Equipment health state operation and maintenance method based on multi-channel perception

Abstract

The invention discloses an equipment health state operation and maintenance method based on multi-channel perception, and relates to the technical field of equipment health monitoring, and the method comprises the steps: collecting the perception data of equipment in a steady-state operation stage and a dynamic working condition transition stage; according to the method, time domain segmentation marking is carried out on sensing data of a dynamic working condition transition section, transient response parameters and parameter change rate characteristics of all segments are extracted, a dynamic working condition characteristic library is established, early degradation signs are found by identifying the consistent trend of parameter slow change characteristics and interval shrinkage, and self-adaptive operation and maintenance of the equipment health state are achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of equipment health monitoring, and specifically relates to an equipment health state operation and maintenance method based on multi-channel sensing. BACKGROUND

[0002] In the process of equipment operation, the degradation of the health state often shows early characteristics in the dynamic working condition transition section (such as start-stop instant, load switching, mode conversion). The traditional operation and maintenance method pays more attention to the steady-state operation data, and is insufficient in capturing the transient parameter change under the dynamic working condition, and is easy to ignore the early degradation signs, resulting in equipment failure burst, increasing the maintenance cost and downtime loss. Therefore, how to accurately identify the early degradation by multi-channel sensing of the parameter change under the dynamic working condition and realize adaptive preventive maintenance is the key to improving the equipment operation and maintenance efficiency and reliability. SUMMARY

[0003] The purpose of the application is to provide an equipment health state operation and maintenance method based on multi-channel sensing, which solves the technical problem that the parameter change under the dynamic working condition cannot accurately identify the early degradation and realize adaptive preventive maintenance in the prior art.

[0004] An equipment health state operation and maintenance method based on multi-channel sensing, comprising: acquiring sensing data of the equipment in the steady-state operation stage and the dynamic working condition transition section; time domain segment labeling is performed on the sensing data of the dynamic working condition transition section, transient response parameters and parameter change rate characteristics of each segment are extracted, and a dynamic working condition feature library is established; based on the dynamic working condition feature library, the dynamic working condition type in which the equipment is currently located is identified, the historical parameter fluctuation threshold of the same type of working condition is called, the dynamic reasonable interval of each sensing parameter under the working condition is real-timely delimited, and when the parameter change rate breaks through the historical threshold, interval adaptive contraction is performed; the sensing parameters of the equipment in each dynamic working condition transition section are analyzed in time sequence according to the change trend, the parameter slow change characteristics are judged according to the contraction amplitude of the dynamic reasonable interval, the abnormal deviation and cumulative effect of the parameter change rate are identified, and when the slow change characteristics and the interval contraction present a consistent trend, the early degradation signs of the equipment are determined; when the early degradation signs are detected, a preventive maintenance instruction is triggered and the dynamic working condition feature library is updated, and adaptive operation and maintenance of the equipment health state are realized.

[0005] Further, in the process of identifying the early degradation signs of the equipment, further comprising: when it is detected that the data accuracy of a certain auxiliary judgment factor under the current special working condition is lower than a set threshold, the transient response parameter change curve and the dynamic reasonable interval contraction characteristics corresponding to the factor are extracted; The above features are compared with the historical effective data features of the same type of working condition in the dynamic working condition feature library, and the proportion of the feature overlapping section is counted; If the feature overlapping section proportion exceeds the set proportion, it is determined that the factor data is inaccurate and will not affect the final result, and it is continued to be included in the analysis sequence of degradation sign recognition; If the feature overlapping section proportion does not reach the set proportion, other auxiliary judgment factors of multi-channel perception are called, the parameter features corresponding to the pre-sequence steady stage and the post-sequence steady stage of the current dynamic working condition transition section are extracted, two factors that exist trigger correlation in the pre-sequence stage and result correlation in the post-sequence stage are screened out, and a replacement evaluation sequence is constructed; The parameter trend output by the replacement evaluation sequence is consistent with the trend of the original inaccurate factor, and if the consistency meets the requirements, the result of the replacement evaluation sequence is used to participate in the early degradation sign judgment.

[0006] Further, after the replacement evaluation sequence is put into use, it includes: The parameter mutation time of the pre-sequence trigger factor in the replacement sequence is intercepted as a reference time node, and the parameter fluctuation waveform of each preset period before and after the node is extracted; The fluctuation waveform is aligned with the standard waveform of the same type of working condition in the dynamic working condition feature library, and the number of unaligned wave peaks and troughs is counted; If the unaligned number exceeds the set value, the historical abnormal waveform segment of the device under the current dynamic working condition is selected as a reference to calibrate the parameter fluctuation waveform of the replacement sequence, so that the wave peak and trough difference meets the distribution rule of the reference segment; After calibration, re-alignment, if the unaligned number does not decrease after two consecutive calibrations, the factory standard waveform of the same type of device is called to adjust the parameter sampling frequency of the replacement sequence; Adjust again and compare, until the unaligned number is lower than the set value, associate the waveform calibration parameter and the sampling frequency adjustment record to the current dynamic working condition identifier, and update the dynamic working condition feature library.

[0007] Further, after the replacement evaluation sequence is put into use, it also includes: The parameter mutation time of the pre-sequence trigger factor in the replacement sequence is obtained and marked as a reference time node; The parameter curve slope change of the post-sequence result factor before and after the reference time node is extracted to generate a three-segment feature segment; The three-segment feature segment is compared with the standard segment of the same type of working condition in the dynamic working condition feature library, and the number of curve inflection point coincidences is counted; If the inflection point coincidence number is lower than the set value, the phase offset correction is performed on the parameter curve of the post-sequence result factor, so that the mutation synchronization section of the two curves reaches the threshold value; In the corrected alternative evaluation sequence, implant the environmental feature code of the current dynamic working condition as the matching key for the next call, and update to the dynamic working condition feature library.

[0008] Further, after the environmental feature code is implanted, it also includes: Extract the curve curvature rate of change of the mutation synchronization segment in the three-segment feature segment after the environmental feature code is implanted, and compare it with the historical calibration segment associated with the environmental feature code; if the curvature rate of change difference exceeds the working condition allowed range, intercept the real-time running noise signal of the equipment under the current dynamic working condition, and decompose the interference component overlapping the frequency of the three-segment feature segment; superimpose the interference component to the parameter curve of the post-sequence result factor in reverse, regenerate the corrected three-segment feature segment, and count the number of inflection point coincidences with the standard segment again; after each correction, check whether the degradation warning given by the alternative evaluation sequence fits the actual state of the equipment, if the deviation does not narrow down for two consecutive times, use other frequency bands to decompose the noise signal; until the number of inflection point coincidences and the warning fitting degree meet the requirements, bind the interference decomposition parameters corresponding to the environmental feature code and the calibration method, and store them well, and update to the dynamic working condition feature library for calibration reference in the same type of environment.

[0009] Further, the determination of the dynamic reasonable interval includes: Group the historical parameter fluctuation data according to the type of dynamic working condition, and each group contains a sufficient continuous sampling sequence reflecting the working condition characteristics; Determine the basic interval based on the central distribution range of each group of data, and then dynamically adjust it according to the degree of change of the current working condition parameters: the greater the change, the more obvious the contraction of the interval boundary to the center; Continuously monitor the parameter change state, when the change tends to be flat, gradually widen the interval to the basic range, and adjust the rhythm to adapt to the degree of change flattening.

[0010] Further, the calculation of feature deviation degree includes: Extract key feature points from the transient response parameter change curve that can reflect the curve shape, including the extreme points, turning points and characteristic values of the stable stage; Compare the difference degree of each feature point with the corresponding feature point of the historical effective data, and integrate these differences to get the overall deviation degree; The size of the feature deviation degree is determined by the cumulative effect of the difference of each feature point, and the difference of the core feature points has more significant influence on the deviation degree.

[0011] Further, the operation of amplitude calibration includes: Compare the fluctuation pattern of the current parameter fluctuation waveform with that of the historical abnormal waveform segment, and find out the difference rule of the amplitude of the two; Adjust the amplitude of the parameter fluctuation waveform of the alternative sequence according to the rule, so that the fluctuation characteristics of the waveform are consistent with the pattern distribution of the reference segment; If the calibration does not meet the expectations, adjust again according to the new morphological differences until the key fluctuation characteristics of the waveform are consistent with the reference segment.

[0012] Further, the implantation of environmental feature codes and the matching call include: Screening environmental factors directly related to device parameter fluctuations from multi-channel perception data; Extracting its typical features before and after the reference time node, combining them into a unique environmental feature code according to the influence logic; When called again, compare the newly generated feature code with the stored one in the library, and if it meets the standard, directly call it; If it does not meet the standard, replace the evaluation sequence parameter curve after adjusting according to the difference, update the adjustment record and new feature code to the feature library.

[0013] Further, trigger the pre-maintenance instruction and update the dynamic working condition feature map library to realize adaptive operation and maintenance, including: After determining early degradation, generate a pre-maintenance instruction containing inspection priorities and maintenance window periods in combination with core components and working condition intensity, and match the risk of emergency; Extracting the degradation parameter trajectory, interval record and environmental features as samples, and classifying them into corresponding working condition subsets and labeling the running time; Collect parameter change data after maintenance, and store it in association with the instruction to optimize the interval sensitivity and decision weight.

[0014] Compared with the prior art, the beneficial effects of the present application are: The present application establishes a dynamic working condition feature library through the transient parameters of the multi-channel perception device in the dynamic working condition transition section, adjusts the dynamic reasonable interval in real time in combination with the parameter change rate, identifies the consistent trend of parameter slow change characteristics and interval contraction to find early signs of degradation, and realizes adaptive operation and maintenance of the device health status. This method breaks through the limitations of traditional reliance on steady-state data, can detect potential device failures earlier, and reduces the probability of sudden failures. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 It is a schematic diagram of the system framework structure of the present application. DETAILED DESCRIPTION

[0016] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] Please refer to Figure 1The application provides a device health state operation and maintenance method based on multi-channel perception, comprising: Collecting perception data of the device in a steady-state running phase and a dynamic working condition transition section; wherein the dynamic working condition transition section includes start-stop moments, load switching processes, and running mode conversion phases.

[0018] Time-domain segmentation marking is performed on the perception data of the dynamic working condition transition section, transient response parameters and parameter change rate characteristics of each segment are extracted, and a dynamic working condition characteristic library is established; wherein the transient response parameters include start time, load transient adjustment amplitude, and mode conversion delay duration.

[0019] Based on the dynamic working condition characteristic library, the type of the dynamic working condition currently experienced by the device is identified, the historical parameter fluctuation threshold of the same type of working condition is called, and the dynamic reasonable interval of each perception parameter under this working condition is real-time delimited; when the parameter change rate breaks through the historical threshold, the interval is adaptively contracted; The change trend of the perception parameters of the device in each dynamic working condition transition section is analyzed in time sequence, the parameter slow change characteristic is judged with the contraction amplitude of the dynamic reasonable interval, the abnormal deviation and cumulative effect of the parameter change rate are identified, and when the slow change characteristic and the interval contraction present a consistent trend, early degradation signs of the device are determined; When the early degradation signs are detected, a preventive maintenance instruction is triggered and the dynamic working condition characteristic library is updated, so as to realize adaptive operation and maintenance of the device health state.

[0020] Among them, the perception data refers to the device running state parameters collected by multi-channel sensors, specifically including vibration acceleration (measured by a piezoelectric sensor), temperature (measured by a thermocouple sensor), current and voltage (measured by a Hall sensor), etc., covering stable parameters in the steady-state running phase and transient parameters in the dynamic working condition transition section, for comprehensively reflecting the device running state.

[0021] Among them, the dynamic working condition transition section refers to the transition phase in which the device running mode changes, specifically including start-stop moments, load switching processes, and running mode conversion phases, in which the device parameters change dramatically and early fault characteristics are easily exposed.

[0022] Among them, the transient response parameter refers to the instantaneous change index of the device parameter in the dynamic working condition transition section, specifically including start time, load transient adjustment amplitude, and mode conversion delay duration, which can be obtained through timestamp recording and parameter peak value calculation, and is used to describe the response characteristics of the dynamic working condition.

[0023] Among them, the parameter change rate characteristic refers to the speed of change of the transient response parameter with time, which can be specifically represented by calculating the change amount of the parameter per unit time (such as 5m / s² change of vibration acceleration per second), and is used to judge whether the parameter change is abnormal.

[0024] The dynamic working condition feature library refers to a database storing different dynamic working condition types and their corresponding transient response parameters and variation rate characteristics, which can be established by historical operation data collection and classification labeling, and is used to identify the dynamic working condition type in which the current device is located. The dynamic reasonable interval refers to the parameter normal variation range determined in combination with the historical parameter fluctuation threshold for a specific dynamic working condition type, which is adjusted in real time according to the parameter variation rate, and the interval is contracted when the variation rate exceeds the historical threshold, and is used to determine whether the parameter is in an abnormal state.

[0025] The parameter slow variation characteristic refers to the slow variation trend of the device parameter in long-term operation, which can be extracted by trend analysis (such as linear regression) on time series data, and is used to determine whether the device has early degradation.

[0026] The cumulative effect refers to the superposition effect of the parameter slow variation characteristic over time, and as time goes by, small parameter changes gradually accumulate, which may eventually lead to device failure. The cumulative deviation amount (such as the total deviation from the initial value) of the slow variation parameter can be calculated to evaluate it. The early degradation sign refers to the early feature of the device that indicates performance decline, which is manifested as the consistent trend of the parameter slow variation characteristic and the dynamic reasonable interval contraction range (such as the parameter continuously increasing slowly, while the dynamic reasonable interval continuously contracts), and is used to warn of potential device failure in advance. The pre-maintenance instruction refers to the preventive maintenance instruction issued for the early degradation sign, which specifically includes maintenance time, maintenance content and other information, and can be sent to the operation and maintenance terminal through the device control system, and is used to guide the operation and maintenance personnel to perform maintenance in advance.

[0027] The innovation of the present application is that the transient parameters of the device in the dynamic working condition transition section are perceived through multiple channels, the dynamic working condition feature library is established, the dynamic reasonable interval is adjusted in real time in combination with the parameter variation rate, the consistent trend of the parameter slow variation characteristic and the interval contraction is identified to find the early degradation sign, and the adaptive operation and maintenance of the device health state is realized. This method breaks through the limitation of traditional dependence on steady-state data, can discover potential device failure earlier, and reduces the probability of sudden failure.

[0028] Specifically, in order to further understand the technical solutions, cases are used for illustration: First, the device is equipped with a multi-channel sensor system, including at least a piezoelectric vibration sensor (sampling frequency 1 kHz), a thermocouple temperature sensor, and a Hall current and voltage sensor, for collecting vibration acceleration, temperature, current and voltage sensing data.

[0029] The system marks the dynamic working condition transition section of the equipment, such as the start-stop transient time marked as T1 (about 3 seconds), the load switching process marked as T2 (about 5 seconds), and the running mode conversion stage marked as T3 (about 2 seconds), and extracts the transient response parameters of each section: the start-up time of T1 stage (about 1.5 seconds), the load transient adjustment amplitude of T2 stage (current fluctuation of about 5A), and the mode conversion delay time of T3 stage (about 0.8 seconds), and establishes a dynamic working condition feature library.

[0030] When the equipment is running, the system identifies that the current is in the T2 load switching process by comparing the current perception data with the feature library data, calls the historical parameter fluctuation threshold of this working condition: vibration acceleration fluctuation ±2m / s², current fluctuation ±6A, and draws an initial dynamic reasonable interval. If the real-time monitoring of the vibration acceleration change rate reaches 8m / s² / s (breaks the historical threshold of 5m / s² / s), the system shrinks the dynamic reasonable interval to vibration acceleration ±1.5m / s² and current ±4A.

[0031] Time series analysis of the vibration acceleration of the T2 stage shows that it presents a slow change feature: an increase of 0.3m / s² per month, and the dynamic reasonable interval continues to shrink due to the change rate exceeding the limit many times, and the two tend to be consistent, which is determined as an early degradation sign (possibly caused by bearing wear).

[0032] After obtaining the judgment result, the system immediately triggers a preventive maintenance instruction to notify the operation and maintenance personnel to lubricate and check the bearing within 3 days, and updates the degradation feature of this time to the dynamic working condition feature library. By actively capturing the parameter change law of the transition stage such as start-stop and load switching, the key problems of difficult to find early implicit degradation and high false alarm rate in dynamic working condition in the prior art are solved, which is beneficial to realize the idea change from passive avoidance of dynamic data to active use of dynamic features.

[0033] In some embodiments of the present application, when identifying early degradation signs, the auxiliary judgment factors may have a decrease in data accuracy due to special working conditions. If the analysis is directly carried out using inaccurate data, it may lead to misjudgment (such as mistaking the parameter fluctuation caused by interference as degradation) or omission (such as covering the real degradation trend due to data distortion) of early degradation signs.

[0034] To this end, the application further proposes that when it is detected that the data accuracy of a certain auxiliary judgment factor under the current special working condition is lower than a set threshold, the transient response parameter change curve and the dynamic reasonable interval contraction feature corresponding to the factor are extracted; the above features are compared with the historical effective data features under the same type of working condition in the dynamic working condition feature library in terms of overlap degree, and the proportion of the feature overlapping section is counted; if the proportion of the feature overlapping section exceeds a set proportion, it is determined that the factor data is inaccurate and will not affect the final result, and it is continued to be included in the analysis sequence of degradation signs; if the proportion of the feature overlapping section does not reach the set proportion, other auxiliary judgment factors of multi-channel perception are called, the parameter features corresponding to the pre-steady stage and the post-steady stage of the current dynamic working condition transition section are extracted, two factors that exist trigger correlation in the pre-steady stage and result correlation in the post-steady stage with the inaccurate factor are screened out, and a replacement evaluation sequence is constructed; the parameter slow change trend output by the replacement evaluation sequence is consistent with the trend of the original inaccurate factor, and if the consistency meets the requirements, the result of the replacement evaluation sequence is used to participate in the early degradation sign judgment.

[0035] The innovation of the application lies in that the replacement items are screened from the pre-steady trigger factor and the post-steady result factor, the idea conversion from data synchronization matching to working condition logic chain complementation is realized, the factor replacement is deeply bound with the stage characteristics of the dynamic working condition, the continuity of the analysis process and the accuracy of the result are ensured, and it is beneficial to avoid the influence of single factor data abnormality on the overall judgment.

[0036] Further, in the above, it is explained that the early degradation signs are mined through the dynamic working condition transition section, and in the present scheme, the factor failure problem is solved through the pre-steady-post-steady stage correlation of the dynamic working condition, and the two together constitute a complete system from dynamic data utilization to dynamic factor complementation.

[0037] In the technical scheme, the data accuracy of the auxiliary judgment factor can be determined through a preset threshold (such as the data fluctuation variance exceeding 3 times the historical mean value); the feature overlap degree comparison can use the dynamic time warping (DTW) algorithm to calculate the curve similarity, and the set proportion is usually 70% (that is, the overlapping section proportion exceeding 70% is considered effective); the correlation screening of the replacement evaluation sequence can be based on the cause-effect relationship analysis (such as the trigger correlation between the motor starting current and the vibration acceleration in the pre-steady stage, and the result correlation between the bearing temperature and the vibration acceleration in the post-steady stage); the consistency check can be realized by calculating the Pearson correlation coefficient of the trend curve (the set threshold is 0.8).

[0038] Further, the transient response parameter change curve refers to the continuous waveform of the parameter of the auxiliary judgment factor (such as current, vibration acceleration) with time in the dynamic working condition transition section (such as load switching, start-stop moment), which reflects the complete change process of the parameter from steady state to transient state to new steady state (such as the curve of the current fluctuating from 10A to 15A in the T2 load switching).

[0039] The dynamic reasonable interval contraction feature refers to the behavior feature that the dynamic reasonable interval (such as current ± 6A) is contracted towards the center due to the parameter change rate breaking through the historical threshold, including the contraction amplitude (such as the difference from ± 6A to ± 4A), the contraction timing (associated with the time of parameter mutation), and the interval boundary after contraction (such as ± 4A), and is used to represent the matching relationship between interval adjustment and parameter transient change.

[0040] For the convenience of understanding, further description is made in combination with the foregoing case: under strong electromagnetic interference, the accuracy of Hall current data is lower than the threshold (the fluctuation variance is more than 3 times the historical average), the system starts to verify: the current transient curve (T2 stage 10A-15A) and the interval contraction feature (± 6A-± 4A) are extracted, and are compared with the historical data of the same working condition in the feature library; if the proportion of the coincident segment is 75% (> 70%), the data can continue to be used; if the proportion is 50% (< 70%), the alternative sequence “motor starting signal-motor winding temperature” is constructed; the alternative sequence shows that the starting delay increases by 0.1s per month, the winding temperature increases by 1℃ per month, and the consistency with the original current fluctuation trend (increasing by 0.5A per month) is 0.85 (> 0.8 threshold), so the result is adopted, combined with the slow change of vibration acceleration (increasing by 0.3m / s2 per month) and interval contraction, to determine the early degradation of the bearing.

[0041] In some embodiments of the present application, after the alternative evaluation sequence is put into use, the parameter fluctuation waveform thereof may be misaligned in peaks and troughs with the standard waveform in the dynamic working condition feature library due to individual differences of equipment, installation deviation of sensors or special working condition interference (such as time deviation of the peak value of the vibration acceleration waveform, and inconsistency of the valley amplitude of temperature fluctuation). If the misaligned waveform is directly used for analysis, it will lead to distortion of the parameter slow change trend, affecting the accuracy of early degradation sign identification.

[0042] In this regard, the present application further proposes to take the parameter mutation moment of the pre-sequence trigger factor in the alternative sequence as the reference time node, and extract the parameter fluctuation waveform of each preset period before and after the node; align the fluctuation waveform with the standard waveform of the same type of working condition in the dynamic working condition feature library in peaks and troughs, and count the number of misaligned peaks and troughs; if the number of misaligned peaks and troughs exceeds the set value, select the historical abnormal waveform segment of the equipment under the current dynamic working condition as a reference to calibrate the amplitude of the parameter fluctuation waveform of the alternative sequence, so that the difference between the peaks and troughs conforms to the distribution law of the reference segment; after calibration, realign, if the number of misaligned peaks and troughs does not decrease after two consecutive calibrations, retrieve the factory standard waveform of the same model of the equipment, and adjust the parameter sampling frequency of the alternative sequence with the peak-to-trough interval of the standard waveform as a template; after adjustment, compare again until the number of misaligned peaks and troughs is lower than the set value, associate the waveform calibration parameter and the sampling frequency adjustment record to the current dynamic working condition identifier, and update the dynamic working condition feature library.

[0043] The innovation of the present application lies in that, in the specific scenario of focusing on the dynamic correlation between environmental interference and equipment parameter fluctuation, the cyclic logic of fluctuation peak comparison-interference signal time domain superposition-band separation verification is used to convert the abstract environmental influence into an operable signal superposition correction.

[0044] Further, the technical solution is complementary to the above-mentioned "stage correlation replacement". The stage correlation focusing on the timing logic of the working condition (such as the logical correlation between the pre-sequence start instruction and the post-sequence temperature change in the high-frequency start-stop working condition) in the above-mentioned scheme filters and replaces factors through the stage correlation within the working condition, which is conducive to ensuring that the replacement sequence matches the characteristics of the working condition. The present rule focuses on the physical coupling between environmental interference and equipment parameters, and through signal superposition and band splitting, it strips off the influence of environmental interference on the replacement sequence, ensuring that the replacement result is not disturbed by external environmental fluctuations.

[0045] Furthermore, in response to the above-mentioned "shape correction", the above-mentioned scheme solves the timing misalignment through phase correction, and the present rule solves the environmental distortion through interference superposition. Both of them improve the reliability of the replacement sequence from the two dimensions of timing alignment and environmental anti-interference, and synergistically strengthen the robustness of degradation identification in dynamic working conditions.

[0046] Specifically, after the replacement evaluation sequence is put into use, the system first locks the key fluctuation interval through the reference time node, and checks the waveform matching degree through peak-to-valley alignment. If the deviation is large, the waveform high and low characteristics are first corrected through amplitude calibration, and then the time interval characteristics of the waveform are optimized through sampling frequency adjustment, finally the precise matching of the waveform is realized. At the same time, the adjustment parameters are associated with the working condition identifier and the feature library is updated, so that the processing capability of the system for the waveform is continuously improved with the accumulation of experience, forming a self-optimizing waveform calibration mechanism.

[0047] In the technical solution, the parameter mutation moment of the reference time node can be determined through threshold triggering (such as parameter change amount exceeding 10% / ms); the preset period is usually set according to the characteristics of the working condition (such as 5 cycles of vibration acceleration, each cycle 0.1 seconds); peak-to-valley alignment can use peak detection algorithm (such as extreme point identification based on gradient change); the set value is generally 20% of the total number of standard waveform peak and valley (such as 10 extreme points in the standard waveform, more than 2 unaligned points need to be calibrated); amplitude calibration can use proportional scaling method (adjust the amplitude according to the ratio of the peak-to-valley difference of the historical abnormal waveform to the current waveform); sampling frequency adjustment is realized by changing the data acquisition interval of the sensor (such as adjusting from 1kHz to 2kHz to refine the waveform details).

[0048] In the above-mentioned embodiments of the present application, the steps of calibrating the waveform of the replacement evaluation sequence and adjusting the sampling frequency are proposed to improve the matching degree of the parameter fluctuation waveform and the standard waveform.

[0049] However, after the alternative evaluation sequence is put into use, the parameter changes of the antecedent trigger factor and the subsequent result factor may have a phase shift in time, causing the curve patterns of the two to not correspond accurately, and even after waveform calibration, the judgment of the correlation of the parameter changes may be affected by the phase misalignment. For example, after the parameter mutation of the antecedent trigger factor, the inflection point of the curve of the subsequent result factor appears delayed, which weakens the dynamic correlation characteristics of the two, and further affects the reflection accuracy of the alternative evaluation sequence on the equipment state.

[0050] To this end, the present application further proposes obtaining the parameter mutation time of the antecedent trigger factor in the alternative sequence, and marking it as a reference time node; extracting the parameter curve slope change of the subsequent result factor before and after the reference time node, and generating a three-segment feature segment; comparing the three-segment feature segment with the standard segment of the same type of working condition in the dynamic working condition feature library in terms of pattern, and counting the number of inflection point coincidences; if the number of inflection point coincidences is lower than a set value, correcting the phase shift of the parameter curve of the subsequent result factor, so that the mutation synchronization segments of the two curves coincide to a threshold; in the corrected alternative evaluation sequence, implanting the environmental feature code of the current dynamic working condition as the matching key for next time, and updating it to the dynamic working condition feature library.

[0051] This technical solution of the present application corrects the curve phase of the antecedent and subsequent factors in the alternative evaluation sequence, and introduces the environmental feature code as a matching key, further improving the correlation and reusability of the alternative sequence. Through the pattern comparison of the three-segment feature segment, the inflection point characteristics of the curve slope change are accurately captured, and the time misalignment is eliminated by phase shift correction, ensuring that the dynamic correlation of the antecedent trigger and the subsequent result conforms to the standard law; and the implantation of the environmental feature code enables the feature library to more accurately match the working conditions under different environments, providing a more reliable reference for subsequent calls, and enhancing the adaptability of the system to complex environments.

[0052] Specifically, after the alternative evaluation sequence is put into use, the system takes the parameter mutation time of the antecedent trigger factor as a reference, divides the curve of the subsequent result factor into a three-segment feature segment of “stable segment before mutation—synchronization segment of mutation—stable segment after mutation”, and judges the phase consistency of the two through the inflection point coincidence degree. If the deviation is large, the key change nodes are aligned through phase shift correction, and then the current environmental features (such as temperature, humidity, electromagnetic interference intensity, etc.) are encoded and implanted, so that the feature library can call the corresponding standard segment according to the environmental differences, ensuring that the alternative sequence can maintain high evaluation accuracy under different environments.

[0053] This method strengthens the dynamic correlation of the antecedent and subsequent factors in the alternative evaluation sequence, avoids the correlation misjudgment caused by phase shift, and at the same time realizes the refinement of working condition matching through the environmental feature code, further improves the accuracy of early degradation sign recognition, and provides more stable technical support for device adaptive operation and maintenance.

[0054] In the technical solution, the determination of the parameter mutation moment can adopt the first derivative threshold method (such as the parameter change rate exceeding 5% / ms); the three-section characteristic segment is divided with the reference time node as the boundary, and the time length of each section is dynamically set according to the working condition (such as 1 second for each section); the curve inflection point identification is realized through the zero-crossing detection of the second derivative; the set value is usually 70% of the total number of inflection points of the standard segment (such as 10 inflection points in the standard segment, and the number of coincidences is less than 7, which needs to be corrected); the phase offset correction can adopt the time axis translation method (moving the whole sequence curve forward or backward by 0.01 seconds), and the coincidence threshold of the mutation synchronization segment is set to 80%; the environmental feature code is generated by combining the quantized environmental parameters (such as 25℃ for temperature, 60% for humidity, and 11 for strong electromagnetic interference) (such as 011011).

[0055] In the above embodiments of the present application, the environmental feature code is implanted in the alternative evaluation sequence as a matching key to improve the matching accuracy in subsequent calls.

[0056] However, after the environmental feature code is implanted, even if the phase of the pre-sequence trigger factor and the post-sequence result factor has been corrected, the curve curvature change rate of the mutation synchronization segment in the three-section characteristic segment may still deviate from the historical calibration segment due to the existence of specific interference (such as sudden vibration and transient electromagnetic pulse) in the current environment, thereby affecting the accuracy of the alternative evaluation sequence for degradation warning. If the curvature change rate difference exceeds a reasonable range, the curve shape deviates from the standard feature, which may cause the warning result to be inconsistent with the actual state of the device.

[0057] To this end, the present application further extracts the curve curvature change rate of the mutation synchronization segment in the three-section characteristic segment after the environmental feature code is implanted, and compares it with the historical calibration segment associated with the environmental feature code; if the curvature change rate difference exceeds the allowable range of the working condition, the real-time running noise signal of the device under the current dynamic working condition is intercepted, and the interference component overlapping the frequency of the three-section characteristic segment is decomposed; the interference component is superimposed in reverse to the parameter curve of the post-sequence result factor, to regenerate the corrected three-section characteristic segment, and the number of inflection point coincidences with the standard segment is counted again; after each correction, it is checked whether the degradation warning given by the alternative evaluation sequence fits the actual state of the device, and if the deviation does not decrease for two consecutive times, the noise signal is decomposed in other frequency bands; until the number of inflection point coincidences and the warning fitting degree meet the requirements, the interference decomposition parameters corresponding to the environmental feature code are bound and stored, and are updated to the dynamic working condition feature library for calibration reference in the same type of environment.

[0058] The technical scheme of the present application further eliminates the influence of environmental interference on the replacement evaluation sequence by monitoring the mutation synchronous segment curve curvature change rate and targeted processing of interference components, and ensures the fitting degree of the degradation early warning and the actual state of the equipment. By reverse superposition to realize curve correction, combined with multi-frequency noise decomposition to improve the comprehensiveness of interference processing, and finally binding the effective processing scheme with the environmental characteristic code, the system can quickly call mature calibration strategies in the same type of environment, which significantly improves the stability and reliability of the replacement evaluation sequence in complex interference environment.

[0059] Specifically, after the environmental characteristic code is implanted, the system first judges whether the current curve is affected by abnormal interference through curvature change rate comparison. If there is an excessive difference, focus on the interference components that overlap with the characteristic segment frequency, and offset its influence by reverse superposition. Then, through multiple rounds of correction and early warning fitting degree verification, the accuracy of the curve shape and the early warning result is ensured. At the same time, the effective interference processing parameters are associated with the environmental characteristic code and stored to form an environmental adaptive calibration knowledge base, so that the anti-interference ability of the system is continuously enhanced with practice.

[0060] This method deepens the fine processing of environmental interference, avoids the possible omission of interference in single frequency band decomposition, and ensures that the corrected replacement evaluation sequence can truly reflect the equipment state through closed-loop verification of early warning fitting degree, further reducing the risk of false or missed judgment caused by environmental interference, and providing a more solid guarantee for adaptive operation and maintenance of equipment health state.

[0061] In the technical scheme, the curve curvature change rate can be calculated by the second derivative (such as unit radian / second); the working condition allowable range is usually set as ±15% of the mean value of the curvature change rate of the historical calibration segment; the real-time running noise signal can be collected by a special noise sensor (with a sampling frequency not less than 10 kHz); the interference component decomposition can use Fourier transform or wavelet transform (such as selecting the same frequency range 5-20 Hz as the three-segment characteristic segment); reverse superposition means adding the negative value of the interference component to the original curve; the early warning fitting degree can be determined by comparing the coincidence degree of the early warning result and the actual detection data of the equipment (such as setting a threshold of 80%); other frequency bands can be divided according to frequency intervals (such as 20-50 Hz, 50-100 Hz, etc.), and the early warning fitting degree can be selected by the formula: wherein, is the early warning prediction value, is the actual detection value.

[0062] Through this process, even in complex environments such as strong electromagnetic interference, the replacement evaluation sequence can still output a degradation early warning that is highly consistent with the actual state of the equipment, providing accurate guidance for the preventive maintenance of bearing wear, and further improving the efficiency of adaptive operation and maintenance of the equipment.

[0063] The application further proposes grouping historical parameter fluctuation data according to dynamic working condition types, each group containing a continuous sampling sequence sufficient to reflect the characteristics of the working condition; determining the basic interval based on the central distribution range of each group of data, and then dynamically adjusting according to the severity of the current working condition parameter change: the greater the change, the more obvious the contraction of the interval boundary to the center; continuously monitoring the parameter change state, and when the change tends to be smooth, gradually widening the interval to the basic range, and the adjustment pace is adapted to the degree of change smoothing.

[0064] Specifically, this technical solution of the application realizes the fine and adaptive determination of the dynamic reasonable interval by constructing the basic interval through grouping by working condition and dynamically adjusting the interval boundary according to the severity of parameter change. By grouping according to working condition types, it ensures that the interval reference matches the working condition characteristics; based on the contraction adjustment of the severity of change, the sensitivity of the interval to transient abnormalities is improved; and the gradual widening in the smooth stage avoids normal fluctuations from being misjudged as abnormalities, making the interval determination more in line with the actual change law of the parameters.

[0065] Specifically, the determination of the dynamic reasonable interval first establishes the basic reference of each working condition through historical data grouping, and then dynamically contracts the boundary according to the severity of real-time parameter change (such as change rate, fluctuation amplitude), tightens the interval to capture subtle abnormalities when the parameter changes dramatically, and widens the interval to adapt to normal fluctuations when the change is smooth, and the adjustment pace changes synchronously with the degree of smoothing (such as the faster the smoothing speed, the faster the widening speed), finally realizing the accurate matching of the interval and the parameter dynamics.

[0066] This method enhances the adaptability of the dynamic reasonable interval to different working conditions and parameter change stages, can maintain high sensitivity to abnormalities when the parameter changes dramatically, and can avoid excessive strict judgment in the smooth stage, effectively improving the accuracy and timeliness of abnormal identification, and providing a more reliable reference for the judgment of early signs of degradation.

[0067] In this technical solution, the grouping of historical parameter fluctuation data can be divided according to dynamic working condition types (such as start-stop transient, load switching, mode conversion), and the length of each continuous sampling sequence is set according to the duration of the working condition (such as 1000 sampling points for the load switching process); the determination of the basic interval can use statistical methods (such as taking the 95% confidence interval of each group of data); the severity of parameter change can be quantified by calculating the parameter fluctuation amplitude in unit time (such as the change per second), and the contraction amplitude is positively correlated with the severity (such as a doubling of the change rate, an increase of 50% in the contraction amplitude); the degree of change smoothing is judged by the decay rate of the fluctuation amplitude, and the adjustment pace uses a proportional coefficient (such as a 10% increase in the degree of smoothing, a 10% increase in the interval widening speed).

[0068] Through this process, the dynamic reasonable interval is tightened in the severe stage of load switching to accurately capture abnormal fluctuations, and is relaxed in the smooth stage to adapt to normal steady state, effectively avoiding misjudgment caused by fixed interval, and providing a more accurate judgment basis for identifying early signs of bearing wear.

[0069] The application further proposes to extract key feature points reflecting the curve shape from the transient response parameter change curve, including extreme points, turning points and characteristic values in the stable stage; compare the difference degree of each feature point with the corresponding feature point of the historical effective data, and obtain the overall deviation degree by synthesizing these differences; the size of the feature deviation degree is determined by the cumulative effect of the differences of each feature point, and the difference of the core feature point has more significant influence on the deviation degree.

[0070] This technical solution of the application improves the accuracy and pertinence of feature deviation degree calculation by focusing on the difference analysis of key feature points. By extracting extreme points (peak, trough), turning points (inflection point) and characteristic values (such as mean value in the stable stage) that can determine the curve shape, it ensures that the difference analysis focuses on the core features of the curve; higher weight is given to the core feature points, so that the deviation degree can more truly reflect the substantial difference of the curve shape, providing a reliable quantitative basis for feature overlap degree comparison.

[0071] Specifically, when calculating the feature deviation degree, first, the key nodes are selected from the curve through the feature point detection algorithm, avoiding the interference of irrelevant data; then the difference value (such as coordinate difference value, amplitude difference value) of each feature point and the corresponding point of the historical data is calculated, and weighted summation is performed according to the importance of the feature points (the weight of the core feature point such as the current peak point at load switching is set to 0.3, and the weight of the ordinary turning point is set to 0.1), to obtain the overall deviation degree. The significant difference of the core feature point will cause the deviation degree to rise sharply, thus accurately reflecting the substantial deviation of the curve shape.

[0072] This method enhances the sensitivity of feature deviation degree to the core difference of the curve, which is beneficial to avoid the ambiguity of overall difference calculation, makes the feature overlap degree comparison result more reliable, and thus ensures the judgment accuracy of the effectiveness of the auxiliary judgment factor data, and provides a more solid support for the anti-interference ability of early degradation sign identification.

[0073] In the technical solution, the extraction of key feature points can adopt a derivative-based detection algorithm (such as identifying extreme points by a first-order derivative being zero and identifying turning points by a second-order derivative passing through zero); the stable stage feature value can adopt a sliding window mean value calculation (the window size is 10 sampling points); the feature point difference degree can be quantified by the Euclidean distance (such as the square root of the sum of the squares of the two-dimensional coordinate difference); the weight of the core feature point can be set by expert experience or feature importance analysis (such as the feature importance score of a random forest), and the weight of the core feature point is usually 2-3 times that of the ordinary feature point; the cumulative effect is realized by a weighted summation formula .

[0074] Through this calculation method, the feature deviation degree can accurately reflect the difference of the core features of the curve, and provides a reliable basis for the judgment of the effectiveness of the auxiliary judgment factor data.

[0075] The application further proposes comparing the current parameter fluctuation waveform with the fluctuation pattern of the historical abnormal waveform segment, finding the difference rule of the amplitude of the wave peaks and valleys, adjusting the amplitude of the replacement sequence parameter fluctuation waveform according to the rule, and making the fluctuation characteristics of the waveform consistent with the pattern distribution of the reference segment; if the expected result is not achieved after one calibration, the amplitude is adjusted again according to the new pattern difference, and the process is repeated until the key fluctuation characteristics of the waveform are consistent with the reference segment.

[0076] This technical solution of the application realizes the refinement of amplitude calibration by focusing on the difference rule of the amplitude of the wave peaks and valleys, and ensures that the calibrated waveform is highly consistent with the historical abnormal waveform segment in terms of fluctuation pattern. By analyzing the difference rule first and then adjusting it, the distortion of the pattern caused by blind scaling is avoided; the multi-round calibration mechanism can gradually reduce the deviation, so that the key fluctuation characteristics (such as the relative amplitude of a specific wave peak and the difference ratio of the wave peak and the wave valley) are consistent with the reference segment, providing a more accurate waveform basis for the subsequent wave peak and wave valley alignment.

[0077] Specifically, the amplitude calibration first identifies the difference rule of the current waveform and the historical abnormal waveform in terms of wave peak height, wave valley depth, adjacent wave peak amplitude ratio, etc. (such as the current wave peak amplitude being generally 1.2 times that of the reference segment) through a pattern comparison tool (such as waveform superposition visualization and feature point difference calculation); then the segments are adjusted according to the rule (such as multiplying the amplitude of each wave peak by 0.83), so that the fluctuation characteristics fit the distribution of the reference segment; if some key wave peaks / wave valleys still do not match after one adjustment, the new pattern difference (such as the adjustment of a specific wave peak still has a deviation) is analyzed again, and a second fine-tuning is performed until all the key fluctuation characteristics are consistent.

[0078] This method improves the accuracy and relevance of amplitude calibration, ensures that the waveform morphology remains consistent with the historical anomaly segment after amplitude adjustment, effectively reduces the misalignment of peaks and troughs caused by amplitude deviation, and provides stronger protection for the reliability of the replacement evaluation sequence.

[0079] In this technical solution, the fluctuation pattern comparison can use waveform superposition method (aligning two waveforms by time axis and then superimposing display) and feature point difference matrix calculation (quantifying the amplitude difference of each peak and trough); the difference rule extraction can be realized by statistical analysis (such as calculating the average of the ratio of each corresponding peak amplitude); the amplitude adjustment can use segmented proportional scaling (adjusting coefficient is calculated for each peak and trough separately); the key fluctuation features include the relative amplitude of main peaks, the maximum difference between peaks and troughs, and the amplitude ratio of adjacent peaks, etc., and the consistency determination can set a threshold (such as a difference less than 5%).

[0080] Through this process, the calibrated temperature fluctuation waveform can accurately match the fluctuation pattern of the historical anomaly segment, laying a good foundation for the subsequent peak and trough alignment with the standard waveform, and ensuring the accuracy of the replacement evaluation sequence for equipment state evaluation.

[0081] The present application further proposes to screen the environmental factors directly related to the device parameter fluctuation from the multi-channel perception data; extract their typical features before and after the reference time node, and combine them into a unique environmental feature code according to the influence logic; when called again, the newly generated feature code is compared with the stored one in the library, and if it meets the standard, it is directly called; if it does not meet the standard, the parameter curve of the replacement evaluation sequence is adjusted according to the difference, and the updated adjustment record and new feature code are updated to the feature library.

[0082] This technical solution of the present application improves the effectiveness and reusability of the environmental feature code by accurately screening environmental factors, generating feature codes according to the influence logic, and dynamically adjusting the matching calling mechanism. Screening directly related environmental factors ensures that the feature code can truly reflect the influence of the environment on the device parameters; combining features according to the influence logic makes the feature code have clear physical meaning and uniqueness; the difference adjustment during matching calling realizes the self-adaptation to environmental changes, which is beneficial to avoid the waste of historical data caused by slight differences in the environment, and at the same time, by updating the record to enrich the feature library, the environmental adaptability of the system is enhanced.

[0083] Specifically, the implantation of the environmental feature code first filters out high-correlation factors (such as electromagnetic interference and current fluctuation correlation coefficient above 0.8) from multi-channel perception environmental data (such as temperature, humidity, electromagnetic interference, air pressure, etc.) through correlation analysis (such as calculating the Pearson correlation coefficient of environmental factors and device parameter fluctuations); then extracts typical features (such as electromagnetic interference intensity peak, temperature change rate) of these factors before and after the reference time node, sorts and combines them according to the logic of "impact degree from large to small" (such as electromagnetic interference features occupying 4 positions, temperature features occupying 2 positions), and generates a unique environmental feature code; when called, first calculate the matching degree (such as string similarity) of the feature code of the new environment and the feature code in the library, and if it meets the standard (such as similarity ≥ 90%), directly call the corresponding substitute evaluation sequence; if it does not meet the standard, analyze the difference (such as temperature feature segment difference), adjust the parameter curve related to temperature (such as correcting the temperature coefficient) in the sequence before calling again, and store the adjustment process and new feature code in the feature library.

[0084] This method ensures that the environmental feature code can accurately represent the impact of the environment on the device, and the matching calling mechanism realizes efficient reuse and dynamic optimization of historical data, further improving the applicability of the substitute evaluation sequence in different environments, reducing the evaluation error caused by environmental changes, and providing more flexible and reliable support for device health state operation and maintenance.

[0085] In this technical solution, the screening of environmental factors can use correlation analysis combined with expert experience (such as excluding light factors unrelated to device parameter fluctuations); the extraction of typical features includes extreme values, change rates, and stable values (such as the electromagnetic interference peak in the 10 seconds before the reference time node); the impact logic can be sorted according to the impact weight of the factors on the device parameters (the weight is determined through regression analysis); the combination of feature codes can use binary or hexadecimal encoding (such as each feature is quantized to 2-bit binary); the matching threshold can be set to 90% of the feature code string similarity; and the difference adjustment can be realized by establishing a mapping relationship between feature code differences and parameter curve adjustment amounts (such as a temperature feature segment difference of 1 bit corresponds to a temperature curve overall offset of ±0.5℃).

[0086] Through this process, the environmental feature code can accurately match the impact of the environment on the device, and when called, it can directly reuse effective historical data and adapt to new environments through difference adjustment, which is conducive to improving the calling efficiency and adaptability of the substitute evaluation sequence.

[0087] The present application further proposes that after determining early degradation, a pre-maintenance instruction containing inspection focus and maintenance window period is generated by combining core components and working condition intensity, and the emergency degree matches the risk; the degradation parameter trajectory, interval record, and environmental features are extracted as samples, and are classified into corresponding working condition subsets and labeled with running time; after maintenance, parameter change data is collected, and the correlation instruction is stored to optimize the interval sensitivity and determine the weight.

[0088] Specifically, after triggering the pre-maintenance instruction, the system first locates the associated core component through the degradation feature (such as the slow change of vibration acceleration corresponding to the bearing), evaluates the degradation risk level in combination with the working condition intensity (such as the load switching frequency and the running time), generates an instruction containing the inspection focus (such as bearing gap measurement), the maintenance window period (such as within 3 days), the emergency level (such as 24 hours within the red pre-warning) matched with the risk level, and then integrates the parameter trajectory (such as the monthly increase of 0.3 m / s² of vibration acceleration) in the degradation process, the dynamic reasonable interval contraction record, and the environmental feature code into a sample, and classifies it into the corresponding working condition subset (such as “load switching + strong electromagnetic interference”) and labels the cumulative running time of the equipment; after the maintenance is completed, the parameter change data (such as the vibration acceleration after maintenance is reduced to 0.1 m / s² / month) is continuously collected and stored in association with the original pre-maintenance instruction, and by analyzing the parameter changes before and after the maintenance, the sensitivity of the dynamic reasonable interval (such as reducing the bearing-related interval contraction threshold by 10%) and the determination weight of each feature point (such as increasing the weight of the vibration acceleration inflection point) are optimized.

[0089] This method realizes the closed-loop management from degradation identification to maintenance execution and system optimization, makes the pre-maintenance instruction more practical, and enables the feature library to evolve continuously with maintenance experience, significantly improving the efficiency of equipment operation and maintenance and the recognition accuracy of early degradation of the system, and providing more perfect technical support for the health management of the whole life cycle of the equipment.

[0090] In the technical scheme, the association of the core component can be realized through the fault mode library mapping (such as the vibration frequency of 100-500 Hz corresponding to the rolling bearing); the working condition intensity can be quantified through the load coefficient, the running time, and the cycle number; the risk level evaluation adopts the risk matrix (combining the fault occurrence probability and the impact degree); the running time for sample labeling is used to analyze the correlation between component degradation and service life; the interval sensitivity optimization can adjust the interval boundary (such as reducing by 10%) by comparing the parameter fluctuation range before and after the maintenance; and the determination weight optimization adopts the gradient descent method (adjusting the weight coefficient of each feature point according to the maintenance feedback).

[0091] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A device health status operation and maintenance method based on multi-channel sensing, characterized in that, include: The data collected by the acquisition equipment during steady-state operation and during dynamic transitions are sensory data. The sensing data of the dynamic operating condition transition segment is segmented and labeled in the time domain, and the transient response parameters and parameter change rate characteristics of each segment are extracted to establish a dynamic operating condition feature library. Based on the dynamic operating condition feature library, the current dynamic operating condition type of the equipment is identified, historical parameter fluctuation thresholds of the same type of operating condition are called, and the dynamic reasonable range of each sensing parameter under the operating condition is defined in real time. When the parameter change rate exceeds the historical threshold, the range is adaptively shrunk. The changing trends of the sensing parameters of the equipment in each dynamic operating condition transition segment are analyzed by time series. The slow change characteristics of the parameters are judged by the shrinkage range of the dynamic reasonable range. Abnormal deviations and cumulative effects of parameter change rate are identified. When the slow change characteristics and the range shrinkage show the same trend, it is judged as an early sign of equipment degradation. When early signs of degradation are detected, a pre-maintenance instruction is triggered and the dynamic operating condition feature library is updated to achieve adaptive operation and maintenance of equipment health status.

2. The device health status operation and maintenance method based on multi-channel sensing according to claim 1, characterized in that, The process of identifying early signs of device degradation also includes: When the accuracy of a certain auxiliary judgment factor under the current special working condition is lower than the set threshold, the transient response parameter change curve and dynamic reasonable interval shrinkage feature corresponding to the factor are extracted. The overlap of the above features with the historical valid data features of the same type of working condition in the dynamic working condition feature library is compared, and the percentage of overlapping features is statistically analyzed. If the proportion of overlapping features exceeds the set ratio, it is determined that the inaccuracy of the data for this factor will not affect the final result, and it will continue to be included in the analysis sequence for identifying signs of degradation. If the proportion of overlapping feature segments does not reach the set ratio, other auxiliary judgment factors of multi-channel perception are retrieved, and the parameter features corresponding to the preceding steady-state stage and the following steady-state stage of the current dynamic working condition transition segment are extracted. Two factors that have a triggering association with the inaccurate factor in the preceding stage and a result association in the following stage are selected to construct an alternative evaluation sequence. The consistency between the gradual change trend of the parameters output by the alternative evaluation sequence and the trend of the original inaccurate factors is verified. If the consistency meets the requirements, the results of the alternative evaluation sequence are used to determine the early signs of degradation.

3. The device health status operation and maintenance method based on multi-channel sensing according to claim 2, characterized in that, Once the alternative assessment sequence is put into use, it includes: The parameter mutation moment of the preceding triggering factor in the substitution sequence is taken as the reference time node, and the parameter fluctuation waveforms of each preset period before and after the node are extracted. Align the undulating waveform with the standard waveform of the same type of working condition in the dynamic working condition feature library, and count the number of misaligned peaks and valleys. If the number of misaligned items exceeds the set value, select a historical abnormal waveform segment of the equipment under the current dynamic operating condition as a reference, and perform amplitude calibration on the parameter fluctuation waveform of the replacement sequence so that the peak-to-trough difference conforms to the distribution pattern of the reference segment. After calibration, realign. If the number of misaligned devices does not decrease after two consecutive calibrations, retrieve the factory standard waveform of the same model of equipment, use its peak-valley interval as a template, and adjust the sampling frequency of the replacement sequence parameters. After adjustment, compare again until the number of misaligned items is lower than the set value. Then, associate the waveform calibration parameters with the sampling frequency adjustment record to the current dynamic operating condition identifier and update the dynamic operating condition feature library.

4. The device health status operation and maintenance method based on multi-channel sensing according to claim 2, characterized in that, After the alternative assessment sequence is put into use, it also includes: Obtain the parameter mutation time of the preceding triggering factor in the substitution sequence and mark it as the baseline time node; Extract the slope changes of the parameter curves of subsequent result factors before and after the baseline time node, and generate a three-segment feature segment; The three-segment feature segments are compared with the standard segments of the same type of working condition in the dynamic working condition feature library to count the number of overlapping curve inflection points. If the number of overlapping inflection points is lower than the set value, the phase shift correction is performed on the parameter curve of the subsequent result factors so that the overlap of the synchronous segment of the sudden change of the two curves reaches the threshold. In the corrected alternative evaluation sequence, the environmental feature code of the current dynamic operating condition is implanted as the matching key for the next call and updated to the dynamic operating condition feature library.

5. A device health status operation and maintenance method based on multi-channel sensing according to claim 4, characterized in that, After the environmental signature is implanted, it also includes: After extracting the curve curvature change rate of the abrupt synchronization segment in the three-segment feature segment after implanting the environmental feature code, compare it with the historical calibration segment associated with the same environmental feature code. If the difference in curvature change rate exceeds the allowable range of the operating conditions, extract the real-time operating noise signal of the equipment under the current dynamic operating conditions and decompose the interference component that overlaps with the frequency of the three-segment feature segment. Inversely superimpose the interference component onto the parameter curve of the subsequent result factor to regenerate the corrected three-segment feature segment, and count the number of inflection points that coincide with the standard segment again. After each round of correction, check whether the degradation warning given by the alternative evaluation sequence matches the actual state of the equipment. If the deviation is not reduced in two consecutive rounds, use other frequency bands to decompose the noise signal. Continue until the number of inflection point overlaps and the degree of matching of the warning both meet the requirements. Then, bind and save the interference decomposition parameters corresponding to the environmental feature code with the calibration method, update it to the dynamic operating condition feature library, and provide a calibration reference for the same type of environment.

6. The device health status operation and maintenance method based on multi-channel sensing according to claim 1, characterized in that, The definition of a dynamic reasonable range includes: Historical parameter fluctuation data are grouped according to dynamic operating condition type, and each group contains a continuous sampling sequence that sufficiently reflects the characteristics of the operating condition. The basic interval is determined based on the concentrated distribution range of each set of data, and then dynamically adjusted according to the degree of change of the current operating parameters: the greater the change, the more obvious the contraction of the interval boundary towards the center. Continuously monitor parameter changes, and when the changes tend to level off, gradually widen the range to the basic range, adjusting the pace to match the degree of flatness of the changes.

7. A device health status operation and maintenance method based on multi-channel sensing according to claim 2, characterized in that, The calculation of characteristic deviation includes: Extract key feature points that reflect the shape of the transient response parameter change curve, including the extreme points, turning points and feature values ​​of the steady phase of the curve; By comparing the degree of difference between each feature point and the corresponding feature points in historical valid data, the overall deviation is obtained by summing these differences. The magnitude of the feature deviation is determined by the cumulative effect of the differences between the feature points, with the differences of the core feature points having a more significant impact on the deviation.

8. A device health status operation and maintenance method based on multi-channel sensing according to claim 3, characterized in that, Amplitude calibration operations include: Compare the fluctuation patterns of the current parameter fluctuation waveform with those of historical abnormal waveform segments to identify the differences in the amplitude of the peaks and troughs. Based on this rule, the amplitude of the fluctuation waveform of the substitution sequence parameter is adjusted so that the fluctuation characteristics of the waveform match the morphological distribution of the reference segment; If the calibration fails to meet expectations, adjust again based on the new morphological differences until the key fluctuation characteristics of the waveform match the reference segment.

9. A device health status operation and maintenance method based on multi-channel sensing according to claim 4, characterized in that, The implantation and matching of environmental signatures include: Filter environmental factors directly related to equipment parameter fluctuations from multi-channel sensing data; Extract its typical features before and after the benchmark time point, and combine them into a unique environmental feature code according to the influence logic; When called again, the newly generated feature code is compared with the one stored in the library. If the match is found, the feature code is called directly. If the standard is not met, the alternative evaluation sequence parameter curve is adjusted according to the difference and then called up, and the adjustment record and new feature code are updated to the feature library.

10. A device health status operation and maintenance method based on multi-channel sensing according to claim 1, characterized in that, Triggering pre-maintenance instructions and updating the dynamic operating condition feature map library to achieve adaptive operation and maintenance includes: After early degradation is determined, pre-maintenance instructions are generated based on the core components and operating conditions, including key inspection points and maintenance windows, with the urgency level matched to the risk. Extract degradation parameter trajectories, interval records, and environmental features as samples, classify them into corresponding working condition subsets, and label the running time; After maintenance, collect parameter change data and store associated instructions to optimize interval sensitivity and decision weights.

Images