within time slice t. at D represents the total number of operations performed by access control device a within time slice t. ma B represents the maximum allowable response delay threshold for access control device A in its design. a This represents the basic performance coefficient of access control device 'a'. Historical operating stress index greater than the dynamic health baseline threshold α at The time slices are marked, and the dynamic health baseline threshold α at The set P represents the historical high-load periods of access control device a, calculated by adding one standard deviation to the mean of the pressure index for the same historical period. a P a ={p a1 p a2 , ..., p ak , ..., p aK }, p ak Let k represent the time slice in which access control device a is marked as historically high-load, where k = 1, 2, ..., K, and K represents the number of times access control device a is marked as high-load within the historical period.
5. The method for predicting access control equipment faults based on multi-source data according to claim 4, characterized in that: The methods for determining the basic performance coefficients of access control device 'a' include: Construct an equipment aging sample database: Obtain historical maintenance records for Q access control devices of the same product model, where Q represents the total number of access control devices of the same product model as the target device. Each record corresponds to the status of one device at a specific point in time and includes feature data and tag data. The feature data includes cumulative usage time, cumulative number of operations, preset total number of times the device is used within its lifespan, and average daily usage. The tag data is the actual performance degradation ratio R, defined as R=D. j0 / P0; where D j0 P0 represents the average response time of device sample j during a recent period of low load, while P0 represents the rated response time of this model. Using the feature data as input and the label data R as the prediction target, a supervised learning algorithm is used to train the regression model M; the supervised learning algorithm is either the gradient boosting decision tree algorithm or the random forest algorithm. The current feature data X of acquisition device a a Input it into the trained performance degradation prediction model M to obtain the predicted degradation factor β output by the model. a ; Calculate the basic performance coefficient B a =β a .
6. The method for predicting faults in access control equipment based on multi-source data according to claim 5, characterized in that: The process involves determining the start time and predicted duration for the assessment, obtaining a set of historical high-load periods for the target access control equipment, defining the contextual features of any time slice including weekday attributes, time period attributes, and pedestrian flow level attributes, and classifying different pedestrian flow levels based on the quantiles of historical pedestrian flow data. For each future time slice, a corresponding contextual feature is determined, and a set of historical time slices that perfectly match this feature is selected from historical data. The proportion of historical time slices belonging to historical high-load periods in the selected set is calculated, and a preset proportion threshold is used to determine whether the future time slice is in a high-load risk state. Specific steps include: Determine the starting time point t0 and the predicted duration S to be evaluated. The predicted duration corresponds to S future time slices. For access control device a, obtain its historical high-load period set P. a ={p a1 p a2 , ..., p ak , ..., p aK } Define the context feature triple C=(W,HT,L) of the device in any time slice, where W is the weekday identifier, HT is the time period identifier, which is the hourly segment number in a day, and L is the traffic level identifier. For each future time slice s, s = 1, 2, ..., S, calculate its absolute time point t. s =t0+s×Δt, where Δt represents the length of a single time slice, and determines the context feature vector C of access control device a within that time slice. as =(W s HT s L s ), where W s and HT s By t s The calendar and time are determined, L s By querying the average pedestrian traffic data of access control device A during the same historical period, and mapping it to a level according to a preset threshold, all devices with the same traffic level as C are filtered out from the historical data. as The set of historical time slices V that satisfy the condition of context similarity s The similarity conditions are: the same W value, the same HT value, and the same L value; if there are no matches with the same L value in the history, the condition is relaxed to the point that the L values differ by no more than one level. For each future time slice t s Calculate the corresponding set of similar historical time slices V s In the middle, it belongs to the set P of historically high load periods. a The proportion of time slices, denoted as ρ s， The calculation formula is as follows: ρ s =N / U, where N is the number of integers satisfying t∈V. s And t∈P a The number of time slices t, where U is the set V. s The total number of time slices, if V s If it is an empty set, then define ρ. s =0, if ρ s If the value is greater than δ, then the access control device a is predicted to be in a high-load risk state in time slice ts, where δ is a preset proportional threshold.
7. The method for predicting access control equipment faults based on multi-source data according to claim 6, characterized in that: The number of time slots during which access control equipment is identified as being in a high-risk state within the predicted duration is statistically analyzed, and the comprehensive fault risk score of the access control equipment within the corresponding predicted duration is calculated. Specific steps include: The number of time slices within a statistical prediction period S in which access control device a is predicted to be in a high-load risk state is denoted as K. a According to the formula Rs a =K a / S; Calculate the comprehensive failure risk score of device a within the predicted future time.
8. The method for predicting access control equipment faults based on multi-source data according to claim 7, characterized in that: Generate a preventative maintenance priority queue, and schedule maintenance personnel to carry out maintenance work based on this queue. Specific steps include: The system iterates through all access control devices within the target area to obtain a comprehensive fault risk score set for each device. Based on this, all devices are sorted in descending order to generate a preventive maintenance priority queue. Maintenance personnel are then assigned to perform maintenance according to the preventive maintenance priority queue.
9. A fault prediction system for access control equipment based on multi-source data, applied to the fault prediction method for access control equipment based on multi-source data as described in any one of claims 1-8, characterized in that: The system includes: a data acquisition module, a parameter calculation module, a model training module, a risk prediction module, a score generation module, and a maintenance scheduling module. The data acquisition module acquires historical operational datasets of access control equipment in the target area, selecting any one device's historical data set. The parameter calculation module calculates the historical operational pressure index for the corresponding time slice of the device, compares it with a preset dynamic health baseline threshold, marks time slices exceeding the threshold, and aggregates them into a historical high-load period set. The model training module builds a device aging sample database and trains a performance degradation prediction model using a supervised learning algorithm. The risk prediction module determines the starting time point and prediction duration for evaluation, defines time slice context features, matches each future time slice with a corresponding set of historical time slices, and judges future high-load risk by calculating the proportion of high-load periods in the set. The score generation module counts the number of high-load risk time slices within the prediction duration and calculates the comprehensive equipment failure risk score. The maintenance scheduling module generates a preventative maintenance priority queue based on the comprehensive failure risk score of each device and assigns maintenance personnel to perform maintenance according to the queue.