Spiral wind pipe predictive operation and maintenance method and system based on digital twinning

Abstract

The application discloses a spiral air pipe predictive operation and maintenance method and system based on digital twinning, relates to the technical field of air pipe system state monitoring and operation and maintenance, and comprises the following steps: high-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors are arranged at key nodes of a spiral air pipe, fan start-stop signals and running power are taken as working condition labels, a high-frequency acquisition mode is automatically triggered and noise reduction processing is carried out, and a multi-modal acoustic vibration original signal data set with time sequence, working condition labels and after preliminary noise reduction is output; the application can monitor the health state of a thin-walled structure in real time by arranging high-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors inside the spiral air pipe, can accurately capture weak early fault signals in the spiral air pipe and give early warnings in the fault germination stage such as micro-motion wear and aging of sealant, and realizes real-time monitoring, early warning and accurate operation and maintenance of the spiral air pipe.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of duct system state monitoring and operation and maintenance, in particular to a spiral duct predictive operation and maintenance method and system based on digital twinning. BACKGROUND

[0002] Spiral ducts are widely used in building ventilation and air conditioning ventilation systems, and their operating state directly affects the performance and energy consumption of the entire ventilation system. However, during long-term operation, the duct system may face problems such as micro-motion wear of the engaged joint, aging of the sealant, etc. If these problems are not discovered and addressed in a timely manner, they may lead to serious consequences such as duct leakage, increased energy consumption, and equipment damage.

[0003] For example, spiral ducts are made by spirally winding and engaging steel belts, which are subject to cyclic stresses such as fan start-stop, wind pressure fluctuations, slight deformation or vibration transmission of the building structure, etc. The spiral engaged joint is subject to micro-motion wear due to cyclic stress, which causes micro-cracks to form and expand, eventually leading to fatigue fracture or air leakage. However, this wear is extremely small and cannot be seen with the naked eye, making it impossible for traditional inspections to detect the problem. However, long-term accumulation can lead to a decrease in the tightness of the engagement, forming micro-cracks, until the air leakage significantly increases or the duct locally ruptures. In addition, the connection parts of the spiral duct are usually sealed with sealant. Over time and influenced by factors such as changes in environmental temperature and humidity, the sealant gradually ages, cracks, and fails, leading to leakage of the duct. Leakage not only reduces ventilation efficiency but also can result in energy waste.

[0004] Application No. CN202410565048.0 discloses an air conditioner duct real-time monitoring system and method, which includes a measurement terminal installed inside the air conditioner duct, a high-precision micro-electronic flow sensor, a particle counter, a noise sensor, and an optical fiber temperature and humidity sensor installed inside the measurement terminal, a compressor detection block, a condenser detection module, a throttling device detection module, and an evaporator detection module installed inside the air conditioner host, and a server terminal. Through the setting of the measurement terminal, data such as flow, flow rate, dust quantity, noise, temperature, and humidity in the air conditioner duct can be monitored, and the parameters of each component inside the air conditioner host can be detected to determine whether it is working normally. Through the high-precision digital twinning technology model, early warning can be performed on possible future conditions, which facilitates user prevention. When a fault occurs, the parameters of the abnormal part can be analyzed based on previous data, so that the fault part can be investigated, and the user can be facilitated to repair. However, this method lacks consideration of the micro-motion wear and fatigue cracking of the continuous spiral engaged joint of the duct and the aging and leakage failure mode of the sealant at the connection part during long-term operation, and there are defects of untimely maintenance and high cost.

[0005] The prior art has the following disadvantages: the existing air duct monitoring and operation technology mainly relies on periodic manual inspection or simple post-maintenance, which cannot capture the weak fault signals of the above slow and hidden early faults, the post-maintenance mode leads to unplanned shutdown and high repair cost, and the periodic maintenance lacks pertinence, is poor in economy and low in efficiency, and thus cannot realize precise operation and maintenance of the spiral air duct, therefore, it is urgent to provide a spiral air duct predictive operation and maintenance method and system based on digital twinning, which can realize early detection, precise diagnosis and residual life prediction of micro-motion wear, aging of sealant and other faults of the spiral air duct, and give optimal maintenance decision.

[0006] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and thus it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide a spiral air duct predictive operation and maintenance method and system based on digital twinning, which can accurately capture weak early fault signals in the pipe and give early warning through high-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors, adopt fault anomaly patterns discovered by MVSPC and unsupervised learning algorithm, and use a hybrid diagnostic method combining physical mechanism knowledge base and data-driven model to further accurately judge the fault type, occurrence event and fault location of the fault anomaly pattern, and use a physical-based fatigue cumulative damage model and a data-driven model based on different physical mechanisms to differentially predict the residual life, so as to solve the problems in the above background.

[0008] In order to achieve the above purpose, the present application provides the following technical scheme: a spiral air duct predictive operation and maintenance method based on digital twinning, comprising the following steps:

[0009] S1, deploying high-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors at key nodes of the spiral air duct, taking fan start-stop signals and running power as working condition labels, automatically triggering high-frequency acquisition mode and performing noise reduction processing, outputting a multi-modal acoustic-vibration original signal data set with time sequence, working condition labels and preliminary noise reduction, including acoustic emission signals and vibration signals;

[0010] S2, extracting features from the acoustic emission signals and vibration signals to form multi-dimensional feature vectors, and inputting the multi-dimensional feature vectors into a multivariate statistical process control MVSPC model, monitoring the spiral air duct state by calculating T2 statistics and squared prediction error, learning the fused multi-dimensional feature space by using an unsupervised learning algorithm, identifying abnormal points deviating from the normal baseline state, and outputting a structured event alarm;

[0011] S3, based on the structured event alarm, a physical mechanism knowledge base of fault characteristics is established, combined with multi-classification machine learning model trained by historical data, to judge the fault mode and output confidence, realize fault pipe segment level positioning by analyzing the time difference / energy attenuation of sensor signals, and output enhanced diagnostic report;

[0012] S4, using the enhanced diagnostic report, respectively using the physical-based fatigue cumulative damage model and the data-driven model to differentially predict the remaining life, the remaining life prediction and decision optimization module built-in cost model, the predicted remaining life is converted into risk cost, compared with the planned maintenance cost, using decision tree model to recommend the economic optimal maintenance time and strategy, and generating executable strategy.

[0013] Optionally, the high-frequency acoustic emission AE sensor frequency range is 20 kHz-1 MHz, which is used to capture the transient elastic wave acoustic emission signals released by the structural changes of micro-motion wear, micro-crack initiation and expansion at the spiral occlusal joint of the spiral duct material;

[0014] The low-frequency vibration acceleration sensor frequency range is 0.1 Hz-2 kHz, which is used to monitor the structural vibration signals caused by the decrease of connection stiffness due to sealant aging, air flow impact and fan vibration.

[0015] Optionally, the output steps of the multi-modal acoustic vibration original signal data set are as follows:

[0016] High-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors are deployed at key nodes of the spiral duct;

[0017] Set the working condition label, including the fan start, stop signal and running power, and monitor the fan start-stop signal and running power in real time;

[0018] When the fan start-stop on-off signal changes or the running power change rate exceeds the preset threshold, the high-frequency acquisition mode is automatically triggered;

[0019] When the high-frequency acquisition mode is triggered, start all high-frequency acoustic emission AE sensors deployed on the duct to collect acoustic emission signals at a frequency of 1 MHz, and simultaneously start all low-frequency vibration acceleration sensors deployed on the duct to collect vibration signals at a frequency of 2 kHz, and sample a fixed time length of the total length of the sampled signal;

[0020] The collected acoustic emission signals and vibration signals are processed by wavelet transform denoising with a low-pass filter to generate denoised acoustic emission signals and vibration signals;

[0021] Add time sequence information and working condition label to the denoised acoustic emission signals and vibration signals;

[0022] The acoustic emission signal and the vibration signal with time sequence, working condition label and preliminary noise reduction are combined into a multi-modal acoustic vibration original signal data set for output.

[0023] Optionally, the multi-dimensional feature vector extraction step is as follows:

[0024] The continuous acoustic emission signal and the vibration signal are divided into fixed-length analysis frames, and each frame contains 1024 sampling points;

[0025] A plurality of feature indicators are calculated in parallel for each frame of data, and for the acoustic emission signal, the ring count, energy, amplitude and wavelet packet energy spectrum are extracted; and for the vibration signal, the root mean square value, kurtosis and envelope spectrum are extracted;

[0026] The feature values of all calculated feature indicators are combined in order into a multi-dimensional vector to form a multi-dimensional feature vector.

[0027] Optionally, the monitoring spiral duct state step of the multi-variable statistical process control MVSPC model is as follows:

[0028] Under the normal operating state of the spiral duct, a large amount of multi-dimensional feature vector data is collected as samples, and the sample mean vector and the covariance matrix are calculated;

[0029] For the multi-dimensional feature vector input by the new sampling point, the T2 statistic and the squared prediction error SPE are calculated;

[0030] The control limit is determined according to the distribution of the T2 statistic;

[0031] The real-time calculated T2 statistic is compared with the determined control limit to determine whether the spiral duct is in a normal state.

[0032] Optionally, the step of identifying abnormal points by the unsupervised learning algorithm is as follows:

[0033] Randomly extract a subsample from the multi-dimensional feature vector data as a training set, and use the Isolation Forest in the unsupervised learning algorithm to train the training set;

[0034] Randomly select features and split values to construct multiple isolation trees to construct an isolation tree forest;

[0035] For each data point, traverse each isolation tree and record the path length from the root node to the leaf node;

[0036] The average value of the path length on all trees is calculated to calculate the anomaly score;

[0037] The anomaly score is used to evaluate the abnormal state of the data point, and is used to determine whether each data point is an abnormal point.

[0038] Optionally, the output step of the structured event alarm is as follows:

[0039] Abnormal evidence fusion is performed on the monitoring results of the MVSPC model and the anomaly scores of the unsupervised learning algorithm;

[0040] According to the characteristics of the abnormal points and the regions where they are located, the type of the abnormality is determined;

[0041] Combined with the position information of the sensors, the location where the abnormality occurs is determined;

[0042] The time when the abnormality occurs is recorded;

[0043] According to the deviation of the abnormal points from the normal baseline state, the Euclidean distance is calculated as a severity score of the deviation;

[0044] The type of the abnormality, the location where it occurs, the time, and the severity score are combined into a structured event alarm output.

[0045] Optionally, the establishment step of the physical mechanism knowledge base is as follows:

[0046] List all fault types that need to be diagnosed, including micro-motion wear of the engagement joint, aging leakage of the sealant, external impact, and normal state;

[0047] Collect various data related to the fault of the spiral duct, including structured event alarms, physical principles of the spiral duct, experimental data, and expert experience;

[0048] Extract and map the physical mechanism knowledge related to the characteristics of the fault from the collected data to determine the typical signal characteristics corresponding to each fault mode;

[0049] Mathematically model the typical performance of each multi-dimensional feature under a specific fault mode, represented in the form of probability distribution or numerical interval;

[0050] Integrate the well-represented knowledge into a knowledge base, thus establishing a physical mechanism knowledge base for subsequent query and use;

[0051] The steps of the multi-classification machine learning model to determine the fault mode and output the confidence are as follows:

[0052] Confirm a multi-classification model of logistic regression combined with the physical mechanism knowledge base;

[0053] Divide the historical data into multi-dimensional feature vectors and fault mode types to train the multi-classification model, obtaining a trained multi-classification model;

[0054] Input a new multi-dimensional feature vector into the trained multi-classification model and output the predicted fault mode and corresponding confidence.

[0055] Optionally, the step of predicting the remaining life by the physics-based fatigue cumulative damage model is as follows:

[0056] The low-frequency vibration acceleration sensor is converted into a stress time history at the joint by a mechanical model, a load spectrum of different stress amplitudes and stress cycle times is counted by rain flow counting method on the stress time history, and then a stress spectrum borne by the spiral pipeline during operation is determined;

[0057] Based on the stress spectrum, the fatigue damage model using the Miner linear cumulative damage theory calculates the cumulative damage that has occurred;

[0058] According to the cumulative damage of the current working condition of the spiral pipeline, the remaining service life is predicted;

[0059] The step of predicting the remaining life by the data-driven model is as follows:

[0060] Collect historical data related to the operating state of the spiral pipeline, including a multi-modal acoustic vibration original signal data set, a structured event alarm and an enhanced diagnostic report, and extract a plurality of feature data capable of representing the overall health state of the pipeline to construct a health index;

[0061] Arrange the historical health index HI values in chronological order to form a degradation time series, and use historical data to train a time series prediction LSTM model to learn the degradation pattern of the health index HI;

[0062] Convert the new input feature vector into the current degradation time series, input the trained LSTM model and output the predicted remaining life.

[0063] The predictive operation and maintenance system of the spiral air pipe based on digital twinning includes a spiral air pipe deployment heterogeneous sensor network module: acoustic emission sensors, vibration sensors, data collectors and working condition interfaces are deployed in the spiral air pipe to form a heterogeneous sensor network;

[0064] Data acquisition and preprocessing module: real-time acquisition of sensor signals is realized by using industrial Internet of Things protocol, fan start-stop signals and running power are used as working condition labels, high-frequency acquisition mode is triggered cooperatively and adaptive noise reduction processing is carried out, and a multi-modal acoustic vibration original signal data set after preliminary noise reduction is output;

[0065] Feature extraction and anomaly detection module: the acoustic emission signals and vibration signals in the multi-modal acoustic vibration original signal data set are respectively subjected to feature extraction, multi-dimensional features are input into a multivariate statistical process control MVSPC model and an unsupervised learning algorithm, and a structured event alarm is output;

[0066] Fault diagnosis and positioning module: according to the structured event alarm, combined with the physical mechanism knowledge base and the multi-classification machine learning model, the fault mode and the confidence are judged, the fault pipe segment level positioning is realized through analyzing the sensor signal, and the enhanced diagnosis report is output;

[0067] Remaining life prediction and decision optimization module: using the enhanced diagnosis report, the physical-based fatigue cumulative damage model and the data-driven model are used respectively to predict the remaining life, the risk cost is calculated and compared with the planned maintenance cost, the economic optimal maintenance time and strategy are recommended, and the executable strategy is generated.

[0068] In the above technical solution, the technical effects and advantages provided by the present application are:

[0069] The present application deploys high-frequency acoustic emission AE sensors and low-frequency vibration acceleration sensors inside the spiral duct, uses high-frequency acoustic emission technology to monitor the health status of the thin-walled structure in real time, can accurately capture the weak early fault signal in the spiral duct at the fault germination stage of the spiral duct such as micro-motion wear and aging of sealant, and issue a warning, realizing real-time monitoring, early warning and accurate operation and maintenance of the spiral duct, and achieving the effect of early detection and abnormal warning in the operation process of the spiral duct; by using MVSPC and unsupervised learning algorithm to learn the fault abnormal patterns hidden in the multi-dimensional feature vector space which cannot be directly discovered by human eyes, and by constructing a hybrid diagnosis method combining physical mechanism knowledge base and data-driven model, the fault type, occurrence event and fault location are accurately distinguished, avoiding blind maintenance by artificial, and greatly improving the operation and maintenance efficiency; and by using the physical-based fatigue cumulative damage model and the data-driven model with different physical mechanisms to predict the remaining life, the fault occurrence time is predicted in advance, the maintenance plan is reasonably arranged, the loss of emergency shutdown and unnecessary maintenance cost can be avoided, the operation efficiency and economy of the duct system are improved, and the fault prediction is unified and quantified to the cost dimension, the best maintenance time is recommended by comparing and optimizing the maintenance actions for different faults, which is beneficial to making maintenance plan strategy with the principle of economic optimization and maximizing investment return. BRIEF DESCRIPTION OF DRAWINGS

[0070] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0071] Figure 1 Flowchart of the present application based on digital twin spiral duct predictive operation and maintenance method.

[0072] Figure 2This is a block diagram of the predictive operation and maintenance system for spiral ducts based on digital twins, as described in this invention. Detailed Implementation

[0073] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0074] Example 1

[0075] This invention provides, for example Figure 1 The predictive operation and maintenance method for spiral ducts based on digital twins, as shown, includes the following steps:

[0076] S1. Deploy high-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors at key nodes of the spiral duct. Use the fan start / stop signal and operating power as operating condition labels to automatically trigger the high-frequency acquisition mode and perform noise reduction processing. Output a multi-modal acoustic and vibration raw signal dataset with time sequence, operating condition labels and preliminary noise reduction, including acoustic emission signals and vibration signals.

[0077] Specifically, the high-frequency acoustic emission (AE) sensor has a frequency range of 20kHz-1MHz and is used to capture transient elastic wave acoustic emission signals released by microstructural changes such as fretting wear, microcrack initiation and propagation at the spiral interlocking seam inside the spiral duct material.

[0078] The low-frequency vibration acceleration sensor has a frequency range of 0.1Hz-2kHz and is used to monitor structural vibration signals caused by decreased connection stiffness due to sealant aging, airflow impact, and fan vibration.

[0079] Specifically, the steps for outputting the multimodal acoustic vibration raw signal dataset are as follows:

[0080] High-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors are deployed at key nodes of the spiral duct. The high-frequency AE sensors can capture high-frequency acoustic emission signals generated by internal material deformation and crack propagation in the duct, while the low-frequency vibration acceleration sensors measure the vibration of the duct.

[0081] Set operating condition labels, including fan start-up and shutdown signals and operating power, and monitor fan start-up and shutdown signals and operating power in real time. The expression for the operating condition label is: In the formula, This is represented as a working condition label. This indicates a fan start signal. This indicates a fan shutdown signal. This is expressed as the operating power of the wind turbine;

[0082] When the fan start-stop switch signal is detected to change or the running power change rate exceeds the preset threshold, the high-frequency acquisition mode is automatically triggered, and an expression of automatically triggering the high-frequency acquisition mode is , wherein represents whether the high-frequency acquisition mode is automatically triggered when the fan start-stop switch signal is detected to change or the running power change rate at the moment , represents the running power, represents time, represents a trigger threshold of the fan running power rise rate, represents a trigger threshold of the fan running power drop rate, represents the fan start-stop switch signal, represents confirming that the fan start-stop switch signal is detected to change, represents the power change rate of the fan detected at the moment ;

[0083] When the high-frequency acquisition mode is triggered, all high-frequency acoustic emission AE sensors deployed on the air pipe are started to collect acoustic emission signals at a frequency of 1 MHz, all low-frequency vibration acceleration sensors deployed on the air pipe are started to collect vibration signals at a frequency of 2 kHz, and a fixed time length of the total length of the sampling signals is sampled;

[0084] The collected acoustic emission signals and vibration signals are subjected to wavelet transform denoising processing by a low-pass filter, the input acoustic emission / vibration signals are decomposed into components of different scales and frequencies, and the acoustic emission / vibration signals subjected to wavelet transform denoising are obtained after wavelet decomposition, threshold processing and wavelet reconstruction processing, to generate denoised acoustic emission signals and vibration signals, wherein an expression of the wavelet transform denoising processing of the acoustic emission signals by the low-pass filter is , and , and , wherein represents the value of the acoustic emission signal at the th sampling point, respectively represent the scale and frequency in the wavelet decomposition, represents the wavelet coefficient, represents the wavelet basis function, represents the sign function of the wavelet coefficient, represents the absolute value of the wavelet coefficient, represents the soft threshold, represents the wavelet coefficient after soft threshold function processing, represents the acoustic emission signal subjected to wavelet transform denoising at the The values ​​of each sampling point;

[0085] The simplified expression for wavelet transform noise reduction of vibration signals using a low-pass filter is: In the formula, The vibration signal after wavelet transform and noise reduction is represented as the signal at the th wavelet transform. The value of each sampling point, This is represented as the inverse discrete wavelet transform operation, which reconstructs the signal from the wavelet coefficients after soft thresholding. This is represented as a soft thresholding operation, which will process values ​​less than the soft threshold. The wavelet coefficients are set to zero to remove noise. This is represented by the discrete wavelet transform operation, which decomposes the signal into components of different scales and frequencies. Represented as the vibration signal at the th The values ​​of each sampling point;

[0086] Add timing information and operating condition tags to the denoised acoustic emission and vibration signals, where the timing information is the sampling timestamp;

[0087] The acoustic emission and vibration signals, labeled with time series and operating conditions and preliminarily denoised, are combined to form a multimodal acoustic-vibration raw signal dataset for output. The expression for this multimodal acoustic-vibration raw signal dataset is as follows: In the formula, This is represented as a dataset of raw acoustic and vibration signals from multiple modes. Represented as a sampling timestamp, This is represented as a working condition label. This indicates whether the high-frequency acquisition mode is automatically triggered. This represents the acoustic emission signal as a preliminary noise reduction. This represents the vibration signal as a preliminary noise reduction signal.

[0088] To further clarify, the key nodes for deploying high-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors in spiral ducts include the middle section of the long straight pipe section, supports, elbows, flange connections, and fan inlets and outlets.

[0089] High-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors collect multimodal acoustic and vibration raw signal datasets related to the operating status of the spiral duct in real time. These datasets provide raw data for subsequent analysis and processing, ensuring that the system can accurately perceive the actual operating status of the spiral duct.

[0090] S2, feature extraction is performed on the acoustic emission signal and the vibration signal respectively to form a multi-dimensional feature vector, and the multi-dimensional feature vector is input into a multivariate statistical process control (MVSPC) model, a T2statistic and a squared prediction error are calculated to monitor the spiral duct state, an unsupervised learning algorithm is used to learn the fused multi-dimensional feature space, an abnormal point deviating from a normal baseline state is identified, and a structured event alarm is output, including an abnormal type, a position of occurrence, a time, and a severity score;

[0091] Specifically, the multi-dimensional feature vector extraction step is as follows:

[0092] The continuous acoustic emission signal and the vibration signal are divided into fixed-length analysis frames, and each frame contains 1024 sampling points;

[0093] A plurality of feature indexes are calculated in parallel for each frame of data, the ringing count, the energy, the amplitude, and the wavelet packet energy spectrum are extracted for the acoustic emission signal, and the root mean square value, the kurtosis, and the envelope spectrum are extracted for the vibration signal;

[0094] The expression of the feature index extracted from the acoustic emission signal includes 、 、 、 In the formula, represents the ringing count, represents the counting function, represents the amplitude of the voltage corresponding to the i-th sampling point in the discrete acoustic emission signal, represents the threshold voltage threshold, represents the energy, represents the i-th sampling point, represents the total number of sampling points, represents the amplitude square summation of all sampling points of the discrete acoustic emission signal, represents the amplitude, represents the maximum value function, represents the absolute value of the amplitude of the voltage corresponding to the i-th sampling point, represents the wavelet packet energy spectrum of the i-th frequency band, represents the frequency band index, and , represents the i-th wavelet packet coefficient in the i-th frequency band, represents the i-th time point, represents the number of wavelet packet decomposition layers;

[0095] ​​​​​​​The expressions for the feature indices extracted from vibration signals include: , , In the formula, Represented as root mean square value, This represents the total number of sampling points. Represented as the vibration signal at the th Vibration amplitude at each sampling point Indicated as kurtosis, It is expressed as the mean of the vibration amplitude of the vibration signal at all sampling points. It is expressed as the standard deviation of the vibration amplitude of the vibration signal at all sampling points. Represented as an envelope signal, the envelope spectrum is obtained by performing a Fourier transform (FFT). Represented as the Hilbert transform operator, It is represented as the absolute value of the analytic signal, and , Represented as the imaginary unit;

[0096] The eigenvalues ​​of all calculated feature indicators are combined sequentially into a multidimensional vector, forming a multidimensional feature vector. The expression for the multidimensional feature vector is: In the formula, Represented as a multidimensional feature vector, This is represented as the envelope spectrum extracted from the vibration signal.

[0097] Specifically, the steps for monitoring the status of a spiral duct using the Multivariate Statistical Process Control (MVSPC) model are as follows:

[0098] Under normal operating conditions of the spiral duct, a large amount of multidimensional feature vector data is collected as samples, and the sample mean vector and covariance matrix are calculated. The formula for calculating the sample mean vector is as follows: In the formula, Represented as a sample mean vector, This represents the total number of samples collected for multidimensional feature vector data. Represented as the first A multidimensional feature vector;

[0099] The formula for calculating the covariance matrix is: In the formula, Represented as the sample covariance matrix, Represented as vector transpose;

[0100] For the multidimensional feature vector input at new sampling points, calculate the T² statistic and the squared prediction error (SPE). The formula for calculating the T² statistic is: In the formula, is expressed as a T2 statistic, and the greater the value of the T2 statistic, the farther the new sample point is from the multivariate center of the normal state, is expressed as a multidimensional feature vector input by the new sample point, is expressed as an inverse matrix of the covariance matrix;

[0101] The calculation formula of the squared prediction error SPE is , wherein, is expressed as a squared prediction error, is expressed as an identity matrix, is expressed as a principal load matrix extracted from the normal training data by the PCA model, is expressed as a squared Euclidean norm of a vector, is expressed as a projection matrix, is expressed as a residual projection matrix, is expressed as a residual vector;

[0102] The control limit is determined according to the distribution of the T2 statistic;

[0103] The real-time calculated T2 statistic is compared with the determined control limit to determine whether the spiral duct is in a normal state, and when any calculated T2 statistic exceeds the control limit, it is determined that the spiral duct is in an abnormal state.

[0104] Specifically, the steps of identifying an abnormal point by the unsupervised learning algorithm are as follows:

[0105] Randomly extract a subsample from the multidimensional feature vector data as a training set, and use the Isolation Forest in the unsupervised learning algorithm to train the training set;

[0106] Randomly select features and split values to construct multiple isolation trees to construct an isolation tree forest;

[0107] For each data point, traverse each isolation tree and record the path length from the root node to the leaf node;

[0108] According to the average value of the path length on all trees, the abnormal score is calculated, wherein the calculation formula of the abnormal score is , and , wherein, is expressed as the abnormal score of the subsample dimension feature vector data point, is expressed as the subsample size, is expressed as, is expressed as the subsample dimension feature vector data point The average value of the path length on all isolation trees, is expressed as the harmonic number of the normalized path length of the subsample, is expressed as a harmonic number;

[0109] The data point anomaly state is evaluated according to the anomaly score, and the closer the anomaly score is to 1, the more likely it is an abnormal data point. Points with anomaly scores much less than 0.5 are likely to be normal points, which are used to determine whether each data point is an abnormal point.

[0110] Specifically, the output steps of the structured event alarm are as follows:

[0111] The monitoring results of the MVSPC model and the anomaly scores of the unsupervised learning algorithm are fused as abnormal evidence;

[0112] According to the characteristics and the region of the abnormal point, the abnormal type is determined, including spiral wind pipe impact failure, stiffness change, leakage or blockage, etc.

[0113] Combined with the position information of the sensor, the position of the abnormal occurrence is determined;

[0114] The time of the abnormal occurrence is recorded;

[0115] According to the deviation degree of the abnormal point from the normal baseline state, the Euclidean distance is calculated as a severity score of the deviation;

[0116] The abnormal type, the occurrence position, the time and the severity score are combined into a structured event alarm output.

[0117] S3, based on the structured event alarm, a physical mechanism knowledge base of fault characteristics is established, combined with a multi-classification machine learning model trained by historical data to judge the fault mode and output the confidence, and through the analysis of the time difference / energy attenuation of the sensor signal, the fault pipe segment level positioning is realized, and an enhanced diagnosis report is output, including the fault position, the abnormal reason and the confidence;

[0118] Specifically, the establishment steps of the physical mechanism knowledge base are as follows:

[0119] List all the fault types that need to be diagnosed, including the micro-motion wear of the engagement joint, the aging leakage of the sealant, the external impact and the normal state, wherein the fault type is expressed as , and , wherein, is expressed as the th fault type, is expressed as the fault type of the micro-motion wear of the engagement joint, is expressed as the fault type of the aging leakage of the sealant, is expressed as the fault type of the external impact, is expressed as the normal state;

[0120] Collect various data related to the spiral duct failure, including structured event alarms, spiral duct physical principles, experimental data, and expert experience;

[0121] Extract and map the physical mechanism knowledge related to the fault characteristics from the collected data to determine the typical signal characteristics corresponding to each fault mode, such as the variation of sensor signals under different fault modes, the causes and influencing factors of fault occurrence;

[0122] Mathematically model the typical performance of each multi-dimensional feature under a specific fault mode, represented in the form of probability distribution or numerical interval;

[0123] Integrate the well-represented knowledge into a knowledge base, thus establishing a physical mechanism knowledge base for convenient subsequent query and use.

[0124] Specifically, the steps of the multi-classification machine learning model to determine the fault mode and output confidence are as follows:

[0125] Confirm a multi-classification model of logistic regression combined with the physical mechanism knowledge base;

[0126] Divide the historical data into multi-dimensional feature vectors and fault mode types to train the multi-classification model, obtaining a trained multi-classification model;

[0127] Input a new multi-dimensional feature vector into the trained multi-classification model and output the predicted fault mode and corresponding confidence, wherein the calculation formula of the confidence is , and , and , wherein, represents the confidence of the multi-dimensional feature vector belonging to the fault mode type predicted by the multi-classification model, represents the original score calculated by the multi-classification model according to the input of a new multi-dimensional feature vector belonging to the first fault mode category, represents the original score calculated by the multi-classification model according to the input of a new multi-dimensional feature vector belonging to the first fault mode category, represents the sum of the exponential scores of all fault mode categories, represents the maximum confidence, represents the operator of selecting the maximum probability.

[0128] Specifically, the output steps of the enhanced diagnostic report are as follows:

[0129] Determine the fault occurrence segment location by analyzing the arrival time difference or energy attenuation of the sensor signal combined with the digital twin model of the spiral duct, wherein the fault occurrence segment location expression is , wherein, denotes the distance from the fault point to one sensor, denotes the propagation speed of the signal in the pipeline, denotes the time difference of the signal detected by the two sensors, denotes the distance between the two sensors;

[0130] According to the output of the physical mechanism knowledge base and the multi-classification machine learning model, the abnormal reason of the fault type is analyzed;

[0131] The confidence of the multi-classification machine learning model output and the confidence of the fault position positioning are integrated to obtain a comprehensive confidence, wherein the calculation formula of the comprehensive confidence is , and , wherein, denotes the comprehensive confidence, denotes the weight coefficient corresponding to the confidence of the fault mode type judgment of the multi-classification model, denotes the weight coefficient corresponding to the confidence of the fault position positioning, denotes the confidence of the fault position positioning, denotes the confidence of the fault mode type judgment of the multi-classification model;

[0132] The fault position, the abnormal reason and the comprehensive confidence are combined into an enhanced diagnosis report output.

[0133] S4, using the enhanced diagnosis report, respectively using a physical-based fatigue cumulative damage model and a data-driven model to differentially predict the remaining life, the remaining life prediction and decision optimization module is embedded with a cost model, the predicted remaining life is converted into a risk cost, including expected energy loss, downtime loss, emergency repair cost, compared with planned maintenance cost, using a decision tree model to recommend the economic optimal maintenance time and strategy, and generating an executable strategy.

[0134] Specifically, the steps of predicting the remaining life based on the physical-based fatigue cumulative damage model are as follows:

[0135] The low-frequency vibration acceleration sensor is converted into a stress time history at the joint through a mechanical model, and the load spectrum of different stress amplitudes and stress cycle times is counted by rain flow counting method on the stress time history, and then the stress spectrum borne by the spiral pipeline in the running process is determined;

[0136] Based on the stress spectrum, the fatigue damage model of the Miner linear cumulative damage theory is used to calculate the cumulative damage that has occurred, wherein the expression of the fatigue damage model is , and , wherein, denotes the current accumulated damage degree, and when denotes the fatigue failure of the material, denotes the quantity of the stress level, denotes the number of cycles experienced at the stress amplitude, denotes the total number of cycles required for the material to fail in fatigue at the stress amplitude, obtained by querying the stress-life S-N curve, denotes the material constant, denotes the material stress amplitude, denotes the material coefficient;

[0137] predicts the remaining service life according to the current working condition of the spiral pipe, wherein the prediction expression of the remaining service life is , wherein denotes the fatigue damage model predicting the remaining service life of the spiral pipe, denotes the future accumulated damage degree, denotes the future running time, denotes the future damage accumulation rate, and when the future running condition is the same as the recent one, then .

[0138] Specifically, the steps of predicting the remaining life by the data-driven model are as follows:

[0139] Collect historical data related to the running state of the spiral pipe, including a multi-modal acoustic vibration original signal data set, a structured event alarm and an enhanced diagnosis report, and extract a plurality of characteristic data capable of representing the overall health state of the pipe to construct a health index, wherein the expression of the health index is , and , and , wherein denotes the health index value at time, denotes the aging-related characteristic value extracted from the historical data, denotes the aging-related characteristic value extracted from the historical data, denotes the set of aging-related characteristic values extracted from the historical data, denotes the aging-related characteristic value at time, denotes the weight coefficient of the aging-related characteristic value;

[0140] The historical health index HI values are arranged in chronological order to form a degradation time series, wherein the expression of the degradation time series is , wherein, represents the health index at the time of , and represents the health index at the present time;

[0141] A time series prediction LSTM model is trained using historical data to learn the degradation pattern of the health index HI, wherein the expression of the time series prediction LSTM model is , and , wherein, represents the hidden state and cell state at the current moment, represents the calculation of the time series prediction LSTM model, represents inputting the health index value at the time of into the LSTM model, represents the hidden state and cell state at the previous moment, represents the health index at the next moment predicted by the LSTM model, represents the weight coefficient of the output layer of the LSTM model, represents the bias term of the output layer of the LSTM model.

[0142] The new input feature vector is converted into the current degradation time series, input into the trained LSTM model and output the predicted remaining life, wherein the expression of the predicted remaining life is , and , wherein, represents the remaining service life of the spiral pipe predicted by the LSTM model, represents the predicted time step, represents the minimum value function, represents the number of steps predicted forward from the current moment, represents the health index at the current moment predicted, represents the preset health index failure threshold.

[0143] It is further supplemented that for the remaining life prediction of the micro-motion wear / fatigue crack of the spiral wind pipe, a physical-based fatigue cumulative damage model is adopted, and historical operation data is used as input; for the remaining life prediction of the aging of the sealant of the spiral wind pipe, a data-driven model is adopted, and stiffness feature frequency offset, etc. are used as health indexes.

[0144] Specifically, the conversion steps of the risk cost are as follows:

[0145] determining cost parameters of expected energy consumption loss, downtime loss, emergency repair cost and planned maintenance cost;

[0146] based on the cost model, converting the predicted residual life into a risk cost, wherein the expression of the risk cost is , and 、 、 , wherein, is expressed as a risk cost, is expressed as an expected energy consumption loss, is expressed as an energy consumption loss per unit time, is expressed as a downtime loss, is expressed as a downtime loss per unit time, is expressed as a downtime, is expressed as an emergency repair cost, is expressed as a fixed cost of emergency repair, is expressed as a minimum value function;

[0147] comparing the risk cost with the planned maintenance cost to analyze and plan periodic maintenance.

[0148] Specifically, the steps of the decision tree model for recommending the economically optimal maintenance time and strategy are as follows:

[0149] The risk cost, the planned maintenance cost and the predicted residual life are used as features to construct a training data set to determine the time decision options for the maintenance of the spiral pipe;

[0150] The decision tree algorithm is used to train the training data set to construct a decision tree model, wherein each decision option is a branch of the decision tree, and for each possible decision option, the total expected cost is calculated using the risk cost model;

[0151] The total costs of all decision branches are compared, and the branch with the lowest cost is selected as the decision of the optimal maintenance time.

[0152] Embodiment 2

[0153] The present application provides a spiral wind pipe predictive operation and maintenance system based on digital twinning as shown in Figure 2 The spiral wind pipe predictive operation and maintenance system based on digital twinning comprises a spiral wind pipe deployment heterogeneous sensor network module: acoustic emission sensors, vibration sensors, data collectors and working condition interfaces are deployed in the spiral wind pipe to form a heterogeneous sensor network;

[0154] A data acquisition and preprocessing module: real-time acquisition of sensor signals is realized by using an industrial Internet of Things protocol, fan start-stop signals and running power are used as working condition labels, a high-frequency acquisition mode is triggered cooperatively and adaptive noise reduction processing is performed, and a preliminary noise-reduced multi-modal acoustic vibration original signal data set is output.

[0155] Feature extraction and anomaly detection module: the acoustic emission signals and vibration signals in the multi-modal acoustic vibration original signal data set are respectively subjected to feature extraction, multi-dimensional features are input into a multivariate statistical process control (MVSPC) model and an unsupervised learning algorithm, and a structured event alarm is output;

[0156] Fault diagnosis and positioning module: according to the structured event alarm, a physical mechanism knowledge base and a multi-classification machine learning model are combined to judge a fault mode and a confidence, fault pipe segment level positioning is realized by analyzing sensor signals, and an enhanced diagnosis report is output;

[0157] Residual life prediction and decision optimization module: using the enhanced diagnosis report, a physical-based fatigue cumulative damage model and a data-driven model are respectively used to predict residual life, risk cost is calculated, compared with planned maintenance cost, economic optimal maintenance time and strategy are recommended, and an executable strategy is generated.

[0158] The spiral air pipe predictive operation and maintenance system based on digital twinning provided by the embodiment of the present application is realized through the spiral air pipe predictive operation and maintenance method based on digital twinning, and the specific method and process of the spiral air pipe predictive operation and maintenance system based on digital twinning are described in the embodiment of the spiral air pipe predictive operation and maintenance method based on digital twinning, which will not be repeated here.

[0159] The above formulas are dimensionless values, and the formulas are obtained by collecting a large amount of data to simulate the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0160] The above-described embodiments can be implemented in part or in whole through software, hardware, firmware or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loaded and executed by a computer, the computer instructions or computer programs cause the computer to perform the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing a set of one or more available media. The available media can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0161] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0162] Those skilled in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0163] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any skilled person 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 predictive operation and maintenance method for spiral ducts based on digital twins, characterized in that, Includes the following steps: S1. Deploy high-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors at key nodes of the spiral duct. Use the fan start / stop signal and operating power as operating condition labels to automatically trigger the high-frequency acquisition mode and perform noise reduction processing. Output a multi-modal acoustic and vibration raw signal dataset with time sequence, operating condition labels and preliminary noise reduction, including acoustic emission signals and vibration signals. S2. Features are extracted from the acoustic emission signal and vibration signal to form multidimensional feature vectors, which are then input into the Multivariate Statistical Process Control (MVSPC) model. The status of the spiral duct is monitored by calculating the T² statistic and the squared prediction error. An unsupervised learning algorithm is used to learn the fused multidimensional feature space to identify anomalies that deviate from the normal baseline state and output structured event alarms. The steps of the MVSPC model for monitoring the spiral duct status are as follows: Under normal operating conditions of the spiral duct, a large amount of multidimensional feature vector data is collected as samples, and the sample mean vector and covariance matrix are calculated. For the multidimensional feature vector input at new sampling points, calculate the T² statistic and the squared prediction error (SPE). The formula for calculating the T² statistic is: In the formula, It is expressed as the T² statistic, and the larger the T² statistic value, the further the new sampling point is from the multivariate center of the normal state. This is represented as a multidimensional feature vector input at the new sampling point. It is represented as the inverse of the covariance matrix; The formula for calculating the squared prediction error (SPE) is as follows: In the formula, This is expressed as the squared prediction error. Represented as an identity matrix, This is represented as the principal load matrix extracted from the normal training data by the PCA model. It is expressed as the squared Euclidean norm of a vector. Represented as a projection matrix, Represented as the residual projection matrix, Represented as a residual vector; Determine the control limits based on the distribution of the T² statistic; The real-time calculated T² statistic is compared with the determined control limits to determine whether the spiral duct is in a normal state; The steps of the unsupervised learning algorithm for identifying outliers are as follows: Subsamples are randomly selected from the multidimensional feature vector data as the training set, and the Isolation Forest algorithm in the unsupervised learning algorithm is used to train the training set. Randomly select features and segmentation values ​​to construct multiple isolation trees to build an isolation tree forest; For each data point, traverse each isolation tree and record the path length from the root node to the leaf node. Calculate the anomaly score based on the average path length across all trees; The anomaly score is used to assess the anomaly status of data points and to determine whether each data point is an anomaly. The output steps for the structured event alarm are as follows: The monitoring results of the MVSPC model and the abnormal scores of the unsupervised learning algorithm are fused to form abnormal evidence. Determine the anomaly type based on the characteristics of the anomaly and its location; By combining the location information from the sensors, the location of the anomaly can be determined; Record the time when the exception occurred; The severity of the deviation is scored by calculating the Euclidean distance based on the degree of deviation of outliers from the normal baseline. The anomaly type, location, time, and severity score are combined to form a structured event alert output. S3. Based on structured event alarms, establish a physical mechanism knowledge base of fault characteristics, combine a multi-class machine learning model trained with historical data to determine the fault mode and output confidence level, and achieve fault segment-level localization by analyzing the time difference / energy attenuation of sensor signals and output an enhanced diagnostic report. S4. Utilizing the enhanced diagnostic report, the remaining service life is predicted differentially using both a physics-based fatigue cumulative damage model and a data-driven model. The remaining service life prediction and decision optimization module incorporates a cost model, converting the predicted remaining service life into risk costs. These costs are compared with planned maintenance costs, and a decision tree model is used to recommend the economically optimal maintenance timing and strategy, generating an executable strategy. The steps for predicting remaining service life using the physics-based fatigue cumulative damage model are as follows: The low-frequency vibration acceleration sensor is converted into a stress time history at the joint through a mechanical model. The load spectrum of different stress amplitudes and stress cycle numbers is obtained by statistically analyzing the stress time history using the rainflow counting method, thereby determining the stress spectrum that the spiral pipe experiences during operation. Based on the stress spectrum, the fatigue damage model of Miner's linear cumulative damage theory is used to calculate the accumulated damage that has occurred. The remaining service life is predicted based on the cumulative damage under the current operating conditions of the spiral pipe. The steps for predicting remaining lifetime using the data-driven model are as follows: Collect historical data related to the operating status of the spiral pipeline, including multimodal acoustic and vibration raw signal datasets, structured event alarms and enhanced diagnostic reports, and extract multiple feature data that can characterize the overall health status of the pipeline to construct a health index; The historical health index HI values ​​are arranged in chronological order to form a degradation time series. A time series prediction LSTM model is trained using the historical data to learn the degradation pattern of the health index HI. The new input feature vector is transformed into the current degradation time series, input into the trained LSTM model, and the predicted remaining lifetime is output.

2. The predictive operation and maintenance method for spiral ducts based on digital twins according to claim 1, characterized in that, The high-frequency acoustic emission (AE) sensor has a frequency range of 20kHz-1MHz and is used to capture transient elastic wave acoustic emission signals released by structural changes such as fretting wear, microcrack initiation and propagation at the spiral interlocking seam inside the spiral duct material. The low-frequency vibration acceleration sensor has a frequency range of 0.1Hz-2kHz and is used to monitor structural vibration signals caused by decreased connection stiffness due to sealant aging, airflow impact, and fan vibration.

3. The predictive operation and maintenance method for spiral ducts based on digital twins according to claim 1, characterized in that, The steps for outputting the multimodal acoustic vibration raw signal dataset are as follows: High-frequency acoustic emission (AE) sensors and low-frequency vibration acceleration sensors are deployed at key nodes of the spiral duct. Set operating condition labels, including fan start-up, shutdown signals, and operating power, and monitor fan start-up and shutdown signals and operating power in real time; When a change in the start / stop switch signal of the wind turbine is detected or the rate of change of operating power exceeds a preset threshold, the high-frequency acquisition mode is automatically triggered. When the high-frequency acquisition mode is triggered, all high-frequency acoustic emission (AE) sensors deployed on the duct are activated to acquire acoustic emission signals at a frequency of 1MHz. Simultaneously, all low-frequency vibration acceleration sensors deployed on the duct are activated to acquire vibration signals at a frequency of 2kHz. A fixed duration is sampled as the total length of the sampled signal. The collected acoustic emission and vibration signals are processed by wavelet transform noise reduction using a low-pass filter to generate noise-reduced acoustic emission and vibration signals. Add timing information and operating condition labels to the noise-reduced acoustic emission and vibration signals; The acoustic emission signal and vibration signal, which have been labeled with time sequence and operating condition and have undergone preliminary noise reduction, are combined into a multimodal acoustic and vibration raw signal dataset for output.

4. The predictive operation and maintenance method for spiral ducts based on digital twins according to claim 1, characterized in that, The steps for extracting the multidimensional feature vector are as follows: The continuous acoustic emission signal and vibration signal are divided into analysis frames of fixed length, and each frame contains 1024 sampling points; For each frame of data, multiple feature indicators are calculated in parallel: for acoustic emission signals, ring count, energy, amplitude, and wavelet packet energy spectrum are extracted; for vibration signals, root mean square value, kurtosis, and envelope spectrum are extracted. The feature values ​​of all calculated feature indicators are combined in order into a multidimensional vector to form a multidimensional feature vector.

5. The predictive operation and maintenance method for spiral ducts based on digital twins according to claim 1, characterized in that, The steps for establishing the physical mechanism knowledge base are as follows: List all types of faults that need to be diagnosed, including fretting wear of the interlocking joint, aging and leakage of sealant, external impact, and normal condition; Collect various data related to spiral duct failures, including structured event alerts, spiral duct physics principles, experimental data, and expert experience; Extract and map the physical mechanism knowledge related to fault characteristics from the collected data to determine the typical signal characteristics corresponding to each fault mode; Mathematical modeling is performed on the typical performance of each multidimensional feature under a specific failure mode, and the result is expressed in the form of a probability distribution or numerical range. By integrating well-represented knowledge into a knowledge base, a physical mechanism knowledge base is established, which facilitates subsequent querying and use. The steps of the multi-class machine learning model in determining the fault mode and outputting the confidence score are as follows: A logistic regression multi-class classification model is identified, combined with a physical mechanism knowledge base; Historical data is divided into multi-dimensional feature vectors and fault mode types to train the multi-classification model, resulting in a well-trained multi-classification model. Input a new multidimensional feature vector into the trained multi-classification model and output the predicted fault mode and the corresponding confidence level.

6. A predictive operation and maintenance system for spiral ducts based on digital twins, implemented by the predictive operation and maintenance method for spiral ducts based on digital twins as described in any one of claims 1-5, characterized in that, This includes deploying heterogeneous sensor network modules in spiral ducts: deploying acoustic emission sensors, vibration sensors, data acquisition devices, and operating condition interfaces in spiral ducts to form a heterogeneous sensor network; Data acquisition and preprocessing module: Real-time acquisition of sensor signals using industrial IoT protocols, using fan start / stop signals and operating power as operating condition labels, collaboratively triggering high-frequency acquisition mode and performing adaptive noise reduction processing, outputting a preliminary noise-reduced multimodal acoustic and vibration raw signal dataset; Feature extraction and anomaly detection module: Extracts features from acoustic emission signals and vibration signals in the multimodal acoustic and vibration raw signal dataset, inputs multidimensional features into the multivariate statistical process control (MVSPC) model and unsupervised learning algorithm, and outputs structured event alarms; Fault diagnosis and location module: Based on structured event alarms, combined with a physical mechanism knowledge base and a multi-class machine learning model, it determines the fault mode and confidence level, realizes fault segment-level location by analyzing sensor signals, and outputs an enhanced diagnostic report; The Remaining Life Prediction and Decision Optimization Module utilizes enhanced diagnostic reports to predict remaining life using both a physics-based fatigue cumulative damage model and a data-driven model. It calculates risk costs, compares them with planned maintenance costs, recommends the most economically optimal maintenance timing and strategy, and generates executable strategies.

Images