CN-121980209-A - Low-altitude equipment based on multi-dimensional rule monitoring and intelligent fault analysis processing method

CN121980209ACN 121980209 ACN121980209 ACN 121980209ACN-121980209-A

Abstract

The application discloses a low-altitude equipment and intelligent fault analysis processing method based on multi-dimensional rule monitoring, which comprises the steps of preprocessing collected multi-source heterogeneous data to obtain multi-dimensional feature vectors, carrying out rule extraction on the multi-dimensional feature vectors to obtain a three-dimensional rule set containing equipment dimension rules, environment dimension rules and experience dimension rules, carrying out weighting processing on the three-dimensional rule set through a pre-built situation perception weight model, carrying out fusion calculation on the weighted rules through an improved D-S evidence theory to obtain fault probability distribution of the low-altitude equipment, carrying out state judgment and trend prediction on the fault probability distribution through a time scale evaluation mechanism to obtain comprehensive evaluation scores containing abnormal grades, fault types and evolution trends, and analyzing the fault types through a pre-built knowledge map when the comprehensive evaluation scores are larger than a preset threshold. The application improves the operation safety and maintenance efficiency of low-altitude equipment in a complex environment.

Inventors

ZHOU LIANG
SHAN HAIFENG
SHEN WENBIN
SHI WENBING
LI ZENGKAI
XU XIAO
YANG FAN
LIU DONGLIANG

Assignees

中电信无人科技(江苏)有限公司

Dates

Publication Date: 20260505
Application Date: 20251222

Claims (9)

1. A multi-dimensional rule-based low-altitude monitoring equipment and an intelligent fault analysis processing method are characterized by comprising the following steps: s1, acquiring multi-source heterogeneous data, wherein the multi-source heterogeneous data comprises equipment operation parameter data, environment monitoring data and historical experience data; S2, carrying out rule extraction on the multidimensional feature vector to obtain a three-dimensional rule set containing equipment dimension rules, environment dimension rules and experience dimension rules, wherein each rule is encoded into a condition-result-confidence coefficient triplet structure; S3, weighting the three-dimensional rule set through a pre-constructed situation awareness weight model, and carrying out fusion calculation on the weighted rule through an improved D-S evidence theory to obtain fault probability distribution of low-altitude equipment; S4, carrying out state judgment and trend prediction on the fault probability distribution through a time scale evaluation mechanism to obtain a comprehensive evaluation value containing an abnormal grade, a fault type and an evolution trend; And S5, when the comprehensive evaluation value is greater than a preset threshold value, analyzing the fault type through a pre-constructed knowledge graph to obtain fault cause and treatment measure suggestions, feeding the fault cause and treatment measure suggestions back to the steps S2 and S3, updating the confidence coefficient of the three-dimensional rule set, and updating the weight parameters of the context perception weight model.
2. The multi-dimensional rule-based low-altitude equipment and intelligent fault analysis processing method according to claim 1, wherein the method comprises the following steps: S1, preprocessing acquired multi-source heterogeneous data to obtain a multi-dimensional feature vector, wherein the method comprises the following steps of: carrying out space-time alignment on the collected equipment operation parameter data, environment monitoring data and historical experience data; Denoising the multi-source heterogeneous data after space-time alignment by using a sliding window filter; extracting statistical features from the denoised data; Performing Fast Fourier Transform (FFT) on the denoised data, and extracting spectral energy distribution from a transformation result as a frequency domain feature; And splicing the statistical features and the frequency domain features to obtain a multidimensional feature vector.
3. The multi-dimensional rule-based low-altitude equipment and intelligent fault analysis processing method according to claim 2, wherein the method comprises the following steps: s2, carrying out rule extraction on the multidimensional feature vector to obtain a three-dimensional rule set containing equipment dimension rules, environment dimension rules and experience dimension rules, wherein the method comprises the following steps: The method comprises the steps of extracting equipment parameter vectors from multi-dimensional feature vectors, establishing a threshold reference based on historical equipment data, detecting abnormal values of real-time equipment parameter data through a sliding time window according to the threshold reference, carrying out trend analysis on the equipment parameter data in the sliding time window, acquiring reconstruction deviation between input equipment parameter data and reconstruction data through a pre-trained deep learning model, and constructing equipment dimension rules according to the abnormal value detection result, the trend analysis result and the reconstruction deviation; the method comprises the steps of extracting an environment parameter vector from a multidimensional feature vector, discretizing continuous environment parameters to obtain an environment grade code, establishing a corresponding relation table of environment factors and equipment performance attenuation according to the environment grade code, acquiring a mapping relation between environment parameter combinations and attenuation coefficients in the corresponding relation table through a machine learning algorithm, and converting the mapping relation into an environment dimension rule; The method comprises the steps of extracting fault case data from historical experience data, identifying fault entities and entity relations from the fault case data through natural language processing, and constructing a fault domain knowledge graph; And encoding the equipment dimension rule, the environment dimension rule and the experience dimension rule to generate a condition-result-confidence coefficient triple structure, and summarizing to form a three-dimensional rule set.
4. The multi-dimensional rule-based low-altitude equipment and intelligent fault analysis processing method according to claim 2, wherein the method comprises the following steps: s3, carrying out weighting treatment on the three-dimensional rule set through a pre-constructed situation awareness weight model, and carrying out fusion calculation on the weighted rules through an improved D-S evidence theory to obtain fault probability distribution of low-altitude equipment, wherein the method comprises the following steps: extracting an environment parameter value at the current moment from the multidimensional feature vector, discretizing temperature, humidity, wind speed and electromagnetic interference parameters in the environment parameter value into grade values, and constructing a situation vector according to the type of a task currently executed by low-altitude equipment; Generating a basic weight vector by utilizing a pre-trained context-weight mapping matrix according to the context vector; obtaining the confidence coefficient and the historical execution accuracy of each rule in the three-dimensional rule set, and calculating the real-time accuracy feedback value of each rule ; Based on the base weight vector and the real-time accuracy feedback value Generating a dynamic weight Wi of each rule through weighted combination; Taking rules meeting preset triggering conditions in the three-dimensional rule set as evidence bodies, and constructing support degree distribution according to the result of each rule and the corresponding confidence coefficient; According to dynamic weights The support degree distribution of a plurality of evidence bodies is weighted and fused by utilizing an improved D-S evidence theory, and preliminary fault probability distribution is obtained; And performing time sequence consistency test on the preliminary fault probability distribution through a sliding time window to obtain final fault probability distribution.
5. The multi-dimensional rule-based low-altitude equipment monitoring and intelligent fault analysis processing method according to claim 4, wherein: dynamic weighting By the following formula: ; Wherein, the Dynamic weight of the ith rule, lambda is a time attenuation coefficient, and sigma is a Sigmoid normalization function; the context vector corresponding to the ith rule; and the basic weight value corresponding to the ith rule in the context-weight mapping matrix.
6. The multi-dimensional rule-based low-altitude equipment monitoring and intelligent fault analysis processing method according to claim 4, wherein: real-time accuracy feedback value for each rule Calculated by the following formula: ; Wherein, the Indicating the number of times that the ith rule is correctly triggered in the history execution process; indicating the number of false triggers of the ith rule during historical execution.
7. The multi-dimensional rule-based low-altitude equipment monitoring and intelligent fault analysis processing method according to claim 4, wherein: According to dynamic weights The support degree distribution of a plurality of evidence bodies is weighted and fused by utilizing an improved D-S evidence theory, and the preliminary fault probability distribution is obtained, which comprises the following steps: Taking each rule meeting the preset triggering condition as an evidence body, and calculating probability distribution of the j evidence body , wherein, Representing the jth evidence body versus fault type Is used for the support value of (a), Indicating the fault type pointed to by the j-th rule, Representing the confidence of the ith rule; according to probability distribution, calculating a conflict coefficient K between any two evidence bodies by using a D-S evidence theory; When the conflict coefficient K is larger than a preset threshold value, the probability distribution of all evidence bodies is processed by using a Murphy correction method, and corrected fusion probability distribution is obtained; when the conflict coefficient K is smaller than or equal to a preset threshold value, directly fusing probability distribution of a plurality of evidence bodies by using a standard D-S combination rule to obtain fused probability distribution; And carrying out normalization processing on the corrected fusion probability distribution and the fused probability distribution to obtain the primary fault probability distribution.
8. The multi-dimensional rule-based low-altitude equipment and intelligent fault analysis processing method according to claim 2, wherein the method comprises the following steps: S4, carrying out state judgment and trend prediction on the fault probability distribution through a time scale evaluation mechanism to obtain a comprehensive evaluation value containing an abnormal grade, a fault type and an evolution trend, wherein the method comprises the following steps: Sampling the fault probability distribution according to a preset short period and a preset long period, wherein the short period preset sampling period is that The long period presets the sampling period as ; Less than ; Smoothing the fault probability sequence sampled in a short period by using an exponential weighted moving average algorithm to obtain a short-term fault probability value; Carrying out smoothing treatment on the fault probability sequence sampled in a long period by utilizing a sliding time window to obtain a long-term fault probability value; Extracting a maximum probability value from the short-term fault probability value and the long-term fault probability value, and a fault type corresponding to the maximum probability value; judging the abnormal level of the fault according to the maximum probability value and the corresponding fault type; Constructing a time sequence data set containing fault probability distribution, multidimensional feature vectors and abnormal grades; Predicting the fault probability at the future moment through a pre-trained LSTM neural network model according to the time sequence data set; And weighting and calculating the comprehensive evaluation value of the fault probability according to the current abnormal grade, the fault type corresponding to the maximum probability value and the predicted fault probability.
9. The multi-dimensional rule-based low-altitude equipment and intelligent fault analysis processing method according to claim 2, wherein the method comprises the following steps: S5, when the comprehensive evaluation value is larger than a preset threshold value, analyzing the fault type through a pre-constructed knowledge graph to obtain fault reasons and treatment measure suggestions, wherein the method comprises the following steps: Judging whether the comprehensive evaluation value is larger than a preset threshold value, if so, extracting the fault type with the maximum probability value from the final fault probability distribution as the current fault type; taking the current fault type as a query starting point, and executing reverse path search in a pre-constructed fault domain knowledge graph; Calculating the comprehensive confidence coefficient of each searched path; Sequencing all paths according to the comprehensive confidence coefficient, selecting the root cause node of the path with the highest comprehensive confidence coefficient as a fault cause, and extracting the processing measure nodes associated with the root cause node to obtain a processing measure suggestion; wherein, the comprehensive confidence is calculated by the following formula: ; wherein PathScore represents the path comprehensive confidence from the current fault type node to the root cause node, n represents the total number of nodes contained in the path; a weight value representing an edge between the i-1 th node and the i-th node in the path; And the real-time data support degree of the ith node is represented.

Description

Low-altitude equipment based on multi-dimensional rule monitoring and intelligent fault analysis processing method Technical Field The application relates to the field of fault analysis, in particular to a multi-dimensional rule-based low-altitude monitoring equipment and an intelligent fault analysis processing method. Background With the rapid development of low-altitude economy, the number of low-altitude equipment such as unmanned aerial vehicles, eVTOL (electric vertical take-off and landing aircrafts), low-altitude communication and monitoring equipment and the like is increased, and the application scene of the low-altitude equipment is expanded from the logistics distribution and the agricultural plant protection to the key fields such as urban air traffic, emergency rescue and the like. The operation environment of the low-altitude equipment presents high complexity and dynamic variability characteristics, and is not only required to face the influence of natural environment factors such as temperature, humidity, wind speed and the like, but also is required to cope with urban environment challenges such as electromagnetic interference, building shielding and the like. Under the background, the equipment faults can cause flight accidents, communication interruption and even airspace safety accidents, and the accurate and reliable fault analysis on low-altitude equipment becomes a key technical requirement for guaranteeing low-altitude economic safety development. However, the existing low-altitude equipment fault detection method mainly adopts a monitoring system based on fixed rules, and equipment operation parameters are judged through preset static thresholds. The traditional method has the technical defects that firstly, the rule weight is fixed and cannot be dynamically adjusted according to real-time operation situations, for example, the battery temperature threshold value is correspondingly improved under high-temperature environment, but the fixed rule cannot realize self-adaptive adjustment, secondly, the same parameter abnormality under different environment conditions can represent different fault severity degrees, the rule of the fixed weight cannot distinguish the difference, so that the fault diagnosis accuracy is obviously reduced under complex and changeable environments, thirdly, due to lack of situation awareness capability, the system frequently generates false alarms or missed alarms, for example, the normal signal fluctuation is misjudged as communication faults under strong electromagnetic interference environments, or early signs of equipment performance attenuation under extreme weather conditions are ignored, and finally, a fixed rule system lacks a learning and updating mechanism, and cannot optimize the rule weight according to historical fault cases and actual operation experience, so that the system performance cannot be continuously improved. Therefore, development of an intelligent fault analysis method capable of dynamically adjusting rule weights according to real-time situations, fully fusing multi-source heterogeneous data and having self-adaptive learning capability is needed to meet the safety guarantee requirements of low-altitude equipment in a complex operation environment. Disclosure of Invention Aiming at the problem in the prior art that the rule weight existing when the low-altitude equipment adopts a fixed rule to perform fault analysis in a complex and changeable operating environment can not be dynamically adjusted, the application provides the multi-dimensional rule-based low-altitude equipment monitoring and intelligent fault analysis processing method, which improves the operating safety and maintenance efficiency of the low-altitude equipment in the complex environment. The application provides a low-altitude equipment and intelligent fault analysis processing method based on multi-dimensional rule monitoring, which comprises the steps of S1, collecting multi-source heterogeneous data, preprocessing the collected multi-source heterogeneous data to obtain multi-dimensional feature vectors, S2, extracting rules of the multi-dimensional feature vectors to obtain a three-dimensional rule set containing equipment dimension rules, environment dimension rules and experience dimension rules, wherein each rule is encoded into a condition-result-confidence coefficient triplet structure, S3, weighting the three-dimensional rule set through a pre-built context perception weight model, carrying out fusion calculation on the weighted rules through an improved D-S evidence theory to obtain fault probability distribution of the low-altitude equipment, S4, carrying out state judgment and trend prediction on the fault probability distribution through a time scale evaluation mechanism to obtain comprehensive evaluation scores containing abnormal grades, fault types and evolution trends, S5, analyzing the fault types through pre-built knowledge to obtain fault causes