Industrial equipment health management method and system based on machine learning

Abstract

The invention discloses an industrial equipment health management method and system based on machine learning, and relates to the technical field of industrial equipment health monitoring and predictive maintenance, and the method comprises the steps: collecting operation signals of industrial equipment, and constructing a time sequence feature matrix according to a time sequence through the operation signals; performing time sequence feature learning on the time sequence feature matrix by using a long and short term memory network model, and extracting a first time sequence correlation feature; training a Gaussian process regression model through the first time sequence correlation features, and establishing a health state evaluation model; and inputting a real-time operation signal of to-be-tested industrial equipment into the health state evaluation model, outputting an equipment health management index value, and triggering a hierarchical management strategy according to the equipment health management index value. Through the steps, the maintenance resource allocation is effectively optimized, and the comprehensive effects of reducing the operation and maintenance cost, prolonging the service life of equipment and improving the production efficiency are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment health monitoring and predictive maintenance, and particularly to an industrial equipment health management method and system based on machine learning. Background Art

[0002] Industrial equipment health management is one of the core technologies in the fields of intelligent manufacturing and industrial Internet of Things. It aims to improve production efficiency and reduce the risk of unplanned downtime by real-time monitoring the operating status of equipment, predicting potential failures, and optimizing maintenance strategies. In the prior art, machine learning-based equipment health management methods have been widely applied in fields such as vibration analysis, temperature monitoring, and current signal processing, providing important support for the predictive maintenance of industrial equipment. However, in a complex industrial environment, how to accurately capture the equipment degradation trend, quantify the uncertainty of the health status, and achieve dynamic hierarchical management remains a technical problem to be solved urgently.

[0003] CN115081584A discloses a power equipment health management method, system and medium based on machine learning. It realizes the perception of equipment health status by constructing a fault sample model and combining neural network time series analysis. This method can train the model with historical fault data to trigger an early warning when detecting similar working conditions. However, this method highly depends on a pre-collected fault sample library, and the fault modes of industrial equipment have diversity and dynamic evolution characteristics, and a single fault sample library is difficult to cover all potential abnormal situations. In addition, the traditional neural network has limited modeling ability for long-period time series features, which may lead to missed detection of early degradation signals and cannot meet the high-precision health assessment requirements.

[0004] CN117421582A discloses a device health analysis method based on multi-source data-driven. It improves the accuracy of fault prediction by fusing multi-type sensor data and adopting incremental learning to optimize the model. Through feature learning and a health assessment model, this method can dynamically update model parameters to adapt to new working conditions, thereby reducing false alarms and missed detections. However, this method lacks in-depth mining of the non-linear correlation between signals in the time series feature extraction stage, and the health assessment result only depends on point estimation (such as prediction mean), without considering the uncertainty of model prediction (such as variance), resulting in insufficient evaluation reliability in the sub-healthy state of the equipment. In addition, its multi-source data processing process is complex and the computational overhead is large, making it difficult to be efficiently deployed on resource-constrained edge devices, limiting its application in real-time monitoring scenarios. Summary of the Invention

[0005] In view of the problems existing in the existing industrial equipment health management system in terms of time series feature extraction, state assessment accuracy, and maintenance strategy intelligence, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to achieve the deep time-series feature extraction of the operation data of industrial equipment, establish a high-precision equipment health status evaluation model, and then implement an intelligent classification management strategy based on the health index.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a machine learning-based industrial equipment health management method, which includes:

[0009] Collect the operation signals of industrial equipment and construct a time-series feature matrix for the operation signals in chronological order;

[0010] Use a long short-term memory network model to perform time-series feature learning on the time-series feature matrix and extract the first time-series correlation features;

[0011] Train a Gaussian process regression model through the first time-series correlation features to establish a health status evaluation model;

[0012] Input the real-time operation signals of the industrial equipment to be measured into the health status evaluation model, output the equipment health management index value, and trigger a classification management strategy according to the equipment health management index value.

[0013] As a preferred solution of the machine learning-based industrial equipment health management method of the present invention, where: inputting the real-time operation signals of the industrial equipment to be measured into the health status evaluation model, outputting the equipment health management index value, and triggering a classification management strategy according to the equipment health management index value includes:

[0014] Collect the real-time operation signals of the industrial equipment to be measured and perform time-series feature extraction on the real-time operation signals to obtain a second time-series feature matrix;

[0015] Input the second time-series feature matrix into the trained long short-term memory network model to extract the second time-series correlation features;

[0016] Input the second time-series correlation features into the health status evaluation model, and jointly determine the equipment health management index value through the prediction mean and prediction variance of the Gaussian process regression model;

[0017] Compare the equipment health management index value with a preset three-level health threshold, where the three-level health threshold includes a normal operation threshold, a warning threshold, and an alarm threshold;

[0018] Based on the comparison result, push the equipment health management index value and management decision to the equipment operation and maintenance system, and at the same time record them in the historical database for iterative optimization of the health status evaluation model.

[0019] As a preferred embodiment of the machine learning-based industrial equipment health management method of the present invention, wherein: the predicted mean is used to reflect the confidence level of the prediction result; the confidence level includes a first confidence interval and a second confidence interval; comparing the equipment health management index value with a preset three-level health threshold includes:

[0020] When the equipment health management index value is greater than the normal operation threshold and the predicted variance is less than the first confidence interval, the operating state of the current equipment is a healthy state;

[0021] When the equipment health management index value is between the warning threshold and the normal operation threshold or the predicted variance is between the first confidence interval and the second confidence interval, the operating state of the current equipment is a sub-healthy state;

[0022] When the equipment health management index value is less than the alarm threshold or the predicted variance is greater than the second confidence interval, the operating state of the current equipment is a failure risk state.

[0023] As a preferred embodiment of the machine learning-based industrial equipment health management method of the present invention, wherein: the method for establishing the health state evaluation model is as follows.

[0024] Construct a training data set, using the first time-series correlation feature as the input variable and the corresponding equipment health state index as the output variable;

[0025] Adopt a radial basis kernel function to construct a Gaussian process regression model, and optimize the hyperparameters by maximizing the marginal likelihood function, where the hyperparameters include the signal variance and the length scale parameter;

[0026] Standardize the first time-series correlation feature in the training data set, and calculate the kernel matrix generated by the radial basis kernel function;

[0027] Calculate the prior distribution of the Gaussian process regression model according to the kernel matrix, and update it in combination with the observed data to obtain the posterior distribution, forming a health state evaluation model.

[0028] As a preferred embodiment of the machine learning-based industrial equipment health management method of the present invention, wherein: the method for extracting the first time-series correlation feature is as follows.

[0029] Slice the first time-series feature matrix according to the input time window size to obtain a training sample sequence and a validation sample sequence;

[0030] Input the validation sample sequence into the trained long short-term memory network model, and extract the output of the long short-term memory network model as the first time-series correlation feature.

[0031] As a preferred solution of the industrial equipment health management method based on machine learning according to the present invention, wherein: the long short-term memory network model includes multiple stacked long short-term memory layers; the long short-term memory layer is composed of multiple long short-term memory units; the long short-term memory unit includes an input gate, a forgetting gate, a memory unit, and an output gate; the training method of the long short-term memory network model is as follows.

[0032] A batch normalization layer is introduced between adjacent long short-term memory layers to perform standardization processing on the input training sample sequence.

[0033] Skip connections are established between long short-term memory units in different layers to form a residual learning structure, alleviating the problem of gradient disappearance in deep networks.

[0034] The backpropagation algorithm is used to optimize the parameters of the long short-term memory network model, and the adaptive learning rate optimization algorithm is combined to improve the training efficiency.

[0035] A regularization term is introduced to suppress the complexity of the long short-term memory network model, prevent overfitting, and an early stopping strategy is set to complete the training of the long short-term memory network model.

[0036] As a preferred solution of the industrial equipment health management method based on machine learning according to the present invention, wherein: the construction method of the first time series feature matrix.

[0037] Sensors are arranged at multiple key measurement points of the industrial equipment, and the operating signals of the industrial equipment are collected through the sensors, where the operating signals include vibration signals, temperature signals, and current signals.

[0038] The operating signals are segmented according to a fixed time window to obtain multiple data segments.

[0039] Based on the data segments, time domain features and frequency domain features are extracted to construct the first time series feature matrix.

[0040] In a second aspect, an embodiment of the present invention provides an industrial equipment health management system based on machine learning, which includes:

[0041] A signal acquisition module for collecting the operating signals of the industrial equipment and constructing a time series feature matrix according to the time sequence of the operating signals.

[0042] A feature extraction module for performing time series feature learning on the time series feature matrix by using the long short-term memory network model and extracting the first time series correlation features.

[0043] A construction module for training a Gaussian process regression model through the first time series correlation features to establish a health status evaluation model.

[0044] An output module, configured to input real-time operation signals of an industrial device to be measured into the health status evaluation model, output an equipment health management index value, and trigger a hierarchical management strategy according to the equipment health management index value.

[0045] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the processor executes the computer program, any step of the above-mentioned industrial device health management method based on machine learning is implemented.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by a processor, any step of the above-mentioned industrial device health management method based on machine learning is implemented.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] 1. By the step of collecting operation signals of an industrial device and constructing a time series feature matrix in chronological order, various sensor data such as vibration signals, temperature signals, and current signals are segmented according to a fixed time window, and time domain features and frequency domain features are extracted to construct a complete time series feature matrix, thereby establishing a comprehensive digital mapping of the device operation state and realizing the structured organization and unified representation of multi-source heterogeneous data;

[0049] 2. By the step of using a long short-term memory network model to perform time series feature learning on the time series feature matrix and extracting the first time series correlation features, a multi-layer stacked long short-term memory layer structure is adopted, and a batch normalization layer and a residual learning structure are introduced, which not only solves the problem that traditional recurrent neural networks are difficult to process long sequence data, but also effectively alleviates the gradient disappearance problem in deep network training, and realizes the effective capture of long-term dependence relationships in industrial device operation data;

[0050] 3. By the step of training a Gaussian process regression model with the first time series correlation features to establish a health status evaluation model, a Gaussian process regression model is constructed using a radial basis kernel function, and hyperparameters such as signal variance and length scale parameters are optimized by maximizing the marginal likelihood function, forming a health status evaluation framework based on prior distribution and posterior distribution, and realizing the probabilistic characterization and uncertainty quantification of the device health status;

[0051] 4. By inputting the real-time operation signals of the industrial equipment to be tested into the health status evaluation model, outputting the equipment health management index value, and triggering the hierarchical management strategy according to this index value, comparing the health management index value with the preset normal operation threshold, warning threshold, and alarm threshold, and combining the distribution of the prediction variance in different confidence intervals, a three-level health status classification system (health status, sub-health status, and failure risk status) is constructed, realizing the real-time monitoring of the equipment health status and intelligent decision-making support. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0053] Figure 1 It is a flowchart of an industrial equipment health management method based on machine learning. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0055] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention, but the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0056] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0057] The present invention is described in detail in conjunction with the schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of clarity, the cross-sectional views showing the device structure will be enlarged locally not in accordance with the general scale, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0058] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper, lower, inner, and outer" is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0059] Unless otherwise clearly specified and defined in the present invention, the terms "installation, connection, and coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can also be a mechanical connection, an electrical connection, or a direct connection, and can also be indirectly connected through an intermediate medium, or can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] Embodiment 1

[0061] Referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides an industrial equipment health management method based on machine learning, including:

[0062] S1: Collect the operation signals of industrial equipment and construct a first time series feature matrix for the operation signals in chronological order.

[0063] S1.1: Select multiple key measuring points of industrial equipment to arrange sensors, and collect the operation signals of industrial equipment through the sensors, where the operation signals include vibration signals, temperature signals, and current signals.

[0064] In an alternative embodiment, the sensors include vibration sensors, temperature sensors, and current sensors; the sampling frequency of the vibration sensor is set to 20 kHz, the sampling frequency of the temperature sensor is set to 1 Hz, and the sampling frequency of the current sensor is set to 10 kHz.

[0065] Exemplarily, collect the vibration signals of industrial equipment through vibration sensors, collect the temperature signals of industrial equipment through temperature sensors, and collect the current signals of industrial equipment through current sensors, and the collection duration is 24 hours.

[0066] Exemplarily, preprocess the collected operation signals, including removing outliers, filling in missing values, and signal filtering; specifically, apply a band-pass filter to the vibration signals to remove high-frequency noise and low-frequency drift; perform a moving average process on the temperature signals to smooth the temperature fluctuations; apply a median filter to the current signals to remove spike interference.

[0067] S1.2: Segment the running signals according to a fixed time window to obtain multiple data segments.

[0068] S1.3: Based on the data segments, extract time-domain features and frequency-domain features to construct the first time-series feature matrix.

[0069] In an alternative embodiment, extracting time-domain features and frequency-domain features includes: extracting the mean, standard deviation, peak factor, and kurtosis of the vibration signal as time-domain features, and extracting the spectral energy distribution and main frequency component through fast Fourier transform as frequency-domain features; extracting the mean, fluctuation range, and change rate of the temperature signal as time-domain features; extracting the effective value, harmonic component, and power factor of the current signal as time-domain features; combining the time-domain features and frequency-domain features in chronological order to form the first time-series feature matrix X ∈ R T×B , where D is the feature dimension and B is the time step.

[0070] Exemplarily, the rows of the first time-series feature matrix represent time points, and the columns represent different types of signal features.

[0071] S2: Use the long short-term memory network model to perform time-series feature learning on the first time-series feature matrix and extract the first time-series correlation features.

[0072] S2.1: Segment the first time-series feature matrix according to the input time window size to obtain a training sample sequence and a validation sample sequence.

[0073] S2.2: Input the validation sample sequence into the trained long short-term memory network model and extract the output of the long short-term memory network model as the first time-series correlation features.

[0074] In an alternative embodiment, the long short-term memory network model includes multiple stacked long short-term memory layers; the long short-term memory layer consists of multiple long short-term memory units; the long short-term memory unit includes an input gate, a forget gate, a memory cell, and an output gate, and the relevant formulas are as follows:

[0075]

[0076]

[0077] u t = σ(M ot [r t-1 , v t +q ot )·exp(-η||s t - s t-1 || 2 )

[0078] r t = u t⊙tanh(s t );

[0079] where g t is the input gate state at time t, p t is the forget gate state at time t, s t is the memory cell state at time t, u t is the output gate state at time t, r t is the hidden state at time t, v t is the input vector at time t, M in is the input gate weight matrix, M ft is the forget gate weight matrix, M ct is the candidate state weight matrix, M ot is the output gate weight matrix, q in is the input gate bias, q ft is the forget gate bias, q ct is the candidate state bias, q ot is the output gate bias, δ is the historical information weight, ψ is the residual connection coefficient, s t-k is the memory cell state at time t - k, is the memory cell state of the l1-th layer network at time t - 1, is the memory cell state of the l2-th layer network at time t - 1, D is the variance calculation function, r t-1 is the hidden state at time t - 1, σ(*) is the Sigmoid activation function, tanh(*) is the hyperbolic tangent function, and ⊙ is the element-wise multiplication operation.

[0080] Exemplarily, the input gate is responsible for controlling the proportion of newly input information entering the memory cell at the current moment; the forget gate is responsible for controlling the proportion of historical information retained in the memory cell; the memory cell is responsible for storing long-term memory information; and the output gate is responsible for controlling the proportion of information output from the memory cell.

[0081] In an alternative embodiment, the training method of the long short-term memory network model is to introduce a batch normalization layer between adjacent long short-term memory layers to perform standardization processing on the input training sample sequence; establish skip connections between long short-term memory cells of different layers to form a residual learning structure to alleviate the gradient vanishing problem of deep networks; use the backpropagation algorithm to optimize the parameters of the long short-term memory network model, and combine with an adaptive learning rate optimization algorithm to improve the training efficiency; introduce a regularization term to suppress the complexity of the long short-term memory network model, prevent overfitting, and set an early stopping strategy to complete the training of the long short-term memory network model.

[0082] Exemplarily, the first temporal correlation feature includes the temporal evolution law of the operating state of industrial equipment.

[0083] S3: Train a Gaussian process regression model using the first temporal correlation features to establish a health status assessment model.

[0084] S3.1: Construct a training dataset, using the first temporal correlation features as input variables and the corresponding device health status indicators as output variables.

[0085] It should be noted that the health status indicators are obtained by expert experience annotation.

[0086] S3.2: Construct a Gaussian process regression model using a radial basis kernel function and optimize the hyperparameters by maximizing the marginal likelihood function, where the hyperparameters include signal variance and length scale parameter.

[0087] Specifically, the specific formula of the Gaussian process regression model is as follows:

[0088]

[0089] Among them, K(a n1 , b n2 ) is the kernel function value between the feature vector a of the n1th prediction sample and the feature vector b of the n2th historical sample, ΔR n1n2 is the weighted hidden state difference value between the n1th prediction sample and the n2th historical sample, ρ is the signal variance, ω is the length scale parameter, τ is the time scale parameter, N is the sequence length, z i1 is the i-th current timestamp, y i2 is the i-th historical timestamp, r i1 is the i-th current hidden state, e i2 is the i-th historical hidden state.

[0090] S3.3: Standardize the first temporal correlation features in the training dataset and calculate the radial basis kernel function to generate a kernel matrix.

[0091] S3.4: Calculate the prior distribution of the Gaussian process regression model according to the kernel matrix and update it with the observed data to obtain the posterior distribution, forming a health status assessment model.

[0092] S4: Input the real-time operation signal of the industrial device to be tested into the health status assessment model, output the device health management index value, and trigger a hierarchical management strategy according to the device health management index value.

[0093] S4.1: Collect the real-time operation signal of the industrial device to be tested and extract the temporal features of the real-time operation signal to obtain a second temporal feature matrix;

[0094] S4.2: Input the second temporal feature matrix into the trained long short-term memory network model to extract the second temporal correlation features;

[0095] S4.3: Input the second time-series correlation feature into the health status assessment model, and jointly determine the device health management index value by the predicted mean and predicted variance of the Gaussian process regression model;

[0096] It should be noted that the predicted mean reflects the confidence level of the prediction result; the confidence level includes the first confidence interval and the second confidence interval.

[0097] S4.4: Compare the device health management index value with the preset three-level health thresholds, where the three-level health thresholds include the normal operation threshold, the warning threshold, and the alarm threshold;

[0098] In an alternative embodiment, when the device health management index value is greater than the normal operation threshold and the predicted variance is less than the first confidence interval, the operating state of the current device is a healthy state; when the device health management index value is between the warning threshold and the normal operation threshold or the predicted variance is between the first confidence interval and the second confidence interval, the operating state of the current device is a sub-healthy state; when the device health management index value is less than the alarm threshold or the predicted variance is greater than the second confidence interval, the operating state of the current device is a fault risk state.

[0099] It should be noted that the normal operation threshold is the upper limit of performance tolerance set based on the statistical distribution of the device's historical health status data and expert experience; the warning threshold is the lower limit of the transition interval determined based on the inflection point analysis of the device performance degradation curve and the sensitivity of early fault signs; the alarm threshold is the mandatory intervention line set based on the device failure critical value and the safe operation boundary; the first confidence interval is the model credibility threshold set based on the predicted variance distribution of the Gaussian process regression model in the healthy state (such as the 85th percentile); the second confidence interval is the uncertainty warning line demarcated based on the sudden increase in variance characteristics of the model in the abnormal state (such as the 65th percentile).

[0100] Exemplarily, when the vibration characteristic index value of a certain centrifugal compressor is 0.85 (normal threshold 0.8) and its prediction variance is 0.02 (the first confidence interval threshold is 0.05), the current operating state of the device is a healthy state, triggering a routine maintenance process; when the device health management index value falls between the warning threshold and the normal operating threshold (e.g., 0.7 ≤ device health management index value ≤ 0.8), or the prediction variance is between the first and second confidence intervals (e.g., 0.05 ≤ variance ≤ 0.1), the current operating state of the device is a sub - healthy state; when the device health management index value breaks through the alarm threshold (e.g., the vibration index of the gearbox drops suddenly to 0.5, lower than the alarm threshold of 0.6), or the prediction variance exceeds the second confidence interval (e.g., the variance of the current harmonic characteristics reaches 0.15, exceeding the threshold of 0.12), it indicates that the device enters a failure risk state; typical scenarios include abnormal current harmonics caused by insulation deterioration of the motor winding (index value 0.4, variance 0.18), at this time the model shows high uncertainty due to the input data deviating from the training distribution, and immediate shutdown for maintenance is required to avoid catastrophic failure.

[0101] In an alternative embodiment, if it is in a healthy state, routine inspections and records are performed. Vibration and temperature data are inspected every shift, lubricating oil samples are collected regularly for analysis, and the cleanliness of the equipment operating environment is maintained.

[0102] Exemplarily, the healthy state means that the wear degree of each component of the device is within an acceptable range, the lubrication condition is good, the mechanical clearance meets the design requirements, and the environmental temperature and humidity are within the process specifications.

[0103] In an alternative embodiment, if it is in a sub - healthy state, the detection frequency is increased to once every 4 hours, focusing on the change trend of abnormal parameters, appropriately increasing the lubrication frequency, checking and recording the change of environmental parameters, and preparing spare parts.

[0104] Exemplarily, the sub - healthy state means that the device shows early failure signs but has not affected normal operation. The early failure signs include equipment imbalance, misalignment, or insufficient lubrication.

[0105] In an alternative embodiment, the failure risk state includes a warning state and a dangerous state; if it is in the warning state, a special monitoring plan is started, key parameters are recorded every hour, oil particle analysis and infrared thermography detection are carried out, the timing of spare part replacement is evaluated, a maintenance plan is formulated, and process parameters are adjusted if necessary to reduce the equipment load; if it is in the dangerous state, the operating load is immediately reduced or shutdown for maintenance is carried out, an emergency plan is started, professional technical personnel are organized for fault diagnosis, faulty components are replaced or repaired, all parts of the equipment are comprehensively inspected for damage, the root cause of the fault is analyzed, and the preventive maintenance strategy is optimized.

[0106] Exemplarily, the warning state indicates that the device has medium - level fault symptoms and requires timely intervention. The medium - level fault symptoms include bearing damage, gear wear, and local overheating of the motor winding. The dangerous state indicates that the device has severe - level fault symptoms and there is a risk of sudden shutdown. The severe - level fault symptoms include severe bearing damage, broken gear teeth, and insulation aging of the motor stator winding.

[0107] S4.5: Based on the comparison results, push the device health management index value and management decision to the device operation and maintenance system, and at the same time record them in the historical database for iterative optimization of the health status assessment model.

[0108] In summary, the present invention realizes the structured organization and unified representation of multi - source heterogeneous data by collecting the operation signals of industrial equipment and constructing a time - series feature matrix, providing high - quality input for subsequent analysis. It uses the long short - term memory network model for time - series feature learning, effectively capturing the long - term dependence relationship and degradation trend in the device operation data, and overcoming the limitations of traditional models in processing long - sequence data. Based on the extracted time - series correlation features, a Gaussian process regression model is trained, which not only realizes the probabilistic representation of the device health status but also quantifies the uncertainty of the model through the prediction variance, enhancing the reliability of the evaluation results. By inputting the real - time operation signals into the evaluation model and implementing a hierarchical management strategy in combination with three - level health thresholds, the maintenance resource allocation is effectively optimized, achieving the comprehensive effects of reducing operation and maintenance costs, extending the service life of the device, and improving production efficiency.

[0109] Embodiment 2

[0110] This is the second embodiment of the present invention. This embodiment also provides an industrial device health management system based on machine learning, including:

[0111] A signal acquisition module, which is used to collect the operation signals of industrial equipment and construct a time - series feature matrix for the operation signals in chronological order;

[0112] A feature extraction module, which is used to perform time - series feature learning on the time - series feature matrix by using a long short - term memory network model and extract the first time - series correlation features;

[0113] A construction module, which is used to train a Gaussian process regression model through the first time - series correlation features to establish a health status evaluation model, where the kernel function of the Gaussian process regression model uses a radial basis function;

[0114] An output module, which is used to input the real - time operation signals of the industrial equipment to be tested into the health status evaluation model, output the device health management index value, and trigger a hierarchical management strategy according to the device health management index value.

[0115] It should be noted that the technical solution of the system for industrial equipment health management based on machine learning and the technical solution of the above-mentioned method for industrial equipment health management based on machine learning belong to the same concept. For the details not described in detail in the technical solution of the system for industrial equipment health management based on machine learning in this embodiment, reference can be made to the description of the technical solution of the above-mentioned method for industrial equipment health management based on machine learning.

[0116] The above-mentioned each unit module can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to each above-mentioned module.

[0117] This embodiment also provides an electronic device, which includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a carrier network, NFC (Near Field Communication) or other technologies. When the computer program is executed by the processor, it realizes a multi-task edge computing resource scheduling method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or can be a button, a trackball or a touchpad set on the shell of the computer device, or can also be an external keyboard, a touchpad or a mouse, etc.

[0118] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by the processor, it realizes the method proposed in the above-mentioned embodiment.

[0119] The storage medium proposed in this embodiment and the method proposed in the above-mentioned embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above-mentioned embodiment, and this embodiment has the same beneficial effects as the above-mentioned embodiment.

[0120] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disc of a computer, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the method of the embodiments of the present invention.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0122] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages.

[0123] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0124] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0126] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn of the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0127] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. An industrial equipment health management method based on machine learning, characterized in that: including collecting the operation signals of industrial equipment and constructing a first time-series feature matrix for the operation signals in chronological order; performing time-series feature learning on the first time-series feature matrix by using a long short-term memory network model to extract first time-series correlation features; training a Gaussian process regression model by the first time-series correlation features to establish a health status evaluation model; inputting the real-time operation signals of the industrial equipment to be measured into the health status evaluation model, outputting an equipment health management index value, and triggering a hierarchical management strategy according to the equipment health management index value.

2. The method for industrial equipment health management based on machine learning according to claim 1, wherein: Inputting the real-time operation signals of the industrial equipment to be measured into the health status evaluation model, outputting an equipment health management index value, and triggering a hierarchical management strategy according to the equipment health management index value, including: collecting the real-time operation signals of the industrial equipment to be measured and performing time-series feature extraction on the real-time operation signals to obtain a second time-series feature matrix; inputting the second time-series feature matrix into the trained long short-term memory network model to extract second time-series correlation features; inputting the second time-series correlation features into the health status evaluation model and jointly determining the equipment health management index value through the prediction mean and prediction variance of the Gaussian process regression model; comparing the equipment health management index value with preset three-level health thresholds, where the three-level health thresholds include a normal operation threshold, a warning threshold, and an alarm threshold; based on the comparison result, pushing the equipment health management index value and management decision to the equipment operation and maintenance system, and simultaneously recording them in the historical database for iterative optimization of the health status evaluation model.

3. The industrial equipment health management method based on machine learning according to claim 2, wherein: The prediction mean is used to reflect the confidence level of the prediction result; the confidence level includes a first confidence interval and a second confidence interval; Comparing the equipment health management index value with preset three-level health thresholds, including: when the equipment health management index value is greater than the normal operation threshold and the prediction variance is less than the first confidence interval, the operation state of the current equipment is a healthy state; when the equipment health management index value is between the warning threshold and the normal operation threshold or the prediction variance is between the first confidence interval and the second confidence interval, the operation state of the current equipment is a sub-healthy state; when the equipment health management index value is less than the alarm threshold or the prediction variance is greater than the second confidence interval, the operation state of the current equipment is a fault risk state.

4. The method for industrial equipment health management based on machine learning according to claim 3, characterized in that: The method for establishing the health status evaluation model is constructing a training data set, using the first time-series correlation features as input variables and the corresponding equipment health status indicators as output variables; constructing a Gaussian process regression model by using a radial basis kernel function and optimizing hyperparameters by maximizing the marginal likelihood function, where the hyperparameters include signal variance and length scale parameters; performing standardization processing on the first time-series correlation features in the training data set and calculating the kernel matrix generated by the radial basis kernel function; calculating the prior distribution of the Gaussian process regression model according to the kernel matrix and updating to obtain the posterior distribution by combining the observed data to form a health status evaluation model.

5. The industrial equipment health management method based on machine learning according to claim 4, wherein: The method for extracting the first time-series correlation features is The first time-series feature matrix is sliced according to the input time window size to obtain a training sample sequence and a validation sample sequence; The validation sample sequence is input into the trained long short-term memory network model, and the output of the long short-term memory network model is extracted as the first time-series correlation feature.

6. The method for industrial equipment health management based on machine learning according to claim 5, wherein: The long short-term memory network model includes multiple stacked long short-term memory layers; the long short-term memory layer is composed of multiple long short-term memory units; the long short-term memory unit includes an input gate, a forget gate, a memory unit, and an output gate; the training method of the long short-term memory network model is as follows. A batch normalization layer is introduced between adjacent long short-term memory layers to perform normalization processing on the input training sample sequence; Skip connections are established between long short-term memory units in different layers to form a residual learning structure, alleviating the problem of gradient disappearance in deep networks; The backpropagation algorithm is used to optimize the parameters of the long short-term memory network model, and an adaptive learning rate optimization algorithm is combined to improve the training efficiency; A regularization term is introduced to suppress the complexity of the long short-term memory network model, prevent overfitting, and an early stopping strategy is set to complete the training of the long short-term memory network model.

7. The method for industrial equipment health management based on machine learning according to claim 5, wherein: The construction method of the first time-series feature matrix, Select multiple key measurement points of industrial equipment to arrange sensors, and collect the operating signals of the industrial equipment through the sensors, where the operating signals include vibration signals, temperature signals, and current signals; The operating signals are segmented according to a fixed time window to obtain multiple data segments; Based on the data segments, time-domain features and frequency-domain features are extracted to construct the first time-series feature matrix.

8. An industrial equipment health management system based on machine learning, based on the industrial equipment health management method based on machine learning according to any one of claims 1 to 7, characterized in that: including, A signal acquisition module for collecting the operating signals of industrial equipment and constructing a time-series feature matrix for the operating signals in chronological order; A feature extraction module for using a long short-term memory network model to perform time-series feature learning on the time-series feature matrix and extract the first time-series correlation feature; A construction module for training a Gaussian process regression model through the first time-series correlation feature to establish a health status evaluation model; An output module for inputting the real-time operating signals of the industrial equipment to be tested into the health status evaluation model, outputting the device health management index value, and triggering a hierarchical management strategy according to the device health management index value.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the industrial equipment health management method based on machine learning according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the industrial equipment health management method based on machine learning according to any one of claims 1 to 7 are implemented.

Images