Petrochemical area inspection system based on inspection robot

By adopting path optimization and dynamic redirection modules in the petrochemical regional inspection system, combining neural network model and time series analysis, the problems of insufficient flexibility of inspection paths and insane data processing in the existing technology are solved, and more efficient and safety inspection and failure prediction.

CN118656734BActive Publication Date: 2024-11-05TANJIE DIGITAL TECH (SHANDONG) GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411146584.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-11-05
Estimated Expiration
2044-08-21

AI Technical Summary

Technical Problem

The existing technology lacks flexibility and adaptability in petrochemical regional inspections, and cannot respond to emergencies or environmental changes in a timely manner, resulting in equipment damage or production interruption risks. Data processing and analysis lacks automation and intelligence, affecting the accuracy of fault prediction. and efficiency.

Method used

Adopt the petrochemical regional patrol system based on inspection robots, and through the path optimization module, time series analysis module and dynamic redirection module, combined with neural network model and time series analysis, the patrol path is dynamically adjusted to optimize node dependence and data flowability. Indicators, real-time monitoring of data abnormalities, improve inspection efficiency and security.

Benefits of technology

Dynamic optimization of inspection paths and abnormal monitoring are realized, inspection efficiency and safety are improved, potential failures can be identified and prevented in a timely manner, and production safety and continuous operation are ensured.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118656734B_ABST
    Figure CN118656734B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of robot technology, specifically to a petrochemical area inspection system based on an inspection robot, the system comprising: a path optimization module based on real-time collection of sensor data between inspection points, analysis of data between multiple inspection points, integration of node dependencies and data mobility indicators, and calculation of the interaction strength of each node through a neural network model. In the present invention, by dynamically integrating and analyzing sensor data to calculate the interaction strength and dependency between nodes, the inspection path configuration is optimized, the inspection efficiency and safety are improved, and the understanding of data patterns is enhanced through time series analysis and clustering, allowing more accurate identification and classification of equipment status, early detection and prevention of potential faults, dynamic redirection applications adjust the inspection path according to real-time data and deviation analysis, can respond to changes immediately, maintain efficient operation, and abnormal mode monitoring can identify abnormal conditions in a timely manner by setting thresholds and performing data anomaly detection, effectively preventing problems from expanding.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of robot technology, and in particular to a petrochemical area inspection system based on an inspection robot. Background Art

[0002] Existing technologies mainly rely on fixed patterns and preset paths for inspections, which lack flexibility and adaptability and cannot respond to emergencies or environmental changes in a timely manner. In complex and dynamically changing production environments, the limitations of this approach are obvious. For example, when equipment fails suddenly or environmental conditions change, the system may not be able to adjust the inspection route in real time, and the response is slow, which may increase the risk of equipment damage or production interruption. Existing technologies generally lack automation and intelligence in data processing and analysis, making it difficult to extract useful information from large amounts of data, affecting the accuracy and efficiency of fault prediction. These deficiencies may lead to major safety accidents or a serious decline in production efficiency in emergency situations, causing economic losses and safety hazards to enterprises. Summary of the invention

[0003] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a petrochemical area inspection system based on an inspection robot.

[0004] In order to achieve the above purpose, the present invention adopts the following technical scheme: The petrochemical area inspection system based on the inspection robot includes:

[0005] The path optimization module is based on the real-time collection of sensor data between inspection points, analyzes data between multiple inspection points, integrates node dependency and data mobility indicators, calculates the interaction strength of each node through a neural network model, determines the dependency graph, and obtains the optimal path configuration;

[0006] The time series analysis module divides the time windows based on the input of the optimal path configuration, performs time series division of the inspection data, determines the fitting degree of the multi-time window data through cross entropy calculation, classifies and clusters the time series, and obtains the cluster recognition result;

[0007] The dynamic redirection module receives real-time sensor data, analyzes deviations from the cluster identification results, reorders emergency checkpoints on the inspection path according to the degree of deviation, adjusts the path to prioritize access to the most critical areas, and establishes a dynamically updated path;

[0008] The abnormal mode monitoring module monitors the data collection process of the inspection robot according to the dynamic update path, sets the abnormal threshold and performs data anomaly detection, identifies the abnormal time series segment, and performs trend analysis to generate abnormal monitoring results.

[0009] As a further solution of the present invention, the steps of obtaining the node dependency and data liquidity index are specifically as follows:

[0010] Perform time series analysis on the sensor data at each checkpoint and calculate the standard deviation using statistical methods and mean , using the formula:

[0011] and , generate a list of standard deviations and means;

[0012] in, Representation Node The volatility of data, Representation Node The average data traffic is the total time unit, Is a node In time The data value of , is the adjustment factor;

[0013] Based on the standard deviation and mean lists, the data flow dependency between nodes is evaluated using the formula: , generating the dependency matrix ;

[0014] in, Representation Node and nodes The strength of the dependence between , is the mean result, , is the standard deviation result, To ensure numerical stability, is the intensity adjustment coefficient;

[0015] The data mobility of the nodes is analyzed according to the dependency matrix and the mean list, using the formula: , calculate the data mobility index of each node, and obtain the node dependency and data mobility index;

[0016] in, Is a node Data liquidity indicators, is the nonlinear adjustment coefficient.

[0017] As a further solution of the present invention, the step of obtaining the optimal path configuration is specifically as follows:

[0018] According to the data liquidity index results, the connection cost between all nodes is evaluated using the formula: , generate the connection cost matrix;

[0019] in, Representation Node and nodes The connection cost between , It is a multi-node data liquidity indicator. , is the parameter that regulates the influence of the node, is the overall moderating factor;

[0020] Using the connection cost matrix, combined with the node dependency matrix, the benefit of each path is calculated using the formula: , generate the benefit matrix;

[0021] in, Representation Node To Node The total benefit of the path, is the dependency strength between nodes, It is to avoid adding a small amount of zero. , is a nonlinear parameter that adjusts the benefit calculation;

[0022] According to the benefit matrix, the shortest path algorithm is used to capture the path with the lowest cost and the highest benefit, using the formula: , determine the optimal path configuration;

[0023] in, represents the optimal path, It is a parameter that regulates the impact of benefits in path selection.

[0024] As a further solution of the present invention, the step of obtaining the cluster recognition result is specifically:

[0025] Based on the input of the optimal path configuration, the time window is divided with reference to the inspection data characteristics, and the formula is used: , generate time window division results;

[0026] in, Represents the total time range, Represents the number of windows, is the tuning parameter of the path configuration, Represents the length of each time window;

[0027] Time series partitioning is performed on the data in each of the time window partitioning results, using the formula: , generate time series partition results;

[0028] in, Representative The sum of data in time windows, Represents the time window Time point The data value of

[0029] Based on the time series partitioning results, the cross entropy is applied to determine the fitting degree of the multi-time window data, using the formula: , generate the goodness-of-fit result;

[0030] in, Representative Window The degree of fit, represents the probability of the actual distribution, represents the probability of the expected distribution, represents the length of each time window, is a correction term to refer to the deviation of the data, is the weight factor;

[0031] Based on the fit, the time series are classified and clustered using the formula: , get the cluster recognition result;

[0032] in, Representative data points The clustering result of is the cluster center, is the within-class standard deviation, It is a small amount to avoid division by zero.

[0033] As a further solution of the present invention, the step of acquiring the dynamic update path is specifically as follows:

[0034] Receive real-time sensor data and analyze the deviation from the cluster recognition result using the formula: , calculate the weighted deviation of each checkpoint from the cluster center and generate the deviation measurement result;

[0035] in, is the deviation measurement result, Represents real-time data, Representative time The weight factor of Represents total time;

[0036] The priority of the emergency checkpoints of the inspection path is re-ordered according to the deviation measurement results, using the sorting formula: , ensure that high deviation points get high priority and generate reordered urgent checkpoints;

[0037] in, represents an urgent checkpoint for reordering, represents a decimal number that avoids division by zero, Representative area The importance coefficient of

[0038] Based on the reordered checkpoints, the inspection path is adjusted to prioritize access to the critical areas, applying the path adjustment formula: ,Using priority sorting to ensure that key areas are visited first and generate dynamically updated paths;

[0039] in, Represents the original path, Represents a dynamically updated path.

[0040] As a further solution of the present invention, the step of obtaining the abnormal monitoring result is specifically:

[0041] According to the dynamic update path, the data collection process of the inspection robot is monitored using the formula: , generate data collection results;

[0042] in, Represents the data collection results, Represents at time By path Sensors inside Data collected, represents the sensor weight, represents the data validity coefficient;

[0043] Based on the data collection results, anomaly thresholds are set and data anomaly detection is performed using the formula: , generate anomaly detection results;

[0044] in, represents the anomaly detection result, , Respectively represent The mean and standard deviation of represents the time decay coefficient;

[0045] Based on the anomaly detection results, the abnormal time series segments are identified and trend analysis is performed using the formula: , generate abnormal monitoring results;

[0046] in, Represents abnormal monitoring results, represent The standard deviation of Represents the time adjustment parameter.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by dynamically integrating and analyzing sensor data to calculate the interaction intensity and dependencies between nodes, the inspection path configuration is optimized, the inspection efficiency and safety are improved, the inspection is more targeted, and potential safety hazards or inefficient areas can be dealt with first. Through time series analysis and clustering, the understanding of data patterns is enhanced, allowing for more accurate identification and classification of equipment status, and early detection and prevention of potential failures. Dynamic redirection applications adjust inspection paths based on real-time data and deviation analysis, and can respond to changes immediately to maintain efficient operation. Abnormal pattern monitoring identifies abnormal conditions in a timely manner by setting thresholds and performing data anomaly detection, effectively preventing problems from expanding, and ensuring production safety and continuous operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a system flow chart of the present invention;

[0050] Figure 2 A flowchart of the steps for obtaining node dependency and data mobility indicators of the present invention;

[0051] Figure 3 A flowchart of the steps for obtaining the optimal path configuration of the present invention;

[0052] Figure 4 A flow chart of steps for obtaining cluster recognition results of the present invention;

[0053] Figure 5 A flowchart of the acquisition steps of the dynamic update path of the present invention;

[0054] Figure 6 The figure is a flow chart of the steps for obtaining abnormal monitoring results of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0057] Example 1 Please refer to Figure 1 The petrochemical area inspection system based on inspection robots includes:

[0058] The path optimization module is based on the real-time collection of sensor data between inspection points, analyzes data between multiple inspection points, integrates node dependency and data mobility indicators, calculates the interaction strength of each node through a neural network model, determines the dependency graph, and obtains the optimal path configuration;

[0059] The time series analysis module divides the time windows based on the input of the optimal path configuration, performs time series division of the inspection data, determines the fit of the multi-time window data through cross entropy calculation, classifies and clusters the time series, and obtains the cluster recognition results;

[0060] The dynamic redirection module receives real-time sensor data, analyzes the deviation from the cluster identification results, reorders the emergency checkpoints of the inspection path according to the degree of deviation, adjusts the path to prioritize the most critical areas, and establishes a dynamically updated path;

[0061] The abnormal pattern monitoring module monitors the data collection process of the inspection robot according to the dynamically updated path, sets the abnormal threshold and performs data anomaly detection, identifies the abnormal time series segment, and performs trend analysis to generate abnormal monitoring results.

[0062] The optimal path configuration specifically includes the interaction strength and dependency graph of the nodes. The cluster identification results specifically include the time window fitting degree, classification index, and clustering index. The dynamically updated path specifically refers to the order of emergency checkpoints and the priority of key areas. The abnormal monitoring results specifically include the abnormal threshold, abnormal time series, and trend analysis results.

[0063] See also Figure 2 ,The specific steps for obtaining node dependency and data liquidity indicators are:

[0064] Perform time series analysis on the sensor data at each checkpoint and calculate the standard deviation using statistical methods and mean , using the formula: and , generate a list of standard deviations and means;

[0065] in, Representation Node The volatility of data, Representation Node The average data traffic is the total unit of time, Is a node In time The data value of , is the adjustment factor;

[0066] Based on the standard deviation and mean list, the data flow dependency between nodes is evaluated using the formula: , generating the dependency matrix ;

[0067] in, Representation Node and nodes The strength of the dependence between , is the mean result, , is the standard deviation result, To ensure numerical stability, is the intensity adjustment coefficient;

[0068] According to the dependency matrix and mean list, the data mobility of the nodes is analyzed using the formula: , calculate the data mobility index of each node, and obtain the node dependency and data mobility index;

[0069] in, Is a node Data liquidity indicators, is the nonlinear adjustment coefficient.

[0070] formula: ,

[0071] in:

[0072] :node The volatility of data;

[0073] :node Average data traffic volume;

[0074] : Total number of units in the time period;

[0075] :node In time The data value of

[0076] : Adjustment coefficient, used to control data scale and sensitivity;

[0077] Calculation example:

[0078] Assume Node The data values ​​in a 4-hour period are ,calculate and :

[0079] ;

[0080] For the standard deviation, the square of the deviation is calculated for each time point, then the average is taken and a small adjustment factor is added 0.1:

[0081] ;

[0082] formula: ;

[0083] in:

[0084] node With Node The strength of the dependence between

[0085] : Strength adjustment coefficient, set to 0.5 to increase the nonlinear nature of the calculation;

[0086] : A tiny constant, set to 0.01 to avoid division by zero error;

[0087] Calculation example:

[0088] Assume that another node of and ,

[0089] ;

[0090] formula: ;

[0091] in:

[0092] :node Data liquidity indicators;

[0093] : Nonlinear adjustment coefficient, set to 0.5 to enhance model flexibility;

[0094] Calculation example:

[0095] Assumptions There are two dependent nodes and , known , and another node of and ,

[0096] ;

[0097] ;

[0098] The calculation shows that the node Data liquidity indicators 5, indicating a node The total activity or influence in the data stream.

[0099] See also Figure 3 , the specific steps for obtaining the optimal path configuration are:

[0100] According to the data liquidity index results, the connection cost between all nodes is evaluated using the formula: , generate the connection cost matrix;

[0101] in, Representation Node and nodes The connection cost between , It is a multi-node data liquidity indicator. , is the parameter that regulates the influence of the node, is the overall moderating factor;

[0102] Use the connection cost matrix, combined with the node dependency matrix, to calculate the benefit of each path using the formula: , generate the benefit matrix;

[0103] in, Representation Node To Node The total benefit of the path, is the dependency strength between nodes, It is to avoid adding a small amount of zero. , is a nonlinear parameter that adjusts the benefit calculation;

[0104] According to the benefit matrix, the shortest path algorithm is used to capture the path with the lowest cost and the highest benefit, using the formula: , determine the optimal path configuration;

[0105] in, represents the optimal path, It is a parameter that regulates the impact of benefits in path selection.

[0106] formula: ;

[0107] Parameter Description:

[0108] :node and The data liquidity indicator is assumed to reflect the frequency or capacity of data exchange between nodes;

[0109] :adjust and Parameters that influence the intensity of

[0110] : Parameters to adjust overall cost sensitivity;

[0111] Assumed parameter values:

[0112] , , ;

[0113] Calculation process:

[0114] calculate and The adjusted impact of:

[0115] ;

[0116] Add the adjusted values:

[0117] ;

[0118] The sum of The cost is calculated using the power as the denominator: ;

[0119] Calculating connection costs : ;

[0120] Cost Value It means the connection cost is low, which means the communication between these two nodes is very efficient or cost-effective.

[0121] formula: ;

[0122] Parameter Description:

[0123] The strength of dependencies between nodes;

[0124] : Nonlinear parameters that adjust the dependency intensity and cost impact;

[0125] Small amounts that avoid division by zero;

[0126] Assumed parameter values:

[0127] , , ;

[0128] Calculation process:

[0129] Adjustment of reference dependency strength: ;

[0130] Add a small amount to the previously calculated connection cost : ;

[0131] Calculate the benefits: ;

[0132] Benefit value High, indicating path arrive It is advantageous at a given dependency strength and cost.

[0133] formula: ;

[0134] Parameter Description:

[0135] : Parameters that adjust the impact of benefits in path selection;

[0136] Assumed parameter values: ;

[0137] Calculation process:

[0138] Benefit value Make adjustments: ;

[0139] Assuming there are multiple paths, the total cost will be Adjust the sum of the values ​​and select the smallest one as the optimal path.

[0140] See also Figure 4 , the steps to obtain clustering recognition results are as follows:

[0141] Based on the input of the optimal path configuration, refer to the inspection data characteristics, divide the time window, and use the formula: , generate time window division results;

[0142] in, Represents the total time range, Represents the number of windows, is the tuning parameter of the path configuration, Represents the length of each time window;

[0143] Perform time series partitioning on the data within each time window partitioning result using the formula: , generate time series partition results;

[0144] in, Representative The sum of data in time windows, Represents the time window Time point The data value of

[0145] Based on the time series partitioning results, cross entropy is applied to determine the fitting degree of multi-time window data using the formula: , generate the goodness-of-fit result;

[0146] in, Representative Window The degree of fit, represents the probability of the actual distribution, represents the probability of the expected distribution, represents the length of each time window, is a correction term to refer to the deviation of the data, is the weight factor;

[0147] Based on the fit, the time series are classified and clustered using the formula: , get the cluster recognition result;

[0148] in, Representative data points The clustering result of is the cluster center, is the within-class standard deviation, It is a small amount to avoid division by zero.

[0149] formula: ;

[0150] Parameter explanation and examples:

[0151] : The total time range, assumed to be 1000 minutes;

[0152] : The number of time windows, assuming there are 10 windows;

[0153] : Parameter adjusted according to the path configuration. Assuming it is 2, it means that each window will be weighted according to the path characteristics;

[0154] Calculation process: ;

[0155] Here, the length of each window is 50 minutes, which is calculated based on the given total time and number of windows and path configuration tuning parameters.

[0156] formula: ;

[0157] Parameter explanation and examples:

[0158] : No. The sum of data in time windows;

[0159] :window At the point in time data value, assuming that the data value per minute is 10;

[0160] Calculation process:

[0161] Assumptions , ;

[0162] Here, the total amount of data in the window is 500, which means that within 50 minutes, the data value recorded every minute is 10, and the total is accumulated.

[0163] formula: ;

[0164] Parameter explanation and examples:

[0165] :window The fit of

[0166] , : are the probabilities of the actual and expected distributions, respectively, assuming ;

[0167] : Correction terms, assuming each term is 0.1;

[0168] : Weight factor, assumed to be 1;

[0169] Calculation process:

[0170] ;

[0171] The goodness of fit calculated here is 27.9, which represents the difference between the actual data and the expected data and the total impact of the correction term.

[0172] formula: ;

[0173] Parameter explanation and examples:

[0174] : Data points The clustering results of

[0175] : Time window The goodness of fit is 27.9 calculated in the previous step;

[0176] : Cluster center, assumed to be 30;

[0177] : Intra-class standard deviation, assumed to be 5;

[0178] : A small amount to avoid division by zero, assumed to be 0.1;

[0179] Calculation process:

[0180] ;

[0181] Here, the calculated clustering results , which represents the data point The squared distance calculated according to the clustering formula.

[0182] See also Figure 5 , the specific steps for obtaining the dynamic update path are:

[0183] Receive real-time sensor data and analyze the deviation from the cluster recognition results using the formula: , calculate the weighted deviation of each checkpoint from the cluster center and generate the deviation measurement result;

[0184] in, is the deviation measurement result, Represents real-time data, Representative time The weight factor of Represents total time;

[0185] According to the deviation measurement results, the priority of the emergency checkpoints of the inspection path is re-ordered, using the sorting formula: , ensure that high deviation points get high priority and generate reordered urgent checkpoints;

[0186] in, represents an urgent checkpoint for reordering, represents a decimal number that avoids division by zero, Representative area The importance coefficient of

[0187] Based on the reordered checkpoints, the inspection path is adjusted to prioritize access to key areas, and the path adjustment formula is applied: ,Using priority sorting to ensure that key areas are visited first and generate dynamically updated paths;

[0188] in, Represents the original path, Represents a dynamically updated path.

[0189] formula: ;

[0190] in:

[0191] :time Real-time data points;

[0192] : cluster center;

[0193] :time The weight of

[0194] : Total number of time windows.

[0195] Assumed parameters:

[0196] Total time (Assuming 4 time points);

[0197] Real-time data d it =[20,25,30,35] (Real-time data at each time point);

[0198] Cluster Center (assumed average value);

[0199] Weight w t =[1,2,1,0.5] (weight at each time point, weighted bias);

[0200] Calculation process:

[0201] Calculate the square of the deviation at each time point multiplied by the weight:

[0202] Sum these values: ;

[0203] Take the square root of the sum: ;

[0204] Result analysis:

[0205] Represents the weighted Euclidean distance between the real-time data and the cluster center at a given weight and time point. The result is used to compare the degree of deviation under differentiated checkpoints or differentiated real-time conditions.

[0206] formula: ;

[0207] in:

[0208] : Deviation measure;

[0209] : Small amount, avoiding division by zero;

[0210] :area The importance coefficient of

[0211] Assumed parameters:

[0212] ;

[0213] ;

[0214] ;

[0215] Calculation process:

[0216] Substituting the values ​​into the formula: ;

[0217] Result analysis: Indicates the priority of the checkpoint given the deviation and region importance. A higher value indicates a higher priority, and a lower value indicates that the checkpoint is not the most urgent.

[0218] formula: ;

[0219] This step involves logic rather than numerical calculations, and the processing method is based on Value pair original path Sort.

[0220] Assumption Path Including points [A, B, C, D], according to The values ​​are sorted. If The value of , then the sorted path Will be [D,B,A,C] , indicating that point D has the highest priority.

[0221] See also Figure 6 , the specific steps for obtaining abnormal monitoring results are:

[0222] According to the dynamic update path, the data collection process of the inspection robot is monitored using the formula: , generate data collection results;

[0223] in, Represents the data collection results, Represents at time By path Sensors inside Data collected, represents the sensor weight, represents the data validity coefficient;

[0224] Based on the data collection results, set the anomaly threshold and perform data anomaly detection using the formula: , generate anomaly detection results;

[0225] in, represents the anomaly detection result, , Respectively represent The mean and standard deviation of represents the time decay coefficient;

[0226] Based on the anomaly detection results, identify the abnormal time series segments and perform trend analysis using the formula: , generate abnormal monitoring results;

[0227] in, Represents abnormal monitoring results, represent The standard deviation of Represents the time adjustment parameter.

[0228] formula: ;

[0229] in:

[0230] Representative sensor In time The data values ​​collected;

[0231] It is a sensor The weight indicates the importance of sensor data;

[0232] is the data validity coefficient, indicating the quality of the data;

[0233] Assumptions:

[0234] set up Includes three sensors , at time ,sensor The readings are , the weight of the sensor , effectiveness coefficient ;

[0235] Calculation process:

[0236] ;

[0237] Analysis: Results Indicates at time ,After referring to sensor importance and data validity, the weighted total data volume of all sensors is 475.

[0238] formula: ;

[0239] in:

[0240] , They are The mean and standard deviation of

[0241] is the time decay coefficient;

[0242] Assumption: Assume , and use the previous Results ,average value , standard deviation , time decay coefficient ;

[0243] Calculation process:

[0244] D it = 1 3 475 - 475 5 2 ⋅ e - 0.1 ⋅ 1 + 480 - 475 5 2 ⋅ e - 0.1 ⋅ 2 + 470 - 475 5 2 ⋅ e - 0.1 ⋅ 3 D it = 1 3 0 ⋅ e - 0.1 +1 ⋅ e - 0.2 +1 ⋅ e - 0.3 D it = 1 3 [0+0.82+0.74] D it =0.52 ;

[0245] Analysis: Indicates the degree of abnormality of the entire time series with reference to the time decay of the data. The larger the number, the more abnormal the data.

[0246] formula: ;

[0247] in:

[0248] is the set abnormal threshold;

[0249] yes The standard deviation of

[0250] is the time adjustment parameter;

[0251] Assumptions:

[0252] Assumed anomaly threshold , standard deviation , time adjustment parameters ,use Result;

[0253] Calculation process:

[0254] E=max 1 2 ⋅ 0.52 - 0.5 0.1 ⋅ 1 1+0.05 ⋅ 1 E=max[1 ⋅ 0.2 ⋅ 0.95] E=0.19 ;

[0255] Analysis: Results Indicates the indicator value with the most significant abnormal segment in a given time series. This indicates the presence of slightly abnormal behavior in the time series.

[0256] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. The petrochemical area inspection system based on inspection robots is characterized by: The system comprises: The path optimization module is based on the real-time collection of sensor data between inspection points, analyzes data between multiple inspection points, integrates node dependency and data mobility indicators, calculates the interaction strength of each node through a neural network model, determines the dependency graph, and obtains the optimal path configuration; The time series analysis module divides the time windows based on the input of the optimal path configuration, performs time series division of the inspection data, determines the fitting degree of the multi-time window data through cross entropy calculation, classifies and clusters the time series, and obtains the cluster recognition result; The dynamic redirection module receives real-time sensor data, analyzes deviations from the cluster identification results, reorders emergency checkpoints on the inspection path according to the degree of deviation, adjusts the path to prioritize access to the most critical areas, and establishes a dynamically updated path; The abnormal mode monitoring module monitors the data collection process of the inspection robot according to the dynamic update path, sets the abnormal threshold and performs data anomaly detection, identifies the abnormal time series segment, and performs trend analysis to generate abnormal monitoring results; The steps for obtaining the node dependency and data liquidity indicators are specifically as follows: Perform time series analysis on the sensor data at each checkpoint and calculate the standard deviation using statistical methods and mean , using the formula: and Generate lists of standard deviations and means; in, Representation Node The volatility of data, Representation Node The average data traffic is the total time unit, Is a node In time The data value of , is the adjustment factor; Based on the standard deviation and mean lists, the data flow dependency between nodes is evaluated using the formula: Generate dependency matrix ; in, Representation Node and nodes The strength of the dependence between , is the mean result, , is the standard deviation result, To ensure numerical stability, is the intensity adjustment coefficient; The data mobility of the nodes is analyzed according to the dependency matrix and the mean list, using the formula: Calculate the data liquidity index of each node to obtain the node dependency and data liquidity index; in, Is a node Data liquidity indicators, is the nonlinear adjustment coefficient; The steps for obtaining the optimal path configuration are specifically as follows: According to the data liquidity index results, the connection cost between all nodes is evaluated using the formula: Generate a connection cost matrix; in, Representation Node and nodes The connection cost between , It is a multi-node data liquidity indicator. , is the parameter that regulates the influence of the node, is the overall moderating factor; Using the connection cost matrix, combined with the node dependency matrix, the benefit of each path is calculated using the formula: Generate a benefit matrix; in, Representation Node To Node The total benefit of the path, is the dependency strength between nodes, It is to avoid adding a small amount of zero. , is a nonlinear parameter that adjusts the benefit calculation; According to the benefit matrix, the shortest path algorithm is used to capture the path with the lowest cost and the highest benefit, using the formula: Determine the optimal path configuration; in, represents the optimal path, It is a parameter that regulates the impact of benefits in path selection.

2. The petrochemical area inspection system based on the inspection robot according to claim 1 is characterized in that: The steps for obtaining the cluster recognition result are specifically as follows: Based on the input of the optimal path configuration, the time window is divided with reference to the inspection data characteristics, and the formula is used: Generate time window division results; in, Represents the total time range, Represents the number of windows, is the tuning parameter of the path configuration, Represents the length of each time window; Time series partitioning is performed on the data in each of the time window partitioning results, using the formula: Generate time series partitioning results; in, Representative The sum of data in time windows, Represents the time window Time point The data value of Based on the time series partitioning results, the cross entropy is applied to determine the fitting degree of the multi-time window data, using the formula: Generate goodness-of-fit results; in, Representative Window The degree of fit, represents the probability of the actual distribution, represents the probability of the expected distribution, represents the length of each time window, is a correction term to refer to the deviation of the data, is the weight factor; Based on the fit, the time series are classified and clustered using the formula: Get the clustering recognition result; in, Representative data points The clustering result of is the cluster center, is the within-class standard deviation, It is a small amount to avoid division by zero.

3. The petrochemical area inspection system based on the inspection robot according to claim 1 is characterized in that: The steps for obtaining the dynamic update path are specifically as follows: Receive real-time sensor data and analyze the deviation from the cluster recognition result using the formula: Calculate the weighted deviation of each checkpoint from the cluster center and generate a deviation metric result; in, is the deviation measurement result, Represents real-time data, Representative time The weight factor of Represents total time; The priority of the emergency checkpoints of the inspection path is re-ordered according to the deviation measurement results, using the sorting formula: Ensure that high deviation points get high priority and generate reordered urgent checkpoints; in, represents an urgent checkpoint for reordering, represents a decimal number that avoids division by zero, Representative area The importance coefficient of Based on the reordered checkpoints, the inspection path is adjusted to prioritize access to the critical areas, applying the path adjustment formula: Priority sorting is used to ensure that key areas are accessed first, and dynamically updated paths are generated; in, Represents the original path, Represents a dynamically updated path.

4. The petrochemical area inspection system based on the inspection robot according to claim 1 is characterized in that: The steps for obtaining the abnormal monitoring results are specifically as follows: According to the dynamic update path, the data collection process of the inspection robot is monitored using the formula: Generate data collection results; in, Represents the data collection results, Represents at time By path Sensors inside Data collected, represents the sensor weight, represents the data validity coefficient; Based on the data collection results, anomaly thresholds are set and data anomaly detection is performed using the formula: Generate anomaly detection results; in, represents the anomaly detection result, , Respectively represent The mean and standard deviation of represents the time decay coefficient; Based on the anomaly detection results, the abnormal time series segments are identified and trend analysis is performed using the formula: Generate abnormal monitoring results; in, Represents abnormal monitoring results, represent The standard deviation of Represents the time adjustment parameter.