CN-121805756-B - Power equipment fault on-line monitoring system for polyester staple fiber production

CN121805756BCN 121805756 BCN121805756 BCN 121805756BCN-121805756-B

Abstract

The invention relates to the technical field of industrial equipment monitoring, in particular to an on-line power equipment fault monitoring system for polyester staple fiber production, which comprises a feature extraction module, a dynamic analysis module, an evolution prediction module, an abnormality identification module, a feature construction module and a diagnosis matching module. The system collects the operation data of the power equipment in real time, extracts time sequence characteristics, calculates dynamic offset vectors, predicts the state evolution track and identifies preliminary anomalies. The feature construction module can backtrack and extract feature fragments associated with the abnormality, obtain fault component signals with different physical meanings through multi-level signal deconstructing, input a fault feature map generator and construct a three-dimensional fault feature map reflecting a fault evolution rule. The diagnosis matching module determines a specific fault type and an evolution stage thereof by performing pattern matching on the three-dimensional map and a preset fault knowledge base. The invention realizes early discovery of fault germination, accurate identification of fault types and dynamic judgment of fault evolution stages.

Inventors

ZHU LIHUA
XIE MING
ZHANG YONG
Xu Yangxiong
ZHANG QINGXI

Assignees

宿迁逸达新材料有限公司

Dates

Publication Date: 20260512
Application Date: 20260309

Claims (7)

1. An on-line monitoring system for power equipment faults in the production of polyester staple fibers, the system comprising: The characteristic extraction module is used for collecting the running state data of the power equipment in the polyester staple fiber production line in real time, extracting the multidimensional time sequence characteristics and generating a time sequence characteristic sequence representing the running state of the power equipment; The dynamic analysis module calculates a dynamic offset vector between the running state and the standard running state of the power equipment according to the time sequence characteristic sequence; the evolution prediction module is used for predicting an evolution track of the running state of the power equipment at the future moment by using the dynamic offset vector driving fault mode evolution model; The abnormality identification module is used for generating multidimensional abnormal signal thresholds covering different fault types based on the running state evolution track, comparing the time sequence characteristic sequence with the multidimensional abnormal signal thresholds in real time, and identifying and positioning preliminary fault abnormal signals; The feature construction module is used for backtracking the preliminary fault abnormal signals into the time sequence feature sequence, extracting complete feature fragments related to the preliminary fault abnormal signals, performing multi-level signal deconstruction to obtain fault component signals in different physical meanings, inputting the fault component signals into a fault feature map generator, and constructing a three-dimensional fault feature map reflecting a fault evolution rule; The diagnosis matching module is used for carrying out pattern matching according to the three-dimensional fault characteristic map and a preset fault knowledge base, and determining the specific fault type and the evolution stage of the power equipment; The multi-dimensional time sequence feature extraction is carried out to generate a time sequence feature sequence representing the running state of the power equipment, and the method specifically comprises the following steps: Respectively intercepting a voltage data section, a current data section, a power factor data section and a temperature data section of a fixed time window from an original running state data stream; carrying out frequency spectrum analysis on the voltage data segment, and extracting fundamental wave amplitude, main harmonic distortion rate and voltage unbalance degree characteristics; carrying out time domain and frequency domain joint analysis on the current data segment, and extracting the current effective value change rate, the peak factor and the current harmonic total distortion characteristic; Carrying out statistical analysis on the power factor data segment, and extracting the power factor mean value, fluctuation variance and mutation frequency characteristics; trend analysis is carried out on the temperature data segment, and the temperature rising slope, the local extremum point and the temperature rising rate characteristic are extracted; arranging all the extracted features according to the time window sequence, and combining to form the time sequence feature sequence; The method for predicting the running state evolution track of the power equipment at the future moment by using the dynamic offset vector to drive the fault mode evolution model specifically comprises the following steps: inputting the current dynamic offset vector to a pre-trained fault mode evolution model; the fault mode evolution model carries out multi-step iterative deduction by taking the current dynamic offset vector as an initial state according to an evolution rule learned by historical fault data; each iteration deduction predicts a dynamic offset vector after a time step in the future; Connecting dynamic offset vectors predicted by a plurality of continuous time steps in time sequence to form an operation state evolution track of the power equipment at the future moment; The running state evolution track describes the path and degree of the running state of the power equipment deviating from the standard state in a future period of time; The step of backtracking the preliminary fault abnormal signal to the time sequence feature sequence, and extracting the complete feature fragment associated with the preliminary fault abnormal signal specifically comprises the following steps: according to the time point information contained in the preliminary fault abnormal signal, forward backtracking the historical time with the set length; Intercepting all feature data from a backtracking starting point to a current time point from the time sequence feature sequence; Taking the feature dimension triggering the preliminary abnormal mark as a core, and extracting complete numerical sequences of all relevant dimensions in the historical feature data; performing time alignment and standardization processing on the extracted complete numerical sequences of each dimension; and integrating the processed dimensional value sequences to form the complete characteristic fragment reflecting the whole process from latency to manifestation of the abnormality.
2. The system for on-line monitoring of faults of electrical equipment for producing polyester staple fibers according to claim 1, wherein the real-time acquisition of the operation state data of the electrical equipment in the polyester staple fiber production line specifically comprises: Deploying a sensor array at a power supply input end, a key load node and a control loop of the power equipment; Synchronously acquiring three-phase voltage instantaneous values, three-phase current instantaneous values, power factor instantaneous values and temperature instantaneous values of key nodes of the power equipment through the sensor array; performing time stamp alignment and sampling rate unification on the collected three-phase voltage instantaneous value, three-phase current instantaneous value, power factor instantaneous value and temperature instantaneous value; And integrating the data subjected to the time stamp alignment and sampling rate unification to form an original running state data stream with time as an index.
3. The system for on-line monitoring of faults in electrical equipment for the production of polyester staple fibers according to claim 2, wherein the dynamic offset vector between the operating state and the standard operating state of the electrical equipment is calculated according to the time sequence characteristic sequence, and specifically comprises: Invoking a standard characteristic sequence of the power equipment pre-stored in a database under a standard healthy running state; comparing the current feature vector in the time sequence feature sequence with the standard feature vector of the corresponding time point in the standard feature sequence element by element; calculating the difference value of the current feature vector and the standard feature vector in each feature dimension; taking the difference values of all feature dimensions as components to construct a multidimensional vector, wherein the multidimensional vector is the dynamic offset vector; The module length of the dynamic offset vector reflects the overall offset degree, and the direction reflects which type or types of operation characteristics are abnormally offset.
4. An on-line monitoring system for faults in electrical equipment for the production of polyester staple fibers according to claim 3, characterized in that it generates multidimensional anomaly signal thresholds covering different fault types based on the running state evolution trajectory, comprising in particular: extracting the maximum value of the predicted future dynamic offset vector in each characteristic dimension from the running state evolution track; for each known fault type, determining a typical offset range of the fault type induced in each characteristic dimension according to historical data of the fault type; combining the predicted maximum value of the future dynamic offset vector and typical offset ranges of different fault types, and calculating a comprehensive early warning upper limit threshold value for each characteristic dimension; According to the sensitivity difference of different feature dimensions in various faults, different weight coefficients are distributed for the early warning upper limit threshold; finally, a multi-dimensional abnormal signal threshold set formed by a plurality of weighted characteristic dimension thresholds is generated.
5. The on-line monitoring system for faults of electrical equipment for producing polyester staple fibers according to claim 4, wherein the real-time comparison of the time sequence characteristic sequence with the multi-dimensional abnormal signal threshold value is performed, and preliminary fault abnormal signals are identified and positioned, and specifically comprises: acquiring a feature vector of the latest time window from the time sequence feature sequence; comparing each characteristic value in the characteristic vector of the latest time window with an early warning upper limit threshold value of the corresponding characteristic dimension in the multidimensional abnormal signal threshold value set; if a certain characteristic value exceeds a corresponding early warning upper limit threshold value, triggering a preliminary abnormal mark of the characteristic dimension; recording the characteristic dimension triggering the preliminary abnormal mark and the specific numerical value and the time point of overrun of the characteristic dimension triggering the preliminary abnormal mark; And packaging the group of preliminary abnormal marks triggered simultaneously and the related information thereof into one preliminary fault abnormal signal.
6. The system for on-line monitoring of faults of electrical equipment for producing polyester staple fibers according to claim 5, wherein the system for performing multi-level signal deconstructing to obtain fault component signals in different physical meanings comprises: treating the complete signature fragment as a composite multidimensional time series signal; an empirical mode decomposition method is applied to adaptively decompose the time sequence signal of each dimension into a series of eigenmode function components and a residual component; Carrying out Hilbert transformation on the eigenvalue function components obtained by decomposition to obtain the time-frequency distribution characteristics of the eigenvalue function components; According to the similarity of time-frequency distribution characteristics, the eigenvector components from different original characteristic dimensions are recombined to form fault component signals respectively representing trend, periodic fluctuation, impact event and random noise; each of the fault component signals carries information of a physical process of a certain aspect in the original composite signal.
7. The on-line monitoring system for faults of electrical equipment for producing polyester staple fibers according to claim 6, wherein the fault component signals are input to a fault feature map generator to construct a three-dimensional fault feature map reflecting the evolution rule of faults, and specifically comprising: Taking time as a first dimension, the type of the fault component signal as a second dimension, and the amplitude, energy or frequency characteristic of the fault component signal as a third dimension; taking the calculated characteristic value of each fault component signal under the corresponding time point and signal type as a data point in a three-dimensional space; Arranging and connecting data points calculated by all fault component signals in a three-dimensional space according to the relation between time sequence and signal types; Fitting discrete data points into a continuous three-dimensional curved surface or voxel model by a spatial interpolation algorithm, wherein the three-dimensional curved surface or voxel model is the three-dimensional fault characteristic map; the three-dimensional fault characteristic map intuitively displays the interrelation and structural characteristics of the time evolution of different fault components and is used for matching with the fault mode map in the preset fault knowledge base so as to determine the specific fault type and the evolution stage of the specific fault type.

Description

Power equipment fault on-line monitoring system for polyester staple fiber production Technical Field The invention relates to the technical field of industrial equipment monitoring, in particular to an on-line monitoring system for power equipment faults in polyester staple fiber production. Background Currently, on-line monitoring of industrial power equipment generally adopts a single-point alarm based on a fixed threshold value or an abnormality detection method based on a statistical model. Such techniques typically set static safe operating boundaries for single or multi-dimensional time series data such as vibration, temperature, current, etc., and trigger an alarm once the monitored data crosses the boundaries. In addition, some schemes use machine learning models to classify extracted features to determine whether a device is in a "normal" or "malfunctioning" state. The core drawback of these prior art techniques is that they analyze anomalies that remain in the data appearance, the alarm mechanism is relatively late, and the granularity of the alarm information generated is coarse. When an abnormality is detected, the equipment often enters an obvious fault state, and early and weak fault germination cannot be effectively identified. Meanwhile, a single fault label cannot inform operation and maintenance personnel of the specific type of the fault, the root cause component and how the fault is developing, so that timeliness of early warning and pertinence of maintenance guidance are insufficient. The conventional technology lacks the capability of deep analysis and characterization of fault physical processes and dynamic evolution rules thereof. The fault of the power equipment is a dynamic process containing interleaving change of various physical quantities, and the characteristics extracted by the existing method are usually statistical abstract or frequency domain energy of signals, so that characteristic components corresponding to different fault mechanisms cannot be effectively separated and highlighted. This results in a low degree of discrimination between different failure modes, which is prone to false positives and false negatives. More importantly, the complete evolution track of the fault from germination and development to degradation cannot be described, so that the predictive maintenance lacks accurate basis for judging the fault development stage and the residual usable life. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides an on-line monitoring system for faults of power equipment for producing polyester staple fibers. In order to achieve the purpose, the invention adopts the following technical scheme that the power equipment fault on-line monitoring system for polyester staple fiber production comprises: The characteristic extraction module is used for collecting the running state data of the power equipment in the polyester staple fiber production line in real time, extracting the multidimensional time sequence characteristics and generating a time sequence characteristic sequence representing the running state of the power equipment; The dynamic analysis module calculates a dynamic offset vector between the running state and the standard running state of the power equipment according to the time sequence characteristic sequence; the evolution prediction module is used for predicting an evolution track of the running state of the power equipment at the future moment by using the dynamic offset vector driving fault mode evolution model; The abnormality identification module is used for generating multidimensional abnormal signal thresholds covering different fault types based on the running state evolution track, comparing the time sequence characteristic sequence with the multidimensional abnormal signal thresholds in real time, and identifying and positioning preliminary fault abnormal signals; The feature construction module is used for backtracking the preliminary fault abnormal signals into the time sequence feature sequence, extracting complete feature fragments related to the preliminary fault abnormal signals, performing multi-level signal deconstruction to obtain fault component signals in different physical meanings, inputting the fault component signals into a fault feature map generator, and constructing a three-dimensional fault feature map reflecting a fault evolution rule; And the diagnosis matching module is used for carrying out pattern matching according to the three-dimensional fault characteristic map and a preset fault knowledge base, and determining the specific fault type and the evolution stage of the power equipment. As a further scheme of the invention, the real-time acquisition of the operation state data of the electric equipment in the polyester staple fiber production line specifically comprises the following steps: Deploying a sensor array at a power supply input end, a key load node and a control loo