A campus equipment operation and maintenance management and control platform based on intelligent control

Abstract

The application discloses a kind of based on intelligent control's campus equipment operation and maintenance management and control platform, including detection module, abnormality analysis module and fault location processing module;Detection module constructs risk vector and spider web coordinate system, according to the spider web area and axis value dispersion degree start balanced or concentrated detection mode;Abnormality analysis module determines abnormal component and labels priority by constructing component data matrix, analyzing time series etc.;Fault location processing module calculates abnormal component correlation, generates fault feature cluster, matches fault type and sorts priority sequence;The platform solves the problems of multi-dimensional risk analysis missing, fault location extensive in traditional operation and maintenance, realizes the accurate assessment and hierarchical detection of equipment risk, the accurate determination of abnormal component, and the topological positioning and scientific scheduling of fault, improves the accuracy and efficiency of campus equipment operation and maintenance, and guarantees the safe and efficient operation of equipment.

Description

Technical Field

[0001] This invention belongs to the field of campus equipment management technology, specifically relating to a campus equipment operation and maintenance management platform based on intelligent control. Background Technology

[0002] As the level of intelligence of campus equipment increases, the operation and maintenance of equipment such as air conditioners and elevators face challenges such as scattered multi-parameter monitoring data and lack of systematic risk assessment. Traditional single-point threshold alarm methods are difficult to capture the linkage effect of multi-dimensional risks such as abnormal operation and accumulated load pressure, resulting in delayed fault location and blind allocation of maintenance resources, which cannot meet the needs of safe and efficient operation of equipment.

[0003] Existing technologies have three major flaws in campus equipment operation and maintenance:

[0004] Insufficient multi-dimensional risk linkage analysis: It is impossible to quantify the correlation between different risk vectors, making it difficult to identify potential systemic failures, and the comprehensive risk assessment lacks accuracy;

[0005] The fault location mechanism is crude: the single parameter threshold judgment is easily affected by noise interference, and it lacks topological clustering analysis of abnormal components, resulting in insufficient accuracy of composite fault location.

[0006] The lack of quantitative prioritization in operation and maintenance decisions: The failure to build a priority system that integrates risk transmission paths and maintenance efficiency leads to blind scheduling of maintenance resources and delayed response to high-risk faults. To address this, we propose a campus equipment operation and maintenance management platform based on intelligent control. Summary of the Invention

[0007] The purpose of this invention is to provide a campus equipment operation and maintenance management platform based on intelligent control, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a campus equipment operation and maintenance management platform based on intelligent control, comprising: a detection module, an anomaly analysis module, and a fault location and processing module;

[0009] Detection module: Constructs risk vectors for campus maintenance equipment, sets detection cycles and analyzes risk vector values ​​at each detection time, constructs a spider web coordinate system to generate a maintenance spider web diagram, and starts the corresponding detection mode based on the spider web area and axis value dispersion;

[0010] Anomaly Analysis Module: Monitors each risk vector according to the priority sequence corresponding to the detection mode; obtains the values ​​of each component within the risk vector during the detection period, constructs a component data matrix and calculates the vector component threshold; determines abnormal differences and the sum of abnormal times by analyzing the component time series and adjacent difference series; calculates the component priority value; and integrates and labels the priority value after judging the abnormal components.

[0011] Fault location processing module: Calculates the correlation coefficient between any two abnormal components, retains component pairs whose absolute correlation values ​​meet the threshold and converts them into connection edges of an undirected graph, and generates fault feature clusters using a fully connected clustering algorithm; constructs a mapping table between fault feature clusters and fault types, substitutes the current feature cluster into the mapping table to match the fault type, calculates the processing priority value corresponding to each fault type, sorts them to form a priority sequence and sends it to the operation and maintenance terminal.

[0012] Preferably, the specific process for constructing a spider web coordinate system to generate an operation and maintenance spider web diagram is as follows:

[0013] Identify the equipment on campus that requires operation and maintenance management, and construct a risk vector for each type of equipment. The risk vector includes: operational anomaly risk vector, load pressure accumulation vector, environmental erosion risk vector, maintenance hidden danger accumulation vector, and lifespan degradation risk vector.

[0014] Each risk vector contains several vector components; a detection period is set, and the values ​​of the vector components at each detection time within the detection period are obtained;

[0015] Based on the logical relationship between the vector components and their corresponding risk vectors, they are divided into positively correlated and negatively correlated components. After normalizing each vector component, the following formula is used: The risk vector value V is obtained; where Xi is the number of the positively correlated component; i = 1, 2, ..., m; m is the total number of positively correlated components; yg is the number of the negatively correlated component; g = 1, 2, ..., k; k is the total number of negatively correlated components; ai and bg are the preset weight coefficients assigned to each positively correlated component and each negatively correlated component, respectively.

[0016] Using the geometric center of the plane as the origin, five directional axes are evenly distributed along the circumference, with each directional axis corresponding to a risk vector. A spider web coordinate system is constructed. For each detection time, the values ​​of each risk vector corresponding to that detection time are normalized and mapped to the corresponding directional axis of the spider web coordinate system to obtain the corresponding value points. Adjacent data points are connected in a clockwise direction to obtain the operation and maintenance spider web diagram.

[0017] Preferably, the specific process of activating the corresponding detection mode based on the dispersion of the spider web area and axis value is as follows:

[0018] Calculate the area of ​​the maintenance spider web at the current detection time to obtain the spider web area ZS; at the same time, calculate the standard deviation of the five directional axis values ​​in the maintenance spider web to obtain the axis value dispersion ZL;

[0019] Set the spider web area threshold SY and the axis value dispersion threshold LY;

[0020] If the detection time ZS ≥ SY and ZL < LY, start the balancing detection mode and perform detections on each direction axis in sequence according to the balancing priority sequence;

[0021] The process of constructing the balancing priority sequence is as follows:

[0022] For each direction axis in the operation and maintenance spider web diagram, analyze the correlation coefficient between the current direction axis and the other direction axes; and calculate the mean value to obtain the average axis correlation degree XG;

[0023] If the value of the direction axis is greater than or equal to the corresponding threshold, it is recorded as an abnormal direction axis. Count the number of times the direction axis is abnormal at the monitoring time within the current period, and record it as the abnormal detection count; by dividing the abnormal detection count by the total number of detection times up to the current time within the current period, obtain the abnormal risk rate FG;

[0024] Obtain the risk vector value V corresponding to each detection time up to the current time within the current detection period; use the formula: Obtain the risk growth rate FZ; where s is the label of the detection time, t = 1, 2,..., T; T is the total number of detection times that have occurred within the current detection period up to the current time;

[0025] After normalizing the average axis correlation degree XG, abnormal risk rate FG, and risk growth rate FZ, use the formula: P

[0025] , = XY × w1 + FG × w2 + FZ × w3 to obtain the balancing priority value P 均衡 ; where w1, w2, and w3 are preset weight coefficients;

[0026] Sort the risk vectors corresponding to each direction axis according to the balancing priority value to obtain the balancing priority sequence. <0..​​​​​​​​​​​​​​

[0032] After normalizing the failure initiation rate GY, average repair time XF, and shaft base ratio ZB, the formula is used: P 集中 =GY×f1+XF×f2+ZB×f3, to obtain the lumped priority value P 集中 Where f1, f2, and f3 are preset weight coefficients;

[0033] By sorting the risk vectors corresponding to each directional axis according to the centralized priority value, a centralized priority sequence is obtained.

[0034] Preferably, the analysis process for adjacent difference sequences is as follows:

[0035] For each risk vector, obtain the values ​​of each vector component within the risk vector at each detection time in this period up to the current detection time;

[0036] Constructing the component data matrix: Where T is the total number of detection moments that occur within the current detection period up to the current moment; n is the total number of components in the risk vector;

[0037] For each vector component, read the historical normal data statistics, namely: mean μd,j and standard deviation σd,j; use the formula: dth,j=μd,j+3σd,j to obtain the vector component threshold dth,j; where j is the label of the vector component, j=1,2,……,n;

[0038] For each vector component, extract the values ​​of the vector component at all times to obtain the component time series: Xj = [x1, j, x2, j, ... xT, j];

[0039] Calculate the numerical difference between adjacent time points: dt,j = |xt,jx(t-1),j|, to obtain the sequence of adjacent component differences: Dj = [d2,j,d3,j,...dT,j]; where the length is T-1;

[0040] Get the mean of the difference sequence The specific calculation process is as follows:

[0041] Preferably, the specific process for determining the sum of abnormal differences and the number of abnormalities, and calculating the component priority value is as follows:

[0042] For each vector component, the clustering degree Cj of the time series of the vector component is obtained. The specific calculation process is as follows: Where exp is the symbol for the natural exponential function. The smaller the clustering degree, the more dispersed the data is and the higher the probability of anomalies.

[0043] Extract the differences that exceed the threshold of the corresponding vector component from the adjacent difference sequence of components: Dj = [d2, j, d3, j, ... dT, j], and mark them as abnormal differences; calculate the sum of abnormal differences to obtain the abnormal difference sum YZj; at the same time, count the total number of abnormal differences to obtain the abnormal number sum YCj.

[0044] Using the formula: The component priority value FYj is obtained; where max(Dj) represents the maximum adjacent difference selected from the adjacent difference sequence of the component; PV is the priority value of the risk vector corresponding to the vector component; q1 and q2 are preset weight coefficients.

[0045] Preferably, the specific process of integrating and labeling priority values ​​after identifying abnormal components is as follows:

[0046] For each vector component, it is determined to be an anomalous component if it meets one of the following conditions:

[0047] The difference between adjacent components H consecutive times: dt,j≥dth,j; where H is the preset number of times;

[0048] Total number of abnormalities:

[0049] All abnormal components are collected, integrated, and labeled with their corresponding priority values ​​before being sent to the fault location and allocation module.

[0050] Preferably, the specific process for generating fault feature clusters is as follows:

[0051] Extract the time series of all abnormal components within the current detection period, and standardize each component sequence.

[0052] The Pearson correlation coefficient method is used to calculate the correlation coefficient between any two outlier components; a correlation coefficient threshold is preset, and outlier component pairs whose absolute correlation values ​​reach the correlation coefficient threshold are retained; these outlier component pairs are transformed into connecting edges in an undirected graph, and each outlier component is used as a node in the graph.

[0053] A fully connected clustering algorithm is used. Initially, each outlier component is treated as a separate cluster. Then, the maximum correlation between all cluster pairs is calculated. Cluster pairs whose maximum correlation reaches the correlation coefficient threshold are merged.

[0054] When any cluster pair is merged, clustering is stopped if the correlation of all abnormal component pairs in the new cluster is not lower than the correlation coefficient threshold, or if there are no cluster pairs that can be merged, and several fault feature clusters are formed.

[0055] Preferably, the specific process of calculating the processing priority value corresponding to each fault type, sorting them into a priority sequence, and then sending it to the operation and maintenance terminal is as follows:

[0056] Construct a mapping table between fault feature clusters and campus equipment fault types; then substitute the fault feature clusters that appeared in this test into the mapping table for matching, and output the corresponding fault types.

[0057] For each fault type, the average priority value of each abnormal component in the fault feature cluster corresponding to the fault type is calculated to obtain the fault type processing priority value.

[0058] All fault types of campus equipment are sorted according to their corresponding fault type handling priority values ​​to obtain a fault type handling priority sequence.

[0059] The priority sequence of fault types for campus equipment is sent to the maintenance terminal. The maintenance terminal dispatchers then handle each fault type of the campus equipment in sequence according to the priority sequence.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] (1) This campus equipment operation and maintenance management platform based on intelligent control constructs a spider web coordinate system containing various risk vectors, transforms the equipment status into a visual spider web diagram, judges whether there are abnormal risks in the equipment by spider web area and axis value dispersion, and dynamically starts balanced detection or centralized detection mode according to risk distribution characteristics; the balanced mode identifies risk transmission hubs by parameters such as axis average correlation and abnormal risk rate, and blocks multi-dimensional risk linkage; the centralized mode locks single high-risk axes based on fault occurrence rate and repair time, realizes accurate allocation of detection resources, and improves the comprehensiveness of equipment risk assessment and the adaptability of detection strategy.

[0062] (2) This campus equipment operation and maintenance management platform based on intelligent control achieves accurate identification of abnormal fluctuations by constructing a component data matrix, calculating vector component thresholds, and combining component time series and adjacent difference sequence analysis; it introduces clustering calculation to quantify the degree of data dispersion, and combines abnormal difference sum, abnormal number sum and continuous abnormal conditions to determine abnormal components in multiple dimensions; it integrates the difference sequence characteristics and corresponding risk vector priority values, calculates component priority values ​​and labels them, providing a reliable basis for subsequent fault location, and improving the accuracy of abnormal detection and the scientific nature of priority assessment.

[0063] (3) This campus equipment operation and maintenance management platform based on intelligent control uses Pearson correlation coefficient to screen highly correlated abnormal component pairs, constructs an undirected graph and generates fault feature clusters through fully connected clustering to ensure the homogeneity of fault components within the cluster; it achieves fault type matching based on fault feature clusters and fault type mapping table, and generates a fault type processing priority sequence through mean calculation to drive operation and maintenance terminals to dispatch orders according to priority; this mechanism realizes topological positioning from abnormal components to fault types and scientific scheduling of operation and maintenance decisions, improving fault positioning accuracy and maintenance resource utilization efficiency. Attached Figure Description

[0064] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0065] 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.

[0066] Example 1

[0067] Please see Figure 1 This invention provides a campus equipment operation and maintenance management platform based on intelligent control, including: a detection module, an anomaly analysis module, and a fault location and processing module;

[0068] The detection module constructs risk vectors for campus maintenance equipment, sets detection cycles, analyzes risk vector values ​​at each detection time, builds a spider web coordinate system to generate a maintenance spider web diagram, and initiates different detection modes based on the spider web area and axis value dispersion to achieve accurate assessment and graded detection of equipment risks. The specific process is as follows:

[0069] Identify the equipment on campus that requires operation and maintenance management, and construct a risk vector for each type of equipment. The risk vector includes: operational anomaly risk vector, load pressure accumulation vector, environmental erosion risk vector, maintenance hidden danger accumulation vector, and lifespan degradation risk vector.

[0070] The operational anomaly risk vector includes: current harmonic distortion rate, vibration phase difference in different parts of the equipment, and exceeding the limit of rotational speed fluctuation rate, etc.

[0071] The cumulative load pressure vector includes: load entropy value, peak load duration, number of load cycles within a certain period, and overload percentage, etc.

[0072] The environmental erosion risk vector includes: environmental corrosivity equivalent, electromagnetic interference intensity index, dust deposition rate, and the hazard value of temperature and humidity fluctuations, etc.

[0073] The maintenance hazard accumulation vector includes: maintenance intervention complexity, fault recurrence risk index, maintenance plan lag rate, etc.;

[0074] The life attenuation risk vector includes: material micro-damage degree, remaining life warning value, aging rate of key components, etc.;

[0075] For each risk vector, there are several vector components contained; set the detection period, and obtain the values of the vector components at each detection moment within the detection period;

[0076] According to the logical relationship between the vector components and the corresponding risk vectors; divide them into positive correlation components and negative correlation components; after normalizing each vector component, use the formula: Obtain the risk vector value V; where Xi is the number of the positive correlation component; i = 1, 2,..., m; m is the total number of positive correlation components; yg is the number of the negative correlation component; g = 1, 2,..., k; k is the total number of negative correlation components; ai and bg are the preset weight coefficients assigned to each positive correlation component and negative correlation component respectively;

[0077] Taking the plane geometric center as the origin, evenly distribute five direction axes along the circumference, and each direction axis corresponds to a risk vector. Construct a cobweb coordinate system. For each detection moment, after normalizing the values of the corresponding risk vectors at this detection moment, map them to the corresponding direction axes in the cobweb coordinate system to obtain the corresponding numerical points. Connect the adjacent data points in the clockwise direction to obtain the operation and maintenance cobweb diagram;

[0078] Calculate the area of the operation and maintenance cobweb diagram at the current detection moment to obtain the cobweb area ZS; at the same time, calculate the standard deviation of the values of the five direction axes in the operation and maintenance cobweb diagram to obtain the axis value dispersion ZL;

[0079] Set the cobweb area threshold SY and the axis value dispersion threshold LY;

[0080] If ZS ≥ SY and ZL < LY at the detection moment, start the balanced detection mode, and perform detections on each direction axis in turn according to the balanced priority sequence;

[0081] The process of constructing the balanced priority sequence is:

[0082] For each direction axis in the operation and maintenance cobweb diagram, use the Pearson correlation coefficient algorithm to analyze the correlation coefficient between the current direction axis and the other direction axes; and perform mean calculation to obtain the axis average correlation degree XG;

[0083] Set the numerical threshold of the direction axis. If the value of the direction axis is greater than or equal to the corresponding threshold, it is recorded as an abnormal direction axis. Count the number of times the direction axis is abnormal at the monitoring time within the current cycle, and record it as the abnormal detection count. Divide the abnormal detection count by the total number of detection times up to the current time within the current cycle to obtain the abnormal risk rate FG.

[0084] Obtain the risk vector value V corresponding to each detection time within the current detection cycle up to the current time. Use the formula: Get the risk growth rate FZ; where s is the label of the detection time, t = 1, 2,..., T; T is the total number of detection times that have occurred within the current detection cycle up to the current time.

[0085] After normalizing the axis average correlation degree XG, the abnormal risk rate FG, and the risk growth rate FZ, use the formula: P 均衡 = XG × w1 + FG × w2 + FZ × w3 to obtain the balanced priority value P 均衡 ; where w1, w2, and w3 are preset weight coefficients.

[0086] Sort the risk vectors corresponding to each direction axis according to the balanced priority value to obtain the balanced priority sequence.

[0087] It should be noted that if the detection time ZS ≥ SY and ZL < LY, it means that the comprehensive risk of the equipment is high but the distribution is balanced, and the risks of each direction axis deteriorate synergistically, which may cause systematic hidden dangers due to equipment aging or environmental factors. Calculating the priority using the axis average correlation degree, abnormal risk rate, and risk growth rate can identify the risk conduction hub axis, lock the high-frequency abnormal direction axis, warn of the accelerating deterioration risk, and block the systematic failure from the risk network level.

[0088] If the detection time ZS ≥ SY and ZL ≥ LY, start the centralized detection mode; and detect the vector components corresponding to each direction axis according to the centralized priority sequence.

[0089] The construction process of the centralized priority sequence is as follows:

[0090] For each direction axis in the operation and maintenance cobweb diagram, count the number of historical direction axis abnormalities and the number of times of equipment failures caused after the direction axis abnormalities; and divide the number of times of equipment failures caused after the historical direction axis abnormalities by the number of historical direction axis abnormalities to obtain the failure initiation rate GY.

[0091] Calculate the average value of the repair duration each time after a failure is caused by a direction axis abnormality to obtain the average repair duration XF.

[0092] Preset the reference value of the direction axis. Divide the current direction axis value by the corresponding reference value to obtain the axis base ratio ZB.

[0093] After normalizing the failure initiation rate GY, average repair time XF, and shaft base ratio ZB, the formula is used: P 集中 =GY×f1+XF×f2+ZB×f3, to obtain the lumped priority value P 集中 Where f1, f2, and f3 are preset weight coefficients;

[0094] By sorting the risk vectors corresponding to each directional axis according to the centralized priority value, a centralized priority sequence is obtained.

[0095] It should be noted that if the detection time ZS≥SY and ZL≥LY, it indicates that the overall risk of the equipment is high and concentrated on individual axes. The single point of failure dominates the overall risk and is prone to triggering a chain of failures. By using failure initiation rate, average repair time and axis-to-base ratio to calculate priority, high-risk axes that deviate from the baseline can be accurately identified. Resources can be optimized according to failure probability and repair efficiency to curb the spread of failures from the perspective of single point breakthrough.

[0096] Based on two detection models and priority parameters, precise operation and maintenance can be achieved: the balanced mode locates systemic risk hubs by correlation, anomaly rate and growth rate, and blocks multi-dimensional risk linkages; the centralized mode uses fault initiation rate, repair time and axis-to-base ratio to lock single high-risk axes and allocate resources according to the urgency of risk; the combination of the two can greatly improve detection efficiency and ensure the safe and efficient operation of campus equipment.

[0097] The anomaly analysis module monitors campus equipment operating in detection mode according to the priority sequence corresponding to the detection mode. It acquires the values ​​of each component within the risk vector during the detection period, constructs a component data matrix, calculates vector component thresholds, and determines abnormal differences and anomaly counts by analyzing component time series and adjacent difference series. It then calculates component priority values, identifies anomalous components, integrates them, and labels their priority values. The specific process is as follows:

[0098] For campus equipment operating in monitoring mode, monitoring is conducted sequentially on each risk vector in the equipment maintenance spider diagram according to the priority sequence corresponding to the monitoring mode activated by the equipment. The specific process is as follows:

[0099] For each risk vector, obtain the values ​​of each vector component within the risk vector at each detection time in this period up to the current detection time;

[0100] Constructing the component data matrix: Where T is the total number of detection moments that occur within the current detection period up to the current moment; n is the total number of components in the risk vector;

[0101] For each vector component, read the historical normal data statistics, namely: mean μd,j and standard deviation σd,j; use the formula: dth,j=μd,j+3σd,j to obtain the vector component threshold dth,j; where j is the label of the vector component, j=1,2,……,n;

[0102] For each vector component, perform the following calculations independently:

[0103] Extract the values ​​of the vector components at all times to obtain the component time series: Xj = [x1, j, x2, j, ..., xT, j];

[0104] Calculate the numerical difference between adjacent time points: dt,j=|xt,jx(t-1),j|, to obtain the sequence of adjacent component differences: Dj=[d2,j,d3,j,...dT,j]; where the length is T-1;

[0105] Get the mean of the difference sequence The specific calculation process is as follows:

[0106] The specific calculation process for obtaining the clustering degree Cj of the time series of vector components is as follows: Where exp is the symbol for the natural exponential function. The smaller the clustering degree, the more dispersed the data is and the higher the probability of anomalies.

[0107] Extract the differences that exceed the threshold of the corresponding vector component from the adjacent difference sequence of components: Dj=[d2,j,d3,j,...dT,j], and mark them as abnormal differences; calculate the sum of abnormal differences to obtain the abnormal difference sum YZj; at the same time, count the total number of abnormal differences to obtain the abnormal number sum YCj.

[0108] Using the formula: The component priority value FYj is obtained; where max(Dj) represents the maximum adjacent difference selected from the adjacent difference sequence of the component; PV is the priority value of the risk vector corresponding to the vector component; q1 and q2 are preset weight coefficients;

[0109] For each vector component, it is determined to be an anomalous component if it meets one of the following conditions:

[0110] The difference between adjacent components H consecutive times: dt,j≥dth,j; where H is the preset number of times;

[0111] Total number of abnormalities: (The number of anomalies exceeds one-third of the detection time);

[0112] All abnormal components are collected, integrated, and labeled with their corresponding priority values ​​before being sent to the fault location and allocation module.

[0113] It should be noted that the above process, by constructing a component data matrix and combining it with historical normal data statistics to calculate vector component thresholds, can accurately identify abnormal fluctuations. Analysis of component time series and adjacent difference series can capture the dynamic characteristics of data changes, and clustering calculation can quantify the degree of data dispersion, providing multi-dimensional basis for anomaly judgment. The calculation of the sum of abnormal differences and the sum of abnormal occurrences, combined with the condition of consecutive abnormal occurrences, can effectively distinguish between random fluctuations and real faults. The calculation of component priority values ​​integrates the characteristics of the difference series and the corresponding risk vector priority values, making the priority assessment of abnormal components more comprehensive. This process achieves accurate judgment and priority labeling of abnormal components, providing a reliable basis for subsequent fault location, improving the accuracy and efficiency of campus equipment operation and maintenance, enabling timely detection of potential equipment faults, preventing fault escalation, and ensuring the safe and efficient operation of equipment.

[0114] The fault location and processing module calculates the correlation coefficient between any two detected abnormal components, retains component pairs whose absolute correlation values ​​meet a threshold, and converts them into connection edges in an undirected graph. A fully connected clustering algorithm is then used to generate fault feature clusters. A mapping table between fault feature clusters and fault types is constructed. The current feature cluster is substituted into the mapping table to match the fault type, and the processing priority value corresponding to each fault type is calculated. These are then sorted to form a priority sequence and sent to the operation and maintenance terminal. The specific process is as follows:

[0115] Extract the time series of all abnormal components within the current detection period, and standardize each component sequence.

[0116] The Pearson correlation coefficient method is used to calculate the correlation coefficient between any two outlier components; a correlation coefficient threshold is preset, and outlier component pairs whose absolute correlation values ​​reach the correlation coefficient threshold are retained; these outlier component pairs are transformed into connecting edges in an undirected graph, and each outlier component is used as a node in the graph.

[0117] A fully connected clustering algorithm is used. Initially, each outlier component is treated as a separate cluster. Then, the maximum correlation between all cluster pairs is calculated. Cluster pairs whose maximum correlation reaches the correlation coefficient threshold are merged.

[0118] When any cluster pair is merged, clustering is stopped if the correlation of all abnormal component pairs in the new cluster is not lower than the correlation coefficient threshold, or if there are no cluster pairs that can be merged, and several fault feature clusters are formed.

[0119] Construct a mapping table between fault feature clusters and campus equipment fault types; then substitute the fault feature clusters that appeared in this test into the mapping table for matching, and output the corresponding fault types.

[0120] For each fault type, the average priority value of each abnormal component in the fault feature cluster corresponding to the fault type is calculated to obtain the fault type processing priority value.

[0121] All fault types of campus equipment are sorted according to their corresponding fault type handling priority values ​​to obtain a fault type handling priority sequence.

[0122] The priority sequence of fault types for campus equipment is sent to the maintenance terminal. The maintenance terminal dispatchers then handle each fault type of the campus equipment in sequence according to the priority sequence.

[0123] It should be noted that the Pearson correlation coefficient is used to quantify the linear correlation between anomalous components, and highly correlated component pairs are selected to construct an undirected graph, providing topological support for fully connected clustering. Fully connected clustering uses the maximum correlation between clusters as the merging criterion to ensure strong correlation among components within a cluster, and the generated fault feature clusters can accurately reflect the common origin of faults. Mapping table matching realizes the transformation from feature clusters to fault types, and the processing priority value of fault types is calculated and sorted by mean, forming a scientific processing priority sequence. This process realizes accurate positioning and priority ranking from anomalous components to fault types, providing a basis for dispatching parts to maintenance terminals, improving the efficiency and pertinence of campus equipment fault handling, and ensuring the orderliness and effectiveness of equipment operation and maintenance.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A campus equipment operation and maintenance management platform based on intelligent control, comprising: Detection module, anomaly analysis module, and fault location and handling module, characterized in that: Detection module: Construct a risk vector for campus operation and maintenance equipment, set a detection period, analyze the risk vector values at each detection moment, construct a cobweb coordinate system to generate an operation and maintenance cobweb diagram, and start the corresponding detection mode according to the cobweb area and the axis value dispersion degree; Anomaly analysis module: Monitor each risk vector according to the priority sequence corresponding to the detection mode; Obtain the values of each component in the risk vector within the detection period, construct a component data matrix and calculate the vector component threshold, determine the abnormal difference and the sum of abnormal times by analyzing the component time series and the adjacent difference sequence, analyze the component priority value based on the abnormal difference, and integrate the abnormal components and mark their priority values after determining the abnormal components according to the determination conditions of the abnormal difference and the sum of abnormal times; The specific process of analyzing the component priority value based on the abnormal difference is as follows: For each vector component, the clustering degree Cj of the time series of the vector component is obtained. The specific calculation process is as follows: Where exp is the symbol for the natural exponential function; Let be the mean of the adjacent difference sequences corresponding to the vector components. Let be the historical normal data standard deviation of the j-th vector component, where j is the vector component label; Extract the differences exceeding the corresponding vector component threshold from the component adjacent difference sequence and mark them as abnormal differences; Calculate the sum of abnormal differences to obtain the sum of abnormal differences YZj; At the same time, count the total number of occurrences of abnormal differences to obtain the sum of abnormal times YCj; Using the formula: The component priority value FYj is obtained; where max(Dj) represents the maximum adjacent difference selected from the adjacent difference sequence of the components; PV is the priority value of the risk vector corresponding to the vector component; q1 and q2 are preset weight coefficients; Fault location and handling module: Calculate the correlation coefficient between any two abnormal components, retain the component pairs whose absolute value of the correlation meets the threshold and convert them into the connecting edges of an undirected graph, and use the fully connected clustering algorithm to generate fault feature clusters; Construct a mapping table between the fault feature clusters and the fault types, substitute the current feature cluster into the mapping table to match the fault type, calculate the processing priority value corresponding to each fault type, sort to form a priority sequence and send it to the operation and maintenance terminal.

2. The campus equipment operation and maintenance management platform based on intelligent control according to claim 1, characterized in that: The specific process of constructing a cobweb coordinate system to generate an operation and maintenance cobweb diagram is as follows: Obtain the equipment that needs to be operation and maintenance controlled on campus. For each type of equipment, construct a risk vector, which includes: operation anomaly risk vector, load pressure accumulation vector, environmental erosion risk vector, maintenance hidden danger accumulation vector, and life attenuation risk vector; For each risk vector, it contains several vector components; Set a detection period and obtain the values of the vector components at each detection moment within the detection period; Based on the logical relationship between the vector components and their corresponding risk vectors, they are divided into positively correlated and negatively correlated components. After normalizing each vector component, the following formula is used: We obtain the risk vector value V; where Xi is the number of the positively correlated component; i = 1, 2, ..., m; m is the total number of positively correlated components; yg is the number of the negatively correlated component; g = 1, 2, ..., k; k is the total number of negatively correlated components; is the minimum value of the g-th negative correlation component in the historical normal operation data; ai and bg are the preset weight coefficients assigned to each positive and negative correlation component, respectively; Take the plane geometric center as the origin, evenly distribute five direction axes along the circumference, each direction axis corresponds to a risk vector, construct a cobweb coordinate system, and after normalizing the values of each risk vector corresponding to the detection moment, map them to the corresponding direction axes in the cobweb coordinate system to obtain the corresponding numerical points, and connect the adjacent data points in a clockwise direction to obtain the operation and maintenance cobweb diagram.

3. The campus equipment operation and maintenance management platform based on intelligent control according to claim 2, characterized in that: The specific process of starting the corresponding detection mode according to the cobweb area and the axis value dispersion degree is as follows: Calculate the area of the operation and maintenance cobweb diagram at the current detection moment to obtain the cobweb area ZS; At the same time, calculate the standard deviation of the values of the five direction axes in the operation and maintenance cobweb diagram to obtain the axis value dispersion degree ZL; Set the cobweb area threshold SY and the axis value dispersion degree threshold LY; If ZS≥SY and ZL<LY at the detection moment, start the balanced detection mode and perform detection on each direction axis in turn according to the balanced priority sequence; The process of constructing the balanced priority sequence is as follows: For each directional axis in the maintenance spider diagram, analyze the correlation coefficient between the current directional axis and the other directional axes; and calculate the mean to obtain the axis-average correlation coefficient XG. If the value of the direction axis is greater than or equal to the corresponding threshold, it is recorded as a direction axis anomaly. The number of times the direction axis is abnormal at the monitoring time within the current period is counted and recorded as the number of anomaly detections. The anomaly risk rate FG is obtained by dividing the number of anomaly detections by the total number of detection times up to the present in the current period. Obtain the risk vector value V corresponding to each detection time up to the current detection time within the current detection period; using the formula: The risk growth rate FZ is obtained; where t is the index of the detection time, t=1,2,...,T; T is the total number of detection times that have occurred in the current detection period up to the current time. After normalizing the axis-mean correlation XG, the anomaly risk rate FG, and the risk growth rate FZ, the formula is used: To obtain the equilibrium priority value ; Where w1, w2, and w3 are preset weight coefficients; By sorting the risk vectors corresponding to each directional axis according to the equilibrium priority value, an equilibrium priority sequence is obtained.

4. The campus equipment operation and maintenance management platform based on intelligent control according to claim 3, characterized in that: If the detection time ZS≥SY and ZL≥LY, the centralized detection mode is activated; and detection is carried out on each directional axis according to the centralized priority sequence. The process of constructing a centralized priority sequence is as follows: For each directional axis in the maintenance spider diagram, the number of historical directional axis anomalies and the number of equipment failures caused by directional axis anomalies are counted; and the failure initiation rate GY is obtained by dividing the number of equipment failures caused by historical directional axis anomalies by the number of historical directional axis anomalies. The average repair time XF is obtained by calculating the average repair time after each fault caused by the abnormality of the steering axis. The axis-base ratio ZB is obtained by dividing the current directional axis value by the corresponding reference value. After normalizing the failure initiation rate GY, average repair time XF, and shaft base ratio ZB, the following formula is used: To obtain the central priority value Where f1, f2, and f3 are preset weighting coefficients; By sorting the risk vectors corresponding to each directional axis according to the centralized priority value, a centralized priority sequence is obtained.

5. A campus equipment operation and maintenance management platform based on intelligent control according to claim 4, characterized in that: The analysis process for adjacent difference sequences is as follows: For each risk vector, obtain the values ​​of each vector component within the risk vector at each detection time in this period up to the current detection time, and construct a component data matrix; For each vector component, read the historical normal data statistics, i.e., the mean. and standard deviation Using the formula: , to obtain the vector component threshold ; where j is the label of the vector component, j=1,2,...,n; For each vector component, the values ​​of the vector component in the component data matrix at all detection times are extracted to obtain the component time series; Calculate the numerical difference between adjacent time points to obtain the sequence of adjacent component differences; The mean of the difference sequence is obtained by calculating the mean of the adjacent difference sequences of the components. .

6. The campus equipment operation and maintenance management platform based on intelligent control according to claim 5, characterized in that: After determining the outlier components by combining the criteria of outlier difference and outlier frequency, the specific process of integrating the outlier components and labeling their priority values ​​is as follows: For each vector component, it is determined to be an anomalous component if it meets one of the following conditions: The difference between adjacent components of number H consecutive times: Where H is the preset number of times, The difference between adjacent components; Total number of abnormalities: ; All abnormal components are collected, integrated, and labeled with their corresponding priority values ​​before being sent to the fault location and allocation module.

7. A campus equipment operation and maintenance management platform based on intelligent control according to claim 6, characterized in that: The specific process for generating fault feature clusters is as follows: Extract the time series of all abnormal components within the current detection period, and standardize each component sequence. The Pearson correlation coefficient method is used to calculate the correlation coefficient between any two outlier components; a correlation coefficient threshold is preset, and outlier component pairs whose absolute correlation values ​​reach the correlation coefficient threshold are retained; these outlier component pairs are transformed into connecting edges in an undirected graph, and each outlier component is used as a node in the graph. A fully connected clustering algorithm is used, initially treating each outlier component as a separate cluster, and then calculating the maximum correlation between all cluster pairs; Merge cluster pairs whose maximum correlation reaches the correlation coefficient threshold; When any cluster pair is merged, clustering is stopped if the correlation of all abnormal component pairs in the new cluster is not lower than the correlation coefficient threshold, or if there are no cluster pairs that can be merged, and several fault feature clusters are formed.

8. A campus equipment operation and maintenance management platform based on intelligent control according to claim 7, characterized in that: The specific process of calculating the processing priority value corresponding to each fault type, sorting them into a priority sequence, and then sending it to the operation and maintenance terminal is as follows: Construct a mapping table between fault feature clusters and campus equipment fault types; then substitute the fault feature clusters that appeared in this test into the mapping table for matching, and output the corresponding fault types. For each fault type, the average priority value of each abnormal component in the fault feature cluster corresponding to the fault type is calculated to obtain the fault type processing priority value. All fault types of campus equipment are sorted according to their corresponding fault type handling priority values ​​to obtain a fault type handling priority sequence. The priority sequence of fault types for campus equipment is sent to the maintenance terminal. The maintenance terminal dispatchers then handle each fault type of the campus equipment in sequence according to the priority sequence.

Images