Rotating machine fault diagnosis method and system based on time sequence large model

Abstract

The invention relates to the technical field of mechanical fault diagnosis, in particular to a rotating machine fault diagnosis method and system based on a time sequence large model, and the method comprises the following steps: extracting a time sequence value of a vibration signal, measuring the frequency and amplitude fluctuation of a time period, and recognizing an extreme point based on the operation data of a rotating machine; and calculating the fluctuation amplitude difference between the extreme point and the adjacent segment, measuring the time interval between the amplitude peak values, and generating the frequency change segment amplitude. According to the method, the time sequence value of the vibration signal is extracted, frequency and amplitude fluctuation extreme point recognition and adjacent segment amplitude difference calculation are combined, the fine-grained analysis capability of the dynamic characteristics is improved, key frequency point time distribution and amplification value change are detected, and amplitude fluctuation time interval and frequency amplification correlation analysis is combined; deep analysis of frequency and amplitude interaction characteristics, dynamic change characteristic positioning of an interaction ratio and rate quantification are realized, and diagnosis accuracy and evaluation depth are comprehensively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault diagnosis, and particularly to a method and system for diagnosing faults of rotating machinery based on a large time series model. Background Art

[0002] The technical field of mechanical fault diagnosis includes various methods and systems for detecting, analyzing, diagnosing, and evaluating the operating state of mechanical equipment. The core of this technical field lies in monitoring and analyzing characteristic data such as vibration, noise, and temperature during the operation of mechanical equipment to determine whether there are potential faults in the mechanical equipment and to identify and evaluate the fault location, fault type, and fault degree. Overall, the technical field of mechanical fault diagnosis covers links such as sensor data acquisition and processing, feature extraction, fault model establishment and analysis, and diagnostic decision support, providing data support and decision-making basis for the maintenance and management of mechanical equipment by combining modern information technology and mechanical engineering knowledge.

[0003] Among them, the method for diagnosing faults of rotating machinery based on a large time series model refers to a method that uses time series data analysis technology to model and identify fault states in the state data during the operation of rotating machinery. This patent theme aims at the large amount of time series data generated during the operation of rotating machinery. By establishing a suitable large time series model, it captures the characteristic patterns in the data and correlates the fault characteristics under different operating states. Specifically, this method uses a deep learning-based time series model to segment and analyze the collected vibration signal data to extract key time series features in the data, and further classifies and identifies these features to achieve the diagnosis and identification of mechanical fault types and locations.

[0004] In the prior art, when analyzing the operating state of machinery, the extraction and processing of dynamic characteristic data mostly rely on a single data dimension or static rules, and it is difficult to adapt to the diversity of the dynamic interaction characteristics between frequency and amplitude under complex working conditions of rotating machinery. The correlation between frequency fluctuations and amplitude changes is often ignored, restricting the in-depth analysis of fault signals and the extraction of correlation laws. For characteristics such as the change trend of frequency extreme points and the time interval distribution of amplitude fluctuations, the prior art lacks effective extraction and quantification means and is difficult to comprehensively reveal the dynamic fault characteristics in the operation of complex machinery. In addition, in the processing of time series data, the prior diagnostic methods often stay at simple trend analysis or discrete feature extraction and fail to fully capture the high-frequency change patterns in vibration signals and their coupling relationship with fault characteristics. This deficiency may lead to missed detection or misjudgment of key fault characteristics. Especially in the case of high-complexity rotating machinery working conditions, there may be a situation where the accuracy of fault assessment results decreases and the dynamic diagnostic ability is insufficient, limiting its practicality and scalability in refined fault diagnosis. Summary of the Invention

[0005] The object of the present invention is to solve the drawbacks existing in the prior art, and a rotating machinery fault diagnosis method and system based on a time series large model are proposed.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A rotating machinery fault diagnosis method based on a time series large model, comprising the following steps:

[0008] S1: Based on the operation data of the rotating machinery, extract the time series values of the vibration signal, measure the frequency and amplitude fluctuations in the measurement time period, identify the extreme points, calculate the amplitude fluctuation difference between the extreme points and adjacent segments, measure the time interval between the amplitude peaks, and generate the segmented amplitude of the frequency change;

[0009] S2: Based on the segmented amplitude of the frequency change, detect the time distribution of the key frequency points, calculate the increase value of the frequency extreme points in the time period, analyze the change trend of the extreme points, extract the time interval characteristics of the amplitude fluctuation, match the frequency increase and the amplitude fluctuation data, and generate the key frequency increase characteristic value;

[0010] S3: Based on the key frequency increase characteristic value, analyze the amplitude characteristics of the key points of the frequency change, calculate the time span of the frequency change and the amplitude trend direction, analyze the matching relationship between the frequency change and the amplitude direction, extract the characteristics of the interactive change of the frequency and the amplitude, and generate the frequency-amplitude interaction ratio;

[0011] S4: Based on the frequency-amplitude interaction ratio, analyze the change characteristics of the interaction ratio in the time series, locate the peak interval of the ratio change, identify the change range of the frequency intensity, analyze the change trend of the ratio, and generate the interaction ratio characteristic interval;

[0012] S5: Based on the interaction ratio characteristic interval, analyze the time distribution of the frequency intensity change and the amplitude characteristics, calculate the time span data of the frequency intensity and the amplitude characteristics, extract the change rates of the frequency and the amplitude, and generate the change rate of the rotating machinery fault characteristics.

[0013] Optionally, the segmented amplitude of the frequency change includes the frequency fluctuation range, the amplitude fluctuation difference, the time interval between the amplitude peaks, and the position distribution of the extreme points. The key frequency increase characteristic value includes the time distribution of the key frequency points, the increase value of the frequency extreme points, the time interval characteristics of the amplitude fluctuation, the change trend of the frequency increase, and the frequency-amplitude matching characteristics. The frequency-amplitude interaction ratio includes the amplitude characteristic parameters, the time span of the frequency change, the amplitude trend direction, and the interactive change characteristic value of the frequency and the amplitude. The interaction ratio characteristic interval includes the peak range of the ratio change, the change range of the frequency intensity, and the distribution of the ratio change trend. The change rate of the rotating machinery fault characteristics includes the time distribution of the frequency intensity, the time span of the amplitude characteristics, the frequency change rate, and the amplitude change rate.

[0014] Optionally, based on the operating data of the rotating machinery, extract the time series values of the vibration signal, measure the frequency and amplitude fluctuations in the measurement time period, identify the extreme points, calculate the amplitude fluctuation differences between the extreme points and adjacent segments, measure the time intervals between the amplitude peaks. The specific steps for generating the segmented amplitude values of the frequency change are as follows:

[0015] S101: Based on the operating data of the rotating machinery, extract the time series values of the vibration signal, segment the time series according to the measurement time period, convert the data into the change information of frequency and amplitude, and generate the segmented fluctuation results of frequency and amplitude;

[0016] S102: Based on the segmented fluctuation results of the frequency and amplitude, identify the extreme points in the frequency and amplitude, extract the time positions of each extreme point and the amplitude fluctuation ranges of adjacent time periods, calculate the fluctuation differences between each extreme point and its adjacent segment, and generate the fluctuation difference results between the extreme points and adjacent segments;

[0017] S103: Based on the fluctuation difference results between the extreme points and adjacent segments, analyze the time intervals between the extreme points, combine the fluctuation amplitudes of the frequency changes within the time intervals, classify and organize them according to the time periods, and generate the segmented amplitude results of frequency and amplitude.

[0018] Optionally, based on the segmented amplitude values of the frequency change, detect the time distribution of the key frequency points, calculate the increase values of the frequency extreme points within the time period, analyze the change trends of the extreme points, extract the time interval characteristics of the amplitude fluctuations, match the frequency increase and amplitude fluctuation data, and the specific steps for generating the key frequency increase characteristic values are as follows:

[0019] S201: Based on the segmented amplitude values of the frequency change, set and extract the time distribution of the key frequency points, extract the segmented frequency values for each frequency point, and calculate the change ranges within the differential time periods, screen the data with significant frequency change amplitudes, and organize the distribution of the key frequency points according to the time sequence to generate the key frequency time distribution results;

[0020] S202: Based on the key frequency time distribution results, calculate the increase values of the frequency extreme points within the time period, extract the amplitude data, classify the amplitude fluctuation ranges, analyze the amplitude values of the extreme point changes and the time intervals of the amplitude fluctuations in each time period, and identify the change trends of the extreme points according to the increase change rules between adjacent time periods to generate the extreme point change and time interval characteristic results;

[0021] S203: Based on the extreme point change and time interval characteristic results, segmentally match the frequency increase values and the amplitude fluctuation data, establish the corresponding relationship between the data after matching according to the increase and fluctuation values within the time interval, extract the amplitude increase characteristics of the key frequency points within the interval, and generate the key frequency increase characteristic values.

[0022] Optionally, the specific formula for calculating the amplification value is as follows:

[0023]

[0024] Wherein, Z represents the amplification value, F j represents the frequency extreme point in the j-th time period, G j-1 represents the frequency extreme point in the previous time period adjacent to the j-th time period, B j represents the amplitude extreme value in the j-th time period, C j-1 represents the amplitude extreme value in the previous time period adjacent to the j-th time period, R j represents the time position in the j-th time period, S j-1 represents the time position in the previous time period adjacent to the j-th time period, and m represents the total number of time periods.

[0025] Optionally, based on the key frequency amplification characteristic value, analyze the amplitude characteristics of the key points of frequency change, calculate the time span of frequency change and the amplitude trend direction, analyze the matching relationship between frequency change and amplitude direction, extract the characteristics of the interactive change between frequency and amplitude, and the specific steps for generating the frequency-amplitude interaction ratio are as follows:

[0026] S301: Based on the key frequency amplification characteristic value, extract the amplitude characteristics of the key points of frequency change, extract the amplitude value range of each key frequency point, match the extraction results with the frequency amplification value item by item, calculate the amplitude change range and direction characteristic value, and generate the frequency change amplitude characteristic result;

[0027] S302: Based on the frequency change amplitude characteristic result, analyze the time span data of the frequency change points, calculate the time difference between the frequency change points, and extract the amplitude trend direction of the corresponding time interval. Classify and match the frequency amplification direction and amplitude trend characteristics according to the time interval, judge the correlation change trend between the two, extract the data that meets the matching conditions, and organize them into interactive information to generate the frequency-amplitude matching relationship result;

[0028] S303: Based on the frequency-amplitude matching relationship result, extract the frequency change and amplitude interaction characteristic data, group them by time span interval, calculate the ratio of frequency and amplitude change, summarize the distribution law of the ratio, and extract the characteristics to generate the frequency-amplitude interaction ratio.

[0029] Optionally, the specific formula for calculating the time-weighted difference between frequency change points is as follows:

[0030]

[0031] Wherein, T Δf represents the time-weighted difference between frequency change points, t wThe time stamp representing the w-th frequency change point, t w+1 The time stamp representing the (w + 1)-th frequency change point, |t w+1 -t w | represents the absolute value of the time difference between two adjacent frequency change points, f w represents the frequency value of the w-th frequency change point, f w+1 represents the frequency value of the (w + 1)-th frequency change point, |f w+1 -f w | represents the absolute value of the frequency difference between two adjacent frequency change points, represents the square root of the absolute value of the frequency difference, W represents the total number of frequency change points.

[0032] Optionally, based on the frequency-amplitude interaction ratio, analyze the change characteristics of the interaction ratio within the time series, locate the peak interval of the ratio change, identify the change range of the frequency intensity, analyze the change trend of the ratio, and the specific steps to generate the interaction ratio characteristic interval are as follows:

[0033] S401: Based on the frequency-amplitude interaction ratio, extract the interaction ratio data in the time series, calculate the change amplitude data of the ratio, compare the increasing and decreasing trends of the values in the change amplitude, mark the peak positions, sort out the interval range of the ratio peaks, record the start and end times, and generate the interaction ratio peak interval result;

[0034] S402: Based on the interaction ratio peak interval result, extract the frequency intensity data within the peak interval, sort the frequency intensity values within the interval, calculate the difference value between the maximum and minimum values, compare the amplitude characteristics of the frequency intensity change within the interval, and generate the frequency intensity change range result;

[0035] S403: Based on the frequency intensity change range result, analyze the change trend of the interaction ratio within each interval, compare the increasing and decreasing directions of the interaction ratio within different time intervals, reclassify the interval characteristics by combining the time span and the frequency intensity amplitude, and summarize the characteristic distribution of the interaction ratio in the time series to generate the interaction ratio characteristic interval.

[0036] Optionally, based on the interaction ratio characteristic interval, analyze the time distribution of the frequency intensity change and the amplitude characteristic, calculate the time span data of the frequency intensity and the amplitude characteristic, extract the change rate of the frequency and the amplitude, and the specific steps to generate the change rate of the rotating machinery fault characteristics are as follows:

[0037] S501: Based on the interaction ratio characteristic interval, extract the frequency intensity and amplitude characteristic data within each time interval, analyze the distribution characteristics of the frequency intensity data, identify the change law of the distribution values of the amplitude characteristic and the frequency intensity according to the time period, and sort the data within the time period into corresponding distribution records in sequence to generate the frequency and amplitude time distribution result;

[0038] S502: Based on the frequency and amplitude time distribution results, calculate the time span value of the frequency intensity and amplitude characteristic data within a time period, compare the length of the time intervals of the frequency intensity and amplitude changes within different time periods, and generate frequency and amplitude time span data;

[0039] S503: Based on the frequency and amplitude time span data, calculate the change rates of the frequency intensity and amplitude characteristics within a time interval, extract the matching data of the frequency change rate and the amplitude change rate, classify and organize the rate characteristic data by time period, and generate the change rate of the rotating machinery fault characteristics.

[0040] A rotating machinery fault diagnosis system based on a time series large model includes:

[0041] The vibration characteristic extraction module extracts the time series values of the vibration signal based on the operation data of the rotating machinery, measures the frequency and amplitude fluctuations in a time period, calculates the amplitude difference between the extreme points and the adjacent segments, determines the time interval between the amplitude peaks, and generates the segmented amplitude of the frequency change;

[0042] The key frequency analysis module detects the time distribution of the key frequency points based on the segmented amplitude of the frequency change, calculates the increase value of the frequency extreme points within a time period, extracts the time interval characteristics of the amplitude fluctuations, matches the frequency increase and amplitude fluctuation data, and generates the key frequency increase characteristic value;

[0043] The amplitude matching module analyzes the amplitude characteristics of the key points of the frequency change based on the key frequency increase characteristic value, calculates the time span of the frequency change and the amplitude trend direction, analyzes the matching relationship between the frequency change and the amplitude direction, and generates the frequency-amplitude interaction ratio;

[0044] The ratio trend analysis module analyzes the change characteristics of the interaction ratio within the time series based on the frequency-amplitude interaction ratio, identifies the change range of the frequency intensity, analyzes the ratio change trend, and generates the interaction ratio characteristic interval;

[0045] The fault rate determination module analyzes the time distribution of the frequency intensity change and the amplitude characteristics based on the interaction ratio characteristic interval, calculates the time span data of the frequency intensity and the amplitude characteristics, and generates the change rate of the rotating machinery fault characteristics.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0047] In the present invention, by extracting the time series values of vibration signals, combining the identification of extreme points of frequency and amplitude fluctuations and the calculation of amplitude differences between adjacent segments, the fine-grained analysis ability of dynamic characteristics is improved. The detection of the time distribution of key frequency points and the change of amplitude increase value, combined with the correlation analysis of the amplitude fluctuation time interval and frequency increase, realizes the in-depth analysis of the interaction characteristics of frequency and amplitude, locates the dynamic change characteristics of the interaction ratio, and quantifies the rate, comprehensively improving the accuracy of diagnosis and the depth of evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a schematic flowchart of the steps of the present invention;

[0050] Figure 2 It is a flowchart of step S1 of the present invention;

[0051] Figure 3 It is a flowchart of step S2 of the present invention;

[0052] Figure 4 It is a flowchart of step S3 of the present invention;

[0053] Figure 5 It is a flowchart of step S4 of the present invention;

[0054] Figure 6 It is a flowchart of step S5 of the present invention;

[0055] Figure 7 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The following describes the technical solutions in the present invention with reference to the drawings.

[0057] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or design solutions. Exactly, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either of the two can be selected.

[0058] In the embodiments of the present invention, the terms "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, their intended meanings are the same. The terms "of", "corresponding", and "corresponding to" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, their intended meanings are the same.

[0059] In the embodiments of the present invention, sometimes subscripts such as W 1 may be written in a non-subscript form such as W1. When the difference between them is not emphasized, their intended meanings are the same.

[0060] To make the technical problems, technical solutions, and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0061] As Figure 1 shown, a method for diagnosing faults in rotating machinery based on a time-series large model includes the following steps:

[0062] S1: Based on the operating data of the rotating machinery, extract the time-series values of the vibration signal, measure the frequency and amplitude fluctuations in the measurement time period, identify the extreme points, calculate the amplitude fluctuation differences between the extreme points and adjacent segments, determine the time intervals between the amplitude peaks, and generate the segmented amplitude values of the frequency change.

[0063] S2: Based on the segmented amplitude values of the frequency change, detect the time distribution of the key frequency points, calculate the increase values of the frequency extreme points in the time period, analyze the change trends of the extreme points, extract the time interval characteristics of the amplitude fluctuations, match the frequency increase and amplitude fluctuation data, and generate the key frequency increase characteristic values.

[0064] S3: Based on the key frequency increase characteristic values, analyze the amplitude characteristics of the key points of the frequency change, calculate the time span of the frequency change and the amplitude trend direction, analyze the matching relationship between the frequency change and the amplitude direction, extract the characteristics of the interactive change between the frequency and the amplitude, and generate the frequency-amplitude interaction ratio.

[0065] S4: Based on the frequency-amplitude interaction ratio, analyze the change characteristics of the interaction ratio within the time series, locate the peak interval of the ratio change, identify the change range of the frequency intensity, analyze the change trend of the ratio, and generate the characteristic interval of the interaction ratio.

[0066] S5: Based on the characteristic interval of the interaction ratio, analyze the time distribution of the frequency intensity change and the amplitude characteristics, calculate the time span data of the frequency intensity and the amplitude characteristics, extract the change rates of the frequency and the amplitude, and generate the change rate of the fault characteristics of the rotating machinery.

[0067] The segmented amplitude of frequency change includes the frequency fluctuation range, the amplitude fluctuation amplitude difference, the time interval between amplitude peaks, and the position distribution of extreme points. The key frequency increase characteristic values include the time distribution of key frequency points, the increase value of frequency extreme points, the amplitude fluctuation time interval characteristic, the frequency increase change trend, and the frequency-amplitude matching characteristic. The frequency-amplitude interaction ratio includes amplitude characteristic parameters, the time span of frequency change, the amplitude trend direction, and the frequency-amplitude interaction change characteristic value. The interaction ratio characteristic interval includes the peak range of ratio change, the frequency intensity change range, and the ratio change trend distribution. The change rate of the rotating machinery fault characteristic includes the time distribution of frequency intensity, the time span of amplitude characteristics, the frequency change rate, and the amplitude change rate.

[0068] As Figure 2 shown, the specific steps of S1 are as follows:

[0069] S101: Based on the operating data of the rotating machinery, extract the time series values of the vibration signal, segment the time series according to the measurement time period, convert the data into the change information of frequency and amplitude, and generate the segmented fluctuation results of frequency and amplitude;

[0070] First, obtain the time series values through the vibration signal acquisition device, use a high-precision acceleration sensor or vibration sensor to record the vibration characteristics during mechanical operation, ensure that the data sampling frequency meets the highest frequency requirements of the signal, segment the collected time series signal according to the preset measurement time period, perform spectral analysis on the vibration signals of each time period using the short-time Fourier transform or fast Fourier transform, extract the frequency components and amplitude values of each segment of data in the frequency domain, process the data in different amplitude ranges through the normalization method, thereby generating the change curves of frequency and amplitude, segment the signal according to the change curves and identify the segmented fluctuation characteristics, and finally use the segmented frequency and amplitude fluctuation characteristic results as the processing output.

[0071] S102: Based on the segmented fluctuation results of frequency and amplitude, identify the extreme points in frequency and amplitude, extract the time positions of each extreme point and the amplitude fluctuation range of adjacent time periods, calculate the fluctuation difference between each extreme point and its adjacent segment, and generate the extreme point and adjacent segment fluctuation difference results;

[0072] Through the detection of extreme points for the frequency change value and amplitude change value, analyze the slope of the segmented fluctuation result curve using the differential method, extract the maximum and minimum values of frequency and amplitude in each segment of the curve, record the position of the extreme point in the time series at the same time, combine the position information of the extreme point, calculate the amplitude change range of adjacent time periods, define the amplitude fluctuation range as the amplitude difference between the extreme point and its adjacent area, analyze the difference size between each extreme point and its adjacent segment by defining the difference formula, and finally output the extreme point and adjacent segment fluctuation difference results.

[0073] S103: Analyze the time interval between extreme points based on the result of the difference between the extreme point and the fluctuation of the adjacent segment. Combine the fluctuation amplitude of the frequency change within the time interval, classify and organize it by time period, and generate the frequency and amplitude segmented amplitude result;

[0074] Based on the time interval between extreme points, combine the fluctuation amplitude of the frequency change within the time interval, and according to the formula

[0075]

[0076] Calculate the frequency and amplitude segmented amplitude result.

[0077] In the formula, A represents the segmented amplitude result, n represents the total number of time periods, Δf i is the change in frequency of the i-th segment, T i is the time length of the i-th segment, A max,i and A min,i are the maximum amplitude and the minimum amplitude within the i-th segment respectively.

[0078] Detailed explanation of the formula and the derivation process of the formula calculation:

[0079] In the formula, Δf i is calculated from the frequency information obtained by Fourier transform, and the calculation method is

[0080] Δf i = f max,i - f min,i ;

[0081] where f max,i and f min,i are the maximum frequency and the minimum frequency of the i-th segment respectively; T i is directly obtained from the segmented length of the time series; A max,i and A min,i are obtained by detecting the extreme points of the amplitude fluctuation curve.

[0082] The derivation process is as follows:

[0083] Assume that the time series is segmented into 3 segments, and the time lengths of each segment are T 1 = 5 seconds, T 2 = 4 seconds, T 3 = 6 seconds, the frequency changes are f max,1 = 100Hz, f min,1 = 80Hz, f max,2 = 120Hz, f min,2 = 90Hz, f max,3 = 95Hz, f min,3 = 85Hz, and the corresponding amplitude changes are A max,1 = 10, A min,1 = 5, Amax,2 = 15, A min,2 = 7, A max,3 = 12, A min,3 = 6.

[0084] Calculate the amplitude fluctuation amplitudes of each frequency segment:

[0085] Δf 1 = 100 - 80 = 20, Δf 2 = 120 - 90 = 30, Δf 3 = 95 - 85 = 10;

[0086] |A max,1 - A min,1 | = 10 - 5 = 5, |A max,2 - A min,2 | = 15 - 7 = 8, |A max,3 - A min,3 | = 12 - 6 = 6;

[0087] Substitute into the formula for calculation:

[0088]

[0089] The result shows that the total segmented amplitude is 17.17, indicating that the segmented amplitude comprehensively reflects the fluctuations of frequency and amplitude in each time period, and the calculation result can be used as the input feature parameter for subsequent signal pattern recognition.

[0090] As Figure 3 shown, the specific steps of S2 are as follows:

[0091] S201: Based on the segmented amplitude of frequency change, set and extract the time distribution of key frequency points, extract the segmented frequency values for each frequency point, calculate the change range within the differential time period, screen the data with significant frequency change amplitudes, and organize the distribution of key frequency points according to the time sequence to generate the key frequency time distribution result;

[0092] First, extract the frequency change data contained in the vibration signal. Combine the segmented analysis method to set the range of key frequency points for identification. According to the distribution position of the frequency points in the time series, record the time distribution data of each key frequency point. By taking the point-by-point difference of the vibration signal frequency curve, extract the change range of frequency values in each segment. Use the sliding window method to mark and screen the frequency points with significant changes within the segment to ensure that the identified key frequency points have significant change amplitude characteristics. Classify and organize the data according to the time distribution order of the key frequency points, and finally generate the key frequency time distribution result.

[0093] S202: Based on the key frequency time distribution results, calculate the increase value of the frequency extreme points within the time period, extract the amplitude data, classify the amplitude fluctuation range, analyze the amplitude value of the extreme point change and the time interval of the amplitude fluctuation in each period, and identify the change trend of the extreme point according to the increase change rule between adjacent time periods, generating the results of the extreme point change and the time interval characteristics;

[0094] The specific calculation formula for the increase value is:

[0095]

[0096] Among them, Z represents the increase value, F j represents the frequency extreme point of the j-th time period, G j-1 represents the frequency extreme point of the previous time period adjacent to the j-th time period, B j represents the amplitude extreme value of the j-th time period, C j-1 represents the amplitude extreme value of the previous time period adjacent to the j-th time period, R j represents the time position of the j-th time period, S j-1 represents the time position of the previous time period adjacent to the j-th time period, and m represents the total number of time periods.

[0097] Z represents the increase value, which is used to characterize the comprehensive characteristics of the frequency and amplitude changes within the time period;

[0098] F j represents the frequency extreme point within the j-th time period, and its value is obtained by monitoring the frequency data and extracting the maximum frequency value of this time period;

[0099] G j-1 represents the frequency extreme point within the previous time period adjacent to the j-th time period, and the acquisition method is the same as that of F j which is to take the maximum frequency value of the adjacent time period;

[0100] B j represents the extreme value of the amplitude within the j-th time period, and the maximum amplitude value within this time period is extracted by the vibration monitoring device;

[0101] C j-1 represents the amplitude extreme value within the previous time period adjacent to the j-th time period, and the acquisition method is the same as that of B j which is to take the maximum amplitude value of the adjacent time period;

[0102] R j represents the time position of the j-th time period, usually represented by the starting time point of the time series acquisition;

[0103] S j-1 represents the time position of the previous time period adjacent to the j-th time period, and the calculation method is the same as that of Rj Same;

[0104] m is the total number of time periods, determined according to the partitioning rules of the time series.

[0105] Calculation process:

[0106] Assume that the monitoring data comes from three time periods, and the following parameters are obtained through monitoring:

[0107] Time period 1: F 1 = 75 Hz, G 0 = 60 Hz, B 1 = 15, C 0 = 10, R 1 = 5 seconds, S 0 = 0 seconds;

[0108] Time period 2: F 2 = 90 Hz, G 1 = 75 Hz, B 2 = 20, C 1 = 15, R 2 = 10 seconds, S 1 = 5 seconds;

[0109] Time period 3: F 3 = 85 Hz, G 2 = 90 Hz, B 3 = 25, C 2 = 20, R 3 = 15 seconds, S 2 = 10 seconds.

[0110] Calculation steps:

[0111] Calculate the first segment:

[0112]

[0113] Calculate the second segment:

[0114]

[0115] Calculate the third segment:

[0116]

[0117] Total result:

[0118] Z = 54.09 + 75 + 32.02 = 161.11;

[0119] Result analysis:

[0120] The amplification value is 161.11, representing the comprehensive change characteristics of frequency and amplitude within three time periods. This result is used to analyze the correlated changes of frequency and amplitude over time and provide input data for subsequent identification of the change trend of extreme points.

[0121] S203: Based on the results of the change of extreme points and the characteristics of time intervals, segmentally match the frequency amplification value with the amplitude fluctuation data, establish a corresponding relationship for the data after matching according to the amplification and fluctuation values within the time interval, extract the amplitude amplification characteristics of the key frequency points within the interval, and generate the key frequency amplification characteristic value;

[0122] Based on the frequency amplification data and the amplitude fluctuation values within the time interval, establish a corresponding relationship between the two. Use the cross ratio of the amplitude and frequency changes in each interval as the characteristic value for data correlation. Segmentally organize the extracted characteristic values into the amplitude amplification characteristics within the time period, and conduct a secondary analysis of the amplitude change trend in combination with the frequency characteristic change curve within the time interval. Finally, generate the key frequency amplification characteristic value as the output result.

[0123] As Figure 4 shown, the specific steps of S3 are as follows:

[0124] S301: Based on the key frequency amplification characteristic value, extract the amplitude characteristics of the key points of frequency change, extract the amplitude value range of each key frequency point, match the extraction results with the frequency amplification value item by item, calculate the amplitude change range and the direction characteristic value of the amplitude change, and generate the frequency change amplitude characteristic result;

[0125] First, extract the amplitude characteristic data of each key point of frequency change, record the amplitude change range corresponding to each frequency point in sequence. Through one-by-one matching of the frequency points and amplitude values, associate the amplitude change data with the frequency amplification value one-to-one according to the time period, calculate the amplitude change range of each key frequency point in the associated data, and record the direction characteristic of the change at the same time. The amplitude direction characteristic is determined by judging the positive and negative values of the amplitude change. Finally, organize the frequency change amplitude characteristic result according to the matching results as the output data after calculation.

[0126] S302: Based on the frequency change amplitude characteristic result, analyze the time span data of the frequency change points, calculate the time difference between the frequency change points, and extract the amplitude trend direction of the corresponding time interval. Classify and match the frequency amplification direction and the amplitude trend characteristics according to the time interval, judge the change trend of the correlation between the two, extract the data that meet the matching conditions, and organize them into interaction information to generate the frequency and amplitude matching relationship result;

[0127] The specific formula for the time weighted difference between frequency change points is:

[0128]

[0129] Among them, T Δf represents the time-weighted difference between frequency change points, t w represents the timestamp of the w-th frequency change point, t w+1 represents the timestamp of the (w + 1)-th frequency change point, |t w+1 -t w | represents the absolute value of the time difference between two adjacent frequency change points, f w represents the frequency value of the w-th frequency change point, f w+1 represents the frequency value of the (w + 1)-th frequency change point, |f w+1 -f w | represents the absolute value of the frequency difference between two adjacent frequency change points, represents the square root of the absolute value of the frequency difference, and W represents the total number of frequency change points.

[0130] t w represents the timestamp of the w-th frequency change point. By recording and collecting time series data, the timestamps of frequency change points are obtained, with the unit of seconds. The method of collecting time data can be based on high-precision sensor data recording, and the timestamp of each frequency change point is obtained through the time point when the amplitude fluctuation exceeds the set threshold. For example, the collected data is t 1 = 0.5 seconds, t 2 = 1.1 seconds, t 3 = 2.4 seconds, t 4 = 3.8 seconds.

[0131] f w represents the frequency value of the w-th frequency change point, with the unit of hertz. By monitoring the frequency value corresponding to the amplitude change point through a frequency measuring instrument, the frequency value corresponding to the timestamp point is recorded. The data collection method is to calculate the corresponding instantaneous frequency after separating the specific frequency interval of the signal through a band-pass filter. For example, the collected data is f 1 = 50 hertz, f 2 = 52 hertz, f 3 = 55 hertz, f 4 = 58 hertz.

[0132] W represents the total number of frequency change points. By globally scanning the time series data, all frequency change points are identified to determine the total number. For example, if the number of frequency change points collected here is 4, then W = 4.

[0133] t w+1 -t w represents the time difference between two adjacent frequency change points, which is calculated by subtracting the two timestamps before and after. For example:

[0134] t 2 -t 1= 1.1 - 0.5 = 0.6 seconds

[0135] t 3 -t 2 = 2.4 - 1.1 = 1.3 seconds

[0136] t 4 -t 3 = 3.8 - 2.4 = 1.4 seconds

[0137] f w+1 -f w represents the frequency difference between two adjacent frequency change points, which is calculated by subtracting the two previous frequency values. For example:

[0138] f 2 -f 1 = 52 - 50 = 2 Hz

[0139] f 3 -f 2 = 55 - 52 = 3 Hz

[0140] f 4 -f 3 = 58 - 55 = 3 Hz

[0141] |t w+1 -t w | and |f w+1 -f w | are the absolute values of the time difference and the frequency difference respectively. Since the above time difference and frequency difference are both positive, the absolute value operation does not change their values.

[0142] is the square root of the absolute value of the frequency difference. The calculation is as follows:

[0143]

[0144] Substitute the above results into the formula for item-by-item calculation to calculate the weight term

[0145] The first item:

[0146] The second item:

[0147] The third item:

[0148] Sum up the weight terms:

[0149]

[0150] Substitute the sum result into the formula:

[0151]

[0152] The calculation result is 1.842 seconds.

[0153] This result indicates that the time-weighted difference between frequency change points is 1.842 seconds. The larger the value, the longer the average time interval between frequency change points; the smaller the value, the shorter the time interval between change points, which is closely related to the weighted relationship between the frequency difference and the time span. After further combining this value with the amplitude trend direction, it can be used for subsequent analysis of the correlation characteristics of matching frequency and amplitude change trends.

[0154] S303: Based on the result of the frequency and amplitude matching relationship, extract the data of the interaction characteristics of frequency change and amplitude, group them by time span into intervals, calculate the ratio of frequency change to amplitude change, summarize the distribution law of the ratio, and extract the characteristics to generate the frequency-amplitude interaction ratio;

[0155] By extracting the data of the interaction characteristics of frequency change and amplitude, first group the data by time span into intervals. By analyzing the interaction characteristics of frequency change and amplitude change within each interval, calculate the ratio of frequency change to amplitude change, use the piecewise fitting method to summarize the distribution characteristics of the ratio, extract the characteristic data with significant change rules in the ratio, organize the distribution results of different intervals, and combine the characteristics of frequency and amplitude change trends to conduct a classification analysis of the ratio characteristics, and finally generate the result of the frequency-amplitude interaction ratio.

[0156] As Figure 5 shown, the specific steps of S4 are as follows:

[0157] S401: Based on the frequency-amplitude interaction ratio, extract the interaction ratio data in the time series, calculate the change amplitude data of the ratio, compare the increasing and decreasing trends of the values in the change amplitude, mark the peak positions, organize the interval range of the ratio peaks, and record the start and end times to generate the result of the interaction ratio peak interval;

[0158] First, divide the time series into fixed time intervals, conduct a point-by-point analysis of the interaction ratio data in each interval, calculate the change amplitude of the interaction ratio, determine the increasing and decreasing trends of the ratio by the method of successive differences, record the peak point positions in the ratio change process, and statistically analyze the interval range of each peak in the ratio change process, calculate the start and end times of the interval and the interaction ratio range, and finally generate the result of the interaction ratio peak interval as the data input for subsequent analysis.

[0159] S402: Based on the result of the interaction ratio peak interval, extract the frequency intensity data within the peak interval, sort the frequency intensity values within the interval, and calculate the difference value between the maximum and minimum values, compare the amplitude characteristics of the frequency intensity change within the interval to generate the result of the frequency intensity change range;

[0160] Based on the results of the peak interval of the interaction ratio, calculate the result of the frequency intensity change range according to the formula

[0161]

[0162] In the formula, F represents the result of the frequency intensity change range, q is the number of peak intervals, X

[0163] and Y r are respectively the maximum frequency intensity value and the minimum frequency intensity value in the r-th interval, Z r is the time span of the interval, |X r -Y r -Y r | represents the absolute value of the frequency intensity change.

[0164] In the formula, X r and Y r are obtained by sorting the frequency intensity data within the peak interval to extract the maximum and minimum values; Z r is the time span of the peak interval, calculated by the difference between the start and end times of the interval; |X r -Y r | is the absolute value of the frequency intensity change, describing the change amplitude of the frequency intensity within the interval.

[0165] Derivation process:

[0166] Assume that the data contains three peak intervals, and the data is as follows:

[0167] Interval 1: X 1 =50, Y 1 =30, Z 1 =4 seconds;

[0168] Interval 2: X 2 =70, Y 2 =40, Z 2 =6 seconds;

[0169] Interval 3: X 3 =60, Y 3 =35, Z 3 =5 seconds.

[0170] Calculate the 1st interval:

[0171]

[0172] Calculate the 2nd interval:

[0173]

[0174] Calculate the 3rd interval:

[0175]

[0176] The total result is:

[0177] F = 100 + 150 + 125 = 375;

[0178] Result analysis:

[0179] The result of the frequency intensity change range is 375, indicating the total change amplitude of the frequency intensity within the peak interval. This result can be used to further analyze the influence of the frequency intensity change on the distribution of the interaction ratio characteristics.

[0180] S403: Based on the result of the frequency intensity change range, analyze the change trend of the interaction ratio within each interval, compare the increase and decrease directions of the interaction ratio in different differential time intervals, reclassify the interval characteristics by combining the time span and the frequency intensity amplitude, summarize the characteristic distribution of the interaction ratio in the time series, and generate the interaction ratio characteristic interval;

[0181] First, record and count the change direction of the interaction ratio within each interval, classify and compare the increase and decrease changes of the interaction ratio in different time intervals, calculate the difference value of the interaction ratio between adjacent time intervals, combine the change amplitude of the frequency intensity and the time span within the interval, reclassify the interaction ratio, separately classify the interval with a long time span and a significant frequency intensity amplitude into one category, and finally summarize the characteristic distribution of the interaction ratio in the time series to generate the interaction ratio characteristic interval as the output result.

[0182] As Figure 6 shown, the specific steps of S5 are as follows:

[0183] S501: Based on the interaction ratio characteristic interval, extract the frequency intensity and amplitude characteristic data within each time interval, analyze the distribution characteristics of the frequency intensity data, identify the distribution value change rules of the amplitude characteristic and the frequency intensity according to the time period, sort the data within the time period in order to generate the corresponding distribution record, and generate the frequency and amplitude time distribution result;

[0184] Analyze the frequency intensity data according to the interval distribution characteristics through the sorting method. At the same time, extract the amplitude characteristic value as the distribution range of each time interval, calculate the change trend of the frequency intensity within each time period, identify the distribution change rules of the amplitude characteristic and the frequency intensity within the time period by statistically analyzing the increase and decrease changes of the frequency intensity value, and sort the data of each time period in ascending or descending order to generate the distribution record, and finally generate the frequency and amplitude time distribution result as the output.

[0185] S502: Based on the frequency and amplitude time distribution result, calculate the time span value of the frequency intensity and amplitude characteristic data within the time period, compare the time interval lengths of the frequency intensity and amplitude changes in different differential time periods, and generate the frequency and amplitude time span data;

[0186] Based on the frequency and amplitude time distribution results, according to the formula

[0187]

[0188] calculate the frequency and amplitude time span data.

[0189] In the formula, K represents the frequency and amplitude time span data within the time period, q is the number of time periods, A p and B p are respectively the maximum and minimum values of the frequency intensity in the p-th segment, D p and E p are respectively the maximum and minimum values of the amplitude in the p-th segment, C p and F p are the length of the p-th time interval.

[0190] In the formula, A p and B p are obtained by detecting the maximum and minimum values in the frequency intensity distribution curve within the time period; D p and E p are obtained by extracting the amplitude characteristic values within the time period; C p and F p are the total time span of the time period, which is directly calculated by dividing the time series.

[0191] Derivation process:

[0192] Assume that the time series data is divided into three segments, and the parameter data is as follows:

[0193] Time period 1: A 1 = 80, B 1 = 50, D 1 = 20, E 1 = 10, C 1 = 4 seconds, F 1 = 6 seconds;

[0194] Time period 2: A 2 = 100, B 2 = 60, D 2 = 25, E 2 = 15, C 2 = 5 seconds, F 2 = 7 seconds;

[0195] Time period 3: A 3 = 90, B 3 = 70, D 3 = 30, E 3 = 20, C 3 = 6 seconds, F 3 = 8 seconds.

[0196] Calculation section 1:

[0197]

[0198] Calculation section 2:

[0199]

[0200] Calculation section 3:

[0201]

[0202] The total result is:

[0203] K = 9.17 + 9.43 + 4.58 = 23.18;

[0204] Result analysis:

[0205] The sum of the time-span data is 23.18, which represents the distribution-span characteristics of frequency and amplitude in different time periods. The total span value reflects the relative change range of frequency and amplitude characteristics in different intervals, laying a data foundation for further analysis of the change rate.

[0206] S503: Based on the frequency and amplitude time-span data, calculate the change rate of frequency intensity and amplitude characteristics within the time interval, extract the matching data of frequency change rate and amplitude change rate, classify and organize the rate characteristic data by time period, and generate the change rate of rotating machinery fault characteristics;

[0207] Match the frequency change rate and amplitude change rate of each time interval point by point, and at the same time count their change trends. By comparing the change rules of the increase and decrease directions, classify and organize the rate matching results within the time period. Combine the time-span data and rate change values, reclassify the change rates in different time periods, and summarize and process the matched rate characteristic data according to the characteristics of time distribution. Finally, generate the change rate data result of rotating machinery fault characteristics as the output.

[0208] As Figure 7 shown, a rotating machinery fault diagnosis system based on a time-series large model includes:

[0209] The vibration characteristic extraction module extracts the time-series values of the vibration signal based on the operation data of the rotating machinery, measures the frequency and amplitude fluctuations in the measurement time period, calculates the fluctuation amplitude difference between the extreme points and adjacent segments, determines the time interval between the amplitude peaks, and generates the segmented amplitude values of the frequency change;

[0210] The key frequency analysis module detects the time distribution of key frequency points based on the segmented amplitude of frequency changes, calculates the increase value of the frequency extreme points within the time period, extracts the time interval characteristics of amplitude fluctuations, matches the frequency increase and amplitude fluctuation data, and generates the key frequency increase characteristic value;

[0211] The amplitude matching module analyzes the amplitude characteristics of the key points of frequency change based on the key frequency increase characteristic value, calculates the time span of frequency change and the amplitude trend direction, analyzes the matching relationship between frequency change and amplitude direction, and generates the frequency-amplitude interaction ratio;

[0212] The ratio trend analysis module analyzes the change characteristics of the interaction ratio within the time series based on the frequency-amplitude interaction ratio, identifies the change range of frequency intensity, analyzes the ratio change trend, and generates the interaction ratio characteristic interval;

[0213] The fault rate determination module analyzes the time distribution of the frequency intensity change and the amplitude characteristics based on the interaction ratio characteristic interval, calculates the time span data of the frequency intensity and the amplitude characteristics, and generates the change rate of the fault characteristics of the rotating machinery.

[0214] As mentioned above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claimed rights.

Claims

1. A rotating machinery fault diagnosis method based on a time series large model, characterized in that: The following steps are involved: S1: Based on the operating data of the rotating machinery, extract the time series value of the vibration signal, measure the frequency and amplitude fluctuations of the time period, identify the extreme point, calculate the difference in fluctuation amplitude between the extreme point and the adjacent segment, determine the time interval between the amplitude peaks, and generate the frequency change segmented amplitude; S2: Based on the frequency change segmented amplitude, detect the time distribution of key frequency points, calculate the increase value of the frequency extreme point within the time period, analyze the change trend of the extreme point, extract the time interval characteristics of the amplitude fluctuation, match the frequency increase and amplitude fluctuation data, and generate the key frequency increase characteristic value; S3: Based on the key frequency increase characteristic value, analyze the amplitude characteristics of the key points of the frequency change, calculate the frequency change time span and the amplitude trend direction, analyze the matching relationship between the frequency change and the amplitude direction, extract the characteristics of the interactive change of frequency and amplitude, and generate the frequency-amplitude interactive ratio; S4: Based on the frequency-amplitude interaction ratio, analyzing the change characteristics of the interaction ratio in the time series, locating the ratio change peak interval, identifying the change range of the frequency intensity, analyzing the ratio change trend, and generating the interaction ratio characteristic interval; S5: Based on the interaction ratio characteristic interval, analyze the time distribution of the frequency intensity change and the amplitude characteristic, calculate the time span data of the frequency intensity and amplitude characteristics, extract the change rate of the frequency and amplitude, and generate the change rate of the rotating machinery fault characteristic.

2. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: The frequency change segmented amplitude includes the frequency fluctuation range, the amplitude fluctuation amplitude difference, the time interval between amplitude peaks, and the extreme point position distribution; the key frequency amplitude characteristic value includes the key frequency point time distribution, the frequency extreme point amplitude value, the amplitude fluctuation time interval characteristic, the frequency amplitude change trend, and the frequency and amplitude matching characteristic; the frequency-amplitude interaction ratio includes the amplitude characteristic parameter, the frequency change time span, the amplitude trend direction, and the frequency and amplitude interaction change characteristic value; the interaction ratio characteristic interval includes the ratio change peak range, the frequency intensity change range, and the ratio change trend distribution; the rotating machinery fault characteristic change rate includes the frequency intensity time distribution, the amplitude characteristic time span, the frequency change rate, and the amplitude change rate.

3. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: Based on the operation data of rotating machinery, the time series value of the vibration signal is extracted, the frequency and amplitude fluctuations of the time period are measured, the extreme points are identified, the difference in the fluctuation amplitude between the extreme points and the adjacent segments is calculated, the time interval between the amplitude peaks is determined, and the specific steps for generating the frequency change segmented amplitude are as follows: S101: extracting the time series value of the vibration signal based on the rotating machinery operation data, segmenting the time series according to the measurement time period, converting the data into frequency and amplitude change information, and generating frequency and amplitude segmented fluctuation results; S102: Based on the frequency and amplitude segmented fluctuation results, identify extreme points in the frequency and amplitude, extract the time position of each extreme point and the amplitude fluctuation range of the adjacent time period, calculate the fluctuation difference between each extreme point and its adjacent segment, and generate the fluctuation difference result between the extreme point and the adjacent segment; S103: Based on the fluctuation difference results of the extreme point and the adjacent segment, the time interval between the extreme points is analyzed, and the fluctuation amplitude of the frequency change within the time interval is combined, and the time periods are classified and sorted to generate frequency and amplitude segmented amplitude results.

4. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: Based on the frequency change segmented amplitude, the time distribution of key frequency points is detected, the increase value of the frequency extreme point in the time period is calculated, the change trend of the extreme point is analyzed, the time interval characteristics of the amplitude fluctuation are extracted, and the frequency increase and amplitude fluctuation data are matched. The specific steps of generating the key frequency increase characteristic value are as follows: S201: Based on the frequency change segmented amplitude, set and extract the time distribution of key frequency points, extract the segmented frequency value of each frequency point, calculate the change range in the differentiated time period, filter the data with significant frequency change amplitude, sort the distribution of key frequency points according to the time sequence, and generate the key frequency time distribution result; S202: Based on the key frequency time distribution result, the amplitude value of the frequency extreme point in the time period is calculated, the amplitude data is extracted, and the amplitude fluctuation range is classified, and the amplitude change of the extreme point in each time period and the time interval of the amplitude fluctuation are analyzed. According to the variation law of the amplitude between adjacent time periods, the variation trend of the extreme point is identified, and the characteristic result of the extreme point change and the time interval is generated; S203: Based on the extreme point change and time interval characteristic results, the frequency increase value and the amplitude fluctuation data are segmented and matched, and a corresponding relationship is established between the matched data according to the increase and fluctuation value within the time interval, and the amplitude increase characteristics of the key frequency points within the interval are extracted to generate the key frequency increase characteristic value.

5. The rotating machinery fault diagnosis method based on time series large model according to claim 4 is characterized in that: The specific calculation formula of the increase value is: Among them, Z represents the increase value, F j represents the frequency extreme point in the jth time period, G j-1 represents the frequency extreme point of the previous time period adjacent to the jth time period, B j represents the amplitude extreme value of the jth time period, C j-1 represents the amplitude extreme value of the previous time period adjacent to the jth time period, R j represents the time position of the jth time period, S j-1 represents the time position of the previous time period adjacent to the j-th time period, and m represents the total number of time periods.

6. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: Based on the key frequency increase characteristic value, the amplitude characteristics of the key points of the frequency change are analyzed, the time span of the frequency change and the amplitude trend direction are calculated, the matching relationship between the frequency change and the amplitude direction is analyzed, the characteristics of the interactive change of frequency and amplitude are extracted, and the specific steps of generating the frequency-amplitude interactive ratio are as follows: S301: based on the key frequency increase characteristic value, extract the amplitude characteristics of the key points of the frequency change, extract the amplitude value range of each key frequency point, match the extraction results with the frequency increase value item by item, calculate the amplitude and direction characteristic values ​​of the amplitude change, and generate the frequency change amplitude characteristic result; S302: Based on the frequency change amplitude characteristic result, analyze the time span data of the frequency change point, calculate the time difference between the frequency change points, and extract the amplitude trend direction of the corresponding time interval, classify and match the frequency increase direction and the amplitude trend characteristic according to the time interval, and judge the correlation change trend between the two, extract the data that meets the matching conditions, and organize them into interactive information to generate the frequency and amplitude matching relationship result; S303: Based on the frequency and amplitude matching relationship result, extract the frequency change and amplitude interaction characteristic data, group the intervals according to the time span, calculate the ratio of frequency and amplitude change, summarize the distribution law of the ratio, extract the characteristics, and generate the frequency-amplitude interaction ratio.

7. The method for fault diagnosis of rotating machinery based on time series large model according to claim 6 is characterized in that: The specific calculation formula of the time-weighted difference between the frequency change points is: Among them, T Δf Represents the time-weighted difference between frequency change points, t w Represents the timestamp of the wth frequency change point, t w+1 Represents the timestamp of the w+1th frequency change point, |t w+1 -t w | represents the absolute value of the time difference between two adjacent frequency change points, f w represents the frequency value of the wth frequency change point, f w+1 represents the frequency value of the w+1th frequency change point, |f w+1 -f w | represents the absolute value of the frequency difference between two adjacent frequency change points, represents the square root of the absolute value of the frequency difference, and W represents the total number of frequency change points.

8. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: Based on the frequency-amplitude interaction ratio, the specific steps of analyzing the change characteristics of the interaction ratio in the time series, locating the ratio change peak interval, identifying the change range of the frequency intensity, analyzing the ratio change trend, and generating the interaction ratio characteristic interval are as follows: S401: Based on the frequency-amplitude interaction ratio, extract the interaction ratio data in the time series, calculate the change amplitude data of the ratio, compare the increase and decrease trends of the values ​​in the change amplitude, mark the peak position, sort out the interval range of the ratio peak, and record the start and end time to generate the interaction ratio peak interval result; S402: Based on the interaction ratio peak interval result, extract the frequency intensity data within the peak interval, sort the frequency intensity values ​​within the interval, calculate the maximum value and the minimum value difference, compare the amplitude characteristics of the interval frequency intensity change, and generate the frequency intensity change range result; S403: Based on the frequency intensity variation range results, analyze the variation trend of the interaction ratio in each interval, compare the increase and decrease directions of the interaction ratio in the differentiated time intervals, reclassify the interval characteristics in combination with the time span and the frequency intensity amplitude, summarize the characteristic distribution of the interaction ratio in the time series, and generate the interaction ratio characteristic interval.

9. The rotating machinery fault diagnosis method based on time series large model according to claim 1 is characterized in that: Based on the interaction ratio characteristic interval, the specific steps of analyzing the time distribution of the frequency intensity change and the amplitude characteristic, calculating the time span data of the frequency intensity and the amplitude characteristic, extracting the change rate of the frequency and the amplitude, and generating the change rate of the rotating machinery fault characteristic are as follows: S501: extracting frequency intensity and amplitude characteristic data in each time interval based on the interaction ratio characteristic interval, analyzing the distribution characteristics of the frequency intensity data, identifying the distribution value variation rules of the amplitude characteristic and the frequency intensity according to the time period, arranging the data in the time period into corresponding distribution records in sequence, and generating frequency and amplitude time distribution results; S502: Based on the frequency and amplitude time distribution results, calculate the time span value of the frequency intensity and amplitude characteristic data within the time period, compare the time interval length of the frequency intensity and amplitude changes within the differentiated time period, and generate frequency and amplitude time span data; S503: Based on the frequency and amplitude time span data, calculate the change rate of the frequency intensity and amplitude characteristics within the time interval, extract the matching data of the frequency change rate and the amplitude change rate, classify and organize the rate characteristic data by time period, and generate the rotating machinery fault characteristic change rate.

10. A rotating machinery fault diagnosis system based on a time series large model, characterized in that: According to the rotating machinery fault diagnosis method based on the time series large model according to any one of claims 1 to 9, the system comprises: The vibration characteristic extraction module extracts the time series value of the vibration signal based on the operating data of the rotating machinery, measures the frequency and amplitude fluctuations of the time period, calculates the difference in the fluctuation amplitude between the extreme point and the adjacent segment, determines the time interval between the amplitude peaks, and generates the frequency change segment amplitude; The key frequency analysis module detects the time distribution of key frequency points based on the frequency change segment amplitude, calculates the amplitude value of the frequency extreme point within the time period, extracts the time interval characteristics of the amplitude fluctuation, matches the frequency amplitude and amplitude fluctuation data, and generates the key frequency amplitude characteristic value; The amplitude matching module analyzes the amplitude characteristics of the key points of frequency change based on the key frequency increase characteristic value, calculates the frequency change time span and the amplitude trend direction, analyzes the matching relationship between the frequency change and the amplitude direction, and generates a frequency-amplitude interaction ratio; The ratio trend analysis module analyzes the change characteristics of the interaction ratio in the time series based on the frequency-amplitude interaction ratio, identifies the change range of the frequency intensity, analyzes the ratio change trend, and generates the interaction ratio characteristic interval; The fault rate determination module analyzes the time distribution of the frequency intensity change and the amplitude characteristic based on the interaction ratio characteristic interval, calculates the time span data of the frequency intensity and the amplitude characteristic, and generates the rate of change of the rotating machinery fault characteristic.

Images