CN-121388793-B - AI-based loading system equipment fault prediction method

CN121388793BCN 121388793 BCN121388793 BCN 121388793BCN-121388793-B

Abstract

The invention relates to the technical field of machine learning, in particular to an AI-based loading system equipment fault prediction method, which comprises the following steps of collecting load and current signals, dividing and differentiating the load and current signals according to beats, judging direction consistency, extracting control and response time sequence pairing numbers, identifying synchronous offset and path jump characteristics, finishing numbering by aggregating offset extension information, and outputting loading system fault prediction results. According to the invention, a sequential fragment mapping of load and current difference is constructed, synchronous offset position extraction and numbering path marking are combined, a response identification chain under a propulsion beat is established, merging judgment of dynamic evolution characteristics of the numbering path is realized by means of rearrangement and aggregation of a trigger interval difference sequence, reverse switching point positioning jump behavior in a difference direction is extracted, a numbering mapping boundary of the jump path is established, offset extension characteristics are introduced to complete numbering classification aggregation, and capturing capacity of synchronization abnormality among multiple components and forward movement sensing capacity of fault trend are enhanced.

Inventors

PENG ENQI
CHEN QINGTONG
LIN WEI
Chen Shunrui
FAN YONGSHENG
Lai Qingyin
ZHENG SHILONG
Liao Chuanxu
WANG SHUYU

Assignees

龙合智能装备制造有限公司

Dates

Publication Date: 20260512
Application Date: 20251224

Claims (9)

1. The AI-based loading system equipment fault prediction method is characterized by comprising the following steps: S1, collecting load deformation signals of a roller drive axle at the tail part of a conveying structure of train loading equipment and driving current change records of lifting supporting points of a stacking grabbing support, dividing the segments according to the propulsion beats, executing load and current first-order difference on each segment, judging the direction consistency, and summarizing and sequencing to generate a load linkage segment table; s2, based on the load linkage segment table, extracting control trigger and response time of a propulsion component, matching same-number records, calculating time intervals, generating a number interval sequence, merging same-direction change numbers, and forming a control response difference sequence set; And S3, splicing the control response difference sequence set and the load linkage segment table, correspondingly extracting load and interval difference points, calculating a synchronous offset position, identifying repeated path numbers, marking indexes, summarizing a cross coverage relationship, and generating a synchronous response offset layer structure, wherein the specific steps of S3 are as follows: S301, splicing the control response difference sequence set and the same numbered fragments in the load linkage fragment table according to a propulsion sequence, extracting point position index columns corresponding to the spliced load difference values and interval difference values, and establishing a load interval corresponding point column set; S302, respectively extracting a load difference value and an interval difference value under a pushing index according to the load interval corresponding point column set, calculating a difference sequence for the index difference of the two types of values under the same pushing position, extracting a pushing index corresponding to a non-zero difference value, and generating a synchronous offset position sequence; s303, calling the synchronous offset position sequence, searching repeated number paths, marking start-stop pushing index positions of each path, merging the number path interval information with cross coverage, and obtaining a synchronous response offset layer structure; s4, according to the synchronous response offset layer structure, searching repeated numbers, extracting offset indexes and positions, and identifying load and interval difference reverse switching points to form a jump mapping boundary diagram; The synchronous response offset layer structure is characterized in that a pushing position with a non-zero difference value is extracted through splicing load difference data and response interval difference data, and start-stop pushing indexes and cross coverage relations of repeated numbered paths are identified and merged to represent synchronous offset distribution and association relations of multiple numbered paths; The jump mapping boundary diagram is based on a synchronous response offset layer structure, extracts the differential direction of the number at the offset position, verifies the reverse change position as a switching point, constructs a jump path and establishes a mapping relation between the number and the path point, thereby representing the differential direction abrupt change boundary and the path distribution; and S5, searching whether the interval offset direction in the control response difference sequence set extends continuously or not according to the jump mapping boundary diagram, marking an abnormal number, aggregating a time index, and outputting a fault prediction result of the loading system.
2. The AI-based loading system equipment fault prediction method according to claim 1, wherein the load linkage segment table comprises a structure number sequence, a differential direction inconsistent position index and a propulsion beat corresponding time segment, the control response difference sequence set comprises a trigger response interval value, a propulsion sequence number pairing and an adjacent homodromous change merging number, the synchronous response offset layer structure comprises a number path set, a synchronous offset point position, a path start-stop index section and a path cross coverage relation, the jump map boundary map comprises a jump number, a differential direction switching point, a number and path point mapping pair, and the loading system fault prediction result comprises an abnormal number category, a ductility offset direction characteristic and an abnormal number time distribution.
3. The AI-based loading system equipment failure prediction method according to claim 1, wherein the load-to-interval differential reverse switching point refers to a time point at which a load-to-interval change direction is reversed in a control response process.
4. The AI-based loading system equipment failure prediction method according to claim 1, wherein the abnormality number refers to a number record in which a time-spread discontinuity and a direction change abnormality occur in a control response process.
5. The AI-based loading system equipment failure prediction method according to claim 1, wherein the specific steps of S1 are: S101, collecting load deformation signals of a roller drive axle at the tail part of a conveying structure of train loading equipment and driving current change records of lifting support points of a stacking grabbing support, extracting time indexes of the two types of signals, dividing signal data frames according to propulsion beats, and generating a propulsion beat signal fragment set; s102, calling the propulsion beat signal segment set, respectively executing first-order difference on deformation signals and current sequences in each segment, and comparing difference symbol directions to generate a difference direction inconsistent position index table; S103, calling the index table of the inconsistent positions of the differential directions, searching the corresponding conveying structure numbers, and ordering the number information according to the index time sequence to generate a load linkage segment table.
6. The AI-based loading system equipment failure prediction method according to claim 1, wherein the specific steps of S2 are: S201, after loading the load linkage segment table, extracting control trigger time and response mark time of a loading cabin door guide rail propulsion assembly, pairing the two types of time information item by item according to numbers, and rearranging according to the sequence of the trigger time to generate a control response time pairing sequence; S202, calling the control response time pairing sequence, calculating a time interval according to each pair of trigger time and response time, writing a result into a corresponding number sequence, and constructing an interval value sequence corresponding to a propulsion sequence under each group of numbers to obtain a trigger response time interval sequence; S203, according to the trigger response time interval sequence, interval values corresponding to all numbers are arranged according to a pushing sequence, the interval change direction of adjacent numbers is judged, and the numbers which continuously change in the same direction are merged to generate a control response difference sequence set.
7. The AI-based loading system equipment failure prediction method according to claim 1, wherein the specific step of S4 is: S401, calling all repeated path numbers in the synchronous response offset layer structure, searching offset indexes and pushing segment position information corresponding to each number, positioning pushing indexes of offset segments in an associated pushing sequence, and generating a number offset index table; S402, extracting a load differential direction and an interval differential direction at each offset index according to the serial number offset index table, and carrying out sign reverse judgment on the two directions, and recording a pushing index as a switching point if the direction changes from positive to negative and from negative to positive, so as to obtain a direction switching index set; s403, according to the direction switching index set, the pushing positions of the switching points are rearranged according to the number sequence and combined into a hopping path, a one-to-one mapping structure between the numbers and the hopping path points is built for each number, and a hopping mapping boundary diagram is built.
8. The AI-based loading system equipment failure prediction method according to claim 1, wherein the specific step of S5 is: S501, according to the number of the jump map boundary map mark, retrieving interval records of corresponding numbers in the control response difference sequence set, extracting all interval values in the propulsion index range of the propulsion section, and establishing a propulsion section interval track set; s502, calling the pushing section interval track set, traversing each numbered interval sequence according to the pushing index direction, judging whether the change direction of the interval value in the sequence is continuously consistent or not, and obtaining an extension abnormal number list; S503, extracting all time indexes of each number in the propulsion section according to the extended abnormal number list, merging index values into number index groups according to time sequence, aggregating all number index groups, and counting abnormal time distribution frequency to obtain a loading system fault prediction result.
9. The AI-based loading system equipment failure prediction method of claim 8, wherein the range of propulsion indexes refers to a propulsion section defined in terms of a sequence and a time sequence of a propulsion process.

Description

AI-based loading system equipment fault prediction method Technical Field The invention relates to the technical field of machine learning, in particular to an AI-based loading system equipment fault prediction method. Background The technical field of machine learning relates to the technology that an algorithm is utilized to enable a computer system to autonomously learn from historical or real-time data and establish a mathematical model so as to realize the functions of prediction, classification, identification and the like, and the core matters comprise acquisition and preprocessing of multi-source data, feature expression, model training and optimization, reasoning and decision making processes and generalization capability improvement of the algorithm. The field emphasizes the processing capability of the model under the background of unstructured, nonlinear and multidimensional data, is widely applied to technical directions of fault detection, predictive maintenance, image recognition, voice recognition, intelligent manufacturing and the like, and plays an important role in improving the intelligent degree of a system, reducing manual intervention and optimizing resource allocation. The traditional fault prediction method of the loading system equipment is characterized in that the operation state of key equipment such as a conveying unit, a stacking robot, an automatic loading machine and the like in the automatic loading system of the train is pointed by a pointer, single working condition data such as current, vibration, temperature, pressure and the like are monitored through a layout sensor, and a mode of identifying typical fault risks such as conveyor line clamping stagnation, hydraulic system leakage, motor bearing abrasion, stacking mechanism barriers and the like by using a set threshold rule or expert experience model is generally carried out by periodical overhaul or maintenance after faults. In the prior art, the judgment of the running state of equipment is only carried out on static monitoring of single-point signals, the dynamic response association among multiple signals in the propulsion process is ignored, a complete chain with time sequence evolution and structural linkage cannot be constructed, so that a fault propagation path among the equipment is difficult to identify, the front symptoms of abnormal behaviors cannot be captured, a modeling mode of time difference response rules is lacking, misjudgment or missed judgment is easy to occur under the background of frequent fluctuation of multi-component coordination actions, a structural extraction mechanism of response offset and jump behaviors among key nodes is lacking, continuous tracking and accurate positioning of complex linkage anomalies are difficult to realize, the granularity of fault identification and the depth of state evaluation of a loading system are further restricted, and the continuity and the interpretability of a prediction result are influenced. Disclosure of Invention In order to achieve the above purpose, the present invention adopts the following technical scheme, and an AI-based loading system equipment failure prediction method includes the following steps: S1, collecting load deformation signals of a roller drive axle at the tail part of a conveying structure of train loading equipment and driving current change records of lifting supporting points of a stacking grabbing support, dividing the segments according to the propulsion beats, executing load and current first-order difference on each segment, judging the direction consistency, and summarizing and sequencing to generate a load linkage segment table; s2, based on the load linkage segment table, extracting control trigger and response time of a propulsion component, matching same-number records, calculating time intervals, generating a number interval sequence, merging same-direction change numbers, and forming a control response difference sequence set; S3, splicing the control response difference sequence set and the load linkage segment table, correspondingly extracting load and interval difference points, calculating a synchronous offset position, identifying repeated path numbers, marking indexes, summarizing a cross coverage relationship, and generating a synchronous response offset layer structure; s4, according to the synchronous response offset layer structure, searching repeated numbers, extracting offset indexes and positions, and identifying load and interval difference reverse switching points to form a jump mapping boundary diagram; and S5, searching whether the interval offset direction in the control response difference sequence set extends continuously or not according to the jump mapping boundary diagram, marking an abnormal number, aggregating a time index, and outputting a fault prediction result of the loading system. As a further scheme of the invention, the load linkage segment table comprises a structure number sequence