A method, system, and device for determining the operating status of equipment.

Abstract

This invention belongs to the field of equipment operation and maintenance, specifically relating to a method, system, and device for determining the operating status of equipment. It aims to solve the problems of existing technologies, such as delayed fault detection, reliance on human experience, incomplete inspections, strong subjectivity, and the inability to achieve early warning and predictive maintenance. The invention includes: acquiring and preprocessing acoustic sensor signals; extracting envelope spectrum, spectral energy, and modulation features; inputting the features into a trained machine learning model to obtain anomaly scores; dynamically updating a threshold that adaptively reflects the statistical distribution characteristics of equipment status based on the latest and historical score data; determining the equipment status by comparing the score with the threshold, and generating alarms and maintenance suggestions when anomalies occur. This invention achieves intelligent monitoring and adaptive diagnosis of equipment status, enabling early warning and predictive maintenance.

Description

Technical Field

[0001] This invention belongs to the field of equipment operation and maintenance, and specifically relates to a method, system and equipment for determining the operating status of equipment. Background Technology

[0002] Flight simulators, especially the highest-level (such as FFS Level D) full-flight simulators, are indispensable key equipment in modern pilot training systems. Through a complex six-degree-of-freedom (6-DOF) motion system, they provide pilots with a highly realistic sense of aerial motion, serving as an important platform for high-risk subject training, emergency procedure drills, and aircraft type conversion training. The core actuator of this motion system is typically a parallel mechanism consisting of six high-performance electro-hydraulic or electric servo actuators, which, under computer control, perform precise coordinated telescoping motions to drive the simulated cockpit platform to achieve a combination of translational and rotational motions in three-dimensional space.

[0003] These actuators have complex structures, operate under harsh conditions, and are subject to variable loads. Their health status directly affects the reliability, safety, and training effectiveness of the simulator. Currently, the industry generally adopts a combination of "routine manual inspections" and "scheduled maintenance checks" to monitor and maintain the operational status of these critical components. Manual inspections mainly rely on the sensory experience of maintenance engineers (such as listening for abnormal noises and observing signs of leaks) and simple instrument measurements; scheduled maintenance requires more in-depth inspections during planned grounding periods.

[0004] However, this traditional model has many inherent flaws:

[0005] First, it has a significant lag. It can only be detected when the fault develops to a certain extent (such as producing obvious abnormal noise, performance degradation or leakage), and it cannot effectively detect early progressive deterioration signs such as bearing micro-pitting, initial wear of seals, and slight jamming of valve cores, leading to increased maintenance costs and a higher risk of unexpected downtime.

[0006] Secondly, its coverage and continuity are insufficient. Manual inspection is discrete and has a low sampling rate, making it difficult to cover all states of the actuator under full operating conditions and long-term operation. It is very easy to miss intermittent faults or transient abnormal events that only occur under specific operating conditions.

[0007] Third, it is highly subjective and lacks quantitative standards. Judging whether something is "abnormal" depends heavily on the personal experience and hearing of maintenance personnel. There is a lack of consistent and objective evaluation standards among different personnel, making it difficult to effectively pass on experience and resulting in low accuracy in diagnosing complex faults.

[0008] Fourth, the depth of inspection is limited. Relying on senses and simple tools, it cannot effectively detect the condition of internal components of the actuator (such as gears and bearing contact surfaces), and is powerless to detect specific acoustic feature changes caused by early internal wear.

[0009] Finally, this model is inefficient and disruptive to operation, requiring dedicated personnel and valuable training time for downtime maintenance. It is difficult to fundamentally shift from preventative and reactive maintenance to predictive maintenance, and it cannot optimize maintenance plans based on the actual health status of components.

[0010] Based on this, the present invention proposes a method, system, and device for determining the operating status of equipment. Summary of the Invention

[0011] To address the aforementioned problems in the prior art, namely, the existing technology suffers from delayed fault detection, reliance on human experience, incomplete inspection, strong subjectivity, and inability to achieve early warning and predictive maintenance, this invention provides a method, system, and device for determining the operating status of equipment.

[0012] A first aspect of the present invention provides a method for determining the operating state of a device, comprising:

[0013] Acquire raw acoustic signals from acoustic sensors;

[0014] The original acoustic signal is preprocessed, including filtering and noise reduction operations, to extract the target frequency band signal and suppress environmental noise;

[0015] Extract a set of predefined acoustic features from the preprocessed signal, including envelope spectrum features, spectral energy distribution features, and modulation features;

[0016] The acoustic features are input into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions.

[0017] Based on the latest acquired comprehensive anomaly scoring sequence and historical anomaly scoring data, the threshold for judging anomaly status is dynamically updated. The threshold can adaptively reflect the statistical distribution characteristics of the equipment's operating status.

[0018] The currently obtained anomaly score is compared with the threshold, and the device is determined to be in an abnormal state based on the comparison result.

[0019] If an abnormality is detected, an alarm signal is generated and maintenance suggestions are output.

[0020] Furthermore, the raw acoustic signals from the acoustic sensors are acquired using the following method:

[0021] A vibration sensor is installed on the surface of the device housing, and a reference microphone is deployed in the acoustic environment in which the device is located. Signals from the vibration sensor and the reference microphone are collected synchronously as the raw acoustic signals.

[0022] Furthermore, a set of predefined acoustic features is extracted from the preprocessed signal. These acoustic features include envelope spectrum features, spectral energy distribution features, and modulation features. The method is as follows:

[0023] Envelope analysis is performed on the preprocessed signal to obtain its envelope spectrum;

[0024] Within a predetermined frequency range of the envelope spectrum, the maximum value of its amplitude is calculated to characterize the intensity of a specific mechanical impact; wherein the specific mechanical impact includes at least gear meshing impact;

[0025] The ratio of the energy of the envelope spectrum within a preset sideband frequency range to the total energy of the envelope spectrum across the entire frequency band is calculated to characterize the significance of the modulation effect.

[0026] The ratio of the energy of the original spectrum in a preset high-frequency band to the total energy of the original spectrum in the full frequency band is calculated to characterize the energy proportion of high-frequency friction or impact noise.

[0027] The calculated maxima, sideband energy ratio, and high-frequency energy ratio are combined into a set of predefined acoustic feature vectors.

[0028] Furthermore, the acoustic features are input into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions, and the method is as follows:

[0029] Acoustic feature data is continuously collected during a predetermined learning phase after the device is started and confirmed to be in a fault-free operating state;

[0030] When the amount of collected fault-free acoustic feature data reaches a preset minimum requirement, the output is a dataset, which is then used to train a support vector machine model and a Gaussian mixture model.

[0031] The acoustic features extracted in real time are input into the support vector machine model and the Gaussian mixture model respectively to calculate the corresponding first anomaly score and second anomaly score.

[0032] A weighted fusion calculation is performed on the first abnormal score and the second abnormal score to obtain a comprehensive abnormal score.

[0033] Furthermore, based on the latest acquired anomaly scoring sequence and historical anomaly scoring data, the threshold for judging anomaly states is dynamically updated. The method is as follows:

[0034] Continuously acquire comprehensive anomaly scores generated by real-time calculations to form the latest anomaly score sequence;

[0035] The latest anomaly score sequence is fused with the historically stored normal operating condition anomaly score data to generate an updated anomaly score dataset.

[0036] Based on the updated anomaly scoring dataset, the statistical mean and statistical standard deviation of the anomaly scores are calculated;

[0037] Based on the calculated statistical mean and statistical standard deviation, combined with the current running time and load status, the threshold generation calculation is performed again to update the threshold.

[0038] Furthermore, based on the calculated statistical mean and standard deviation, combined with the current running time and load status, the threshold generation calculation is performed again, using the following method:

[0039] Obtain the calculated updated statistical mean and updated statistical standard deviation;

[0040] Obtain the load sensitivity factor corresponding to the current cumulative running time of the device, wherein the value of the load sensitivity factor decreases according to a predetermined rule as the running time increases;

[0041] Obtain the load deviation coefficient, which characterizes the difference between the current operating load state and the baseline load state;

[0042] The updated adaptive threshold is generated by adding the updated statistical mean, the updated statistical standard deviation by a preset multiple, the load sensitivity factor, and the load deviation coefficient.

[0043] Furthermore, the currently obtained anomaly score is compared with the threshold, and the device is determined to be in an abnormal state based on the comparison result. The method is as follows:

[0044] The comprehensive anomaly score calculated in real time is compared with the current threshold.

[0045] If the overall anomaly score is greater than the threshold, a first judgment signal indicating an abnormal health status is generated.

[0046] If the overall abnormal score is less than or equal to the threshold, a second judgment signal indicating a normal health status is generated.

[0047] Based on the generated judgment signal, the final judgment result of the real-time health status of the device is output.

[0048] Furthermore, if an anomaly is detected, an alarm signal is generated and maintenance suggestions are output, as follows:

[0049] If the final determination result is an abnormal health status, an alarm signal with a predetermined level will be generated.

[0050] Based on the magnitude of the comprehensive anomaly score and the degree of degradation of specific indicators in the acoustic features, corresponding maintenance recommendations are selected from a predetermined set of maintenance measures and generated.

[0051] Output the alarm signal and the maintenance recommendations.

[0052] In another aspect, the present invention provides a system for determining the operating state of a device, based on a method for determining the operating state of a device, the system comprising:

[0053] A signal acquisition module configured to acquire raw acoustic signals from an acoustic sensor;

[0054] A preprocessing module is configured to preprocess the original acoustic signal, including filtering and noise reduction operations, to extract the target frequency band signal and suppress ambient noise;

[0055] The feature extraction module is configured to extract a set of predefined acoustic features from the preprocessed signal, the acoustic features including envelope spectrum features, spectral energy distribution features and modulation features;

[0056] An anomaly score calculation module is configured to input the acoustic features into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions.

[0057] The threshold update module is configured to dynamically update the threshold for judging abnormal states based on the latest acquired comprehensive abnormal score sequence and historical abnormal score data. The threshold can adaptively reflect the statistical distribution characteristics of the device's operating status.

[0058] The status judgment module is configured to compare the currently obtained anomaly score with the threshold and determine whether the device is in an abnormal state based on the comparison result.

[0059] The maintenance module is configured to generate an alarm signal and output maintenance suggestions if an abnormality is detected.

[0060] A third aspect of the present invention provides an electronic device comprising:

[0061] At least one processor; and

[0062] A memory communicatively connected to at least one of the processors; wherein,

[0063] The memory stores instructions that can be executed by the processor to implement the method described above for determining the operating state of a device.

[0064] The beneficial effects of this invention are:

[0065] This invention effectively separates and enhances the acoustic characteristics of the actuator by employing dual-channel synchronous signal acquisition and adaptive noise reduction processing, providing high-quality input signals for subsequent analysis. By extracting features such as meshing impact intensity, sideband energy ratio, and high-frequency energy ratio from the envelope spectrum, it achieves directional capture and quantitative characterization of internal fault features such as gear wear.

[0066] A normal acoustic baseline is established by a self-learning architecture that integrates a support vector machine and a Gaussian mixture model, enabling the system to adapt to fluctuations in normal operating conditions. By combining real-time anomaly scoring, historical statistics, and load status to dynamically update the judgment threshold, the monitoring system can adapt to different operating stages and load conditions of the equipment, and continuously maintain high detection sensitivity and low false alarm rate.

[0067] Ultimately, through sliding window decision-making and a multi-level alarm mechanism, reliable judgment and timely early warning of abnormal states were achieved. The entire technical solution forms a complete closed loop from signal perception, feature extraction, intelligent analysis to decision output, ultimately realizing real-time, accurate, and reliable monitoring of the actuator's health status. Attached Figure Description

[0068] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0069] Figure 1 This is a flowchart illustrating a method for determining the operating status of a device according to the present invention. Detailed Implementation

[0070] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0071] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0072] The first embodiment of the present invention provides a method for determining the operating status of a device, comprising:

[0073] Step S10: Acquire raw acoustic signals from the acoustic sensor;

[0074] Step S20: Preprocess the original acoustic signal, including filtering and noise reduction operations, to extract the target frequency band signal and suppress environmental noise;

[0075] Step S30: Extract a set of predefined acoustic features from the preprocessed signal, the acoustic features including envelope spectrum features, spectral energy distribution features and modulation features;

[0076] Step S40: Input the acoustic features into the trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions.

[0077] Step S50: Based on the latest acquired comprehensive anomaly score sequence and historical anomaly score data, dynamically update the threshold for judging anomaly status. The threshold can adaptively reflect the statistical distribution characteristics of the equipment operating status.

[0078] Step S60: Compare the currently obtained anomaly score with the threshold, and determine whether the device is in an abnormal state based on the comparison result;

[0079] If an abnormality is detected, an alarm signal is generated and maintenance suggestions are output.

[0080] To more clearly illustrate the method for determining the operating status of equipment according to the present invention, the following description is provided in conjunction with... Figure 1 The steps in the embodiments of the present invention are described in detail below:

[0081] Step S10: Acquire raw acoustic signals from the acoustic sensor;

[0082] In this embodiment, a vibration sensor is installed on the surface of the device housing, and a reference microphone is deployed in the acoustic environment where the device is located. Signals from the vibration sensor and the reference microphone are collected synchronously as the original acoustic signals.

[0083] The device is an electric servo actuator.

[0084] Specifically, raw acoustic signals from two sensors are simultaneously acquired: a vibration sensor mounted on the surface of the electric servo actuator housing and a reference microphone deployed in the acoustic environment in which the device is located. The vibration sensor directly acquires the sound signal generated by the vibration of the device housing, while the reference microphone is used to acquire the background noise signal in the environment. The two are strictly synchronized in time, providing the raw data foundation for subsequent signal processing and analysis.

[0085] In this embodiment, the vibration sensor is an ICP-type accelerometer, installed on the actuator cylinder housing, with a sampling rate ≥40kHz. The reference microphone is an omnidirectional microphone, deployed in the simulated cabin noise source area; the synchronization accuracy of the two sensor signals is ≤10μs.

[0086] Step S20: Preprocess the original acoustic signal, including filtering and noise reduction operations, to extract the target frequency band signal and suppress environmental noise;

[0087] In this embodiment, the acquired raw acoustic signal is first preprocessed, mainly including two steps: filtering and noise reduction. A 6th-order Butterworth filter with a passband frequency range of [5kHz, 15kHz] is used to bandpass filter the raw signal, and its transfer function is:

[0088] ;

[0089] Where s is a complex frequency variable, Using the center frequency as the reference, high-frequency target frequency band signals related to the meshing and wear of the actuator cylinder gear are extracted.

[0090] To suppress environmental noise interference, an adaptive noise cancellation technique based on the least mean square algorithm is further adopted. This algorithm uses the environmental noise collected by the reference microphone as the reference input. The filtered signal from the vibration sensor is used as the desired response. The filter weight vector is updated iteratively. To minimize error signal The weight update and error calculation follow the formula below:

[0091] ; ;

[0092] in, μ The step size factor has a value range of [0.01, 0.1]. This process can effectively cancel out environmental noise components and finally output a pre-processed, pure target acoustic signal.

[0093] Step S30: Extract a set of predefined acoustic features from the preprocessed signal, the acoustic features including envelope spectrum features, spectral energy distribution features and modulation features;

[0094] Specifically, the steps include the following:

[0095] Step S31: Perform envelope analysis on the preprocessed signal to obtain its envelope spectrum;

[0096] Step S32: Within a predetermined frequency range of the envelope spectrum, calculate the maximum value of its amplitude to characterize the intensity of a specific mechanical impact; wherein the specific mechanical impact includes at least gear meshing impact;

[0097] Step S33: Calculate the ratio of the energy of the envelope spectrum in the preset sideband frequency range to the total energy of the envelope spectrum in the full frequency range, in order to characterize the significance of the modulation effect;

[0098] Step S34: Calculate the ratio of the energy of the original spectrum in the preset high-frequency band to the total energy of the original spectrum in the full frequency band, in order to characterize the energy proportion of high-frequency friction or impact noise.

[0099] Step S35: Combine the calculated maxima, sideband energy ratio, and high-frequency energy ratio into a set of predefined acoustic feature vectors.

[0100] In this embodiment, the preprocessed signal is first subjected to envelope analysis to obtain its envelope spectrum. This process extracts the analytic signal of the signal and calculates its magnitude through Hilbert transform, thereby obtaining the envelope line that reflects the signal amplitude variation law. Then, Fourier transform is performed on this envelope line to obtain the final envelope spectrum.

[0101] On the calculated envelope spectrum, the maximum value of its amplitude is searched and recorded in the range near the pre-set meshing frequency. This maximum value is defined as the meshing impact intensity, which is used to quantitatively characterize the intensity level of a specific impact event generated in the mechanical system.

[0102] Furthermore, the ratio of the energy of the envelope spectrum in the preset sideband frequency range of 1kHz to 3kHz to its total energy across the entire frequency band is calculated. This ratio is defined as the sideband energy ratio, and its magnitude directly reflects the significance of the amplitude modulation effect caused by the fault. Simultaneously, the ratio of the energy of the original spectrum in the high-frequency band of 5kHz to 15kHz to the total energy of the original spectrum across the entire frequency band is calculated. This ratio is defined as the high-frequency energy proportion, used to characterize the energy proportion of the high-frequency noise component generated by abnormal friction or impact in the entire acoustic signal.

[0103] Finally, the three quantitative indicators of meshing impact strength, sideband energy ratio and high-frequency energy ratio are combined into a multi-dimensional vector, which constitutes a set of predefined acoustic feature vectors for subsequent state identification and fault diagnosis.

[0104] Step S40: Input the acoustic features into the trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions.

[0105] Specifically, the steps include the following:

[0106] Step S41: After the device is started and confirmed to be in a fault-free operating state, acoustic feature data is continuously collected during a predetermined learning phase.

[0107] Step S42: When the amount of collected fault-free acoustic feature data reaches a preset minimum requirement, the output is a dataset, and the dataset is used to train a support vector machine model and a Gaussian mixture model.

[0108] Step S43: The acoustic features extracted in real time are input into the support vector machine model and the Gaussian mixture model respectively to calculate the corresponding first anomaly score and second anomaly score.

[0109] Step S44: Perform a weighted fusion calculation on the first abnormal score and the second abnormal score to obtain a comprehensive abnormal score.

[0110] The specific process during implementation is as follows:

[0111] After the equipment is started and confirmed to be in a stable and fault-free initial learning phase, the system begins to continuously collect acoustic feature data. This predetermined learning phase is typically set for the first 50 hours of operation after the equipment is put into service. When the accumulated fault-free feature data reaches the preset minimum sample size requirement, i.e., 500 feature vectors, the dataset is output and used to train two benchmark models in parallel: a single-class support vector machine model and a Gaussian mixture model. The single-class support vector machine model uses a radial basis function as the kernel function, with its kernel parameter γ set to 0.1 and outlier scaling parameter ν set to 0.01. The Gaussian mixture model is set to have 3 components to better model the probability distribution of normal operating condition data.

[0112] After training, the entire fault-free dataset is re-input into the two trained models, and the anomaly scores are calculated for each. The first anomaly score output by the support vector machine model ranges from -1 to 1, with higher values ​​indicating greater anomalies. The second anomaly score output by the Gaussian mixture model is the negative log-likelihood value, with higher values ​​also indicating greater anomalies. Based on these two anomaly scores for all normal samples, their statistical mean and standard deviation are calculated.

[0113] Subsequently, the system combines the calculated statistical mean and standard deviation, a preset load sensitivity factor that decays exponentially based on equipment uptime, and the deviation coefficient between the current real-time load and the baseline load state to dynamically calculate and update the adaptive threshold for health status assessment in real time. The threshold is calculated following the principle of adding three times the standard deviation to the mean, and then superimposing the product of the load sensitivity factor and the load deviation coefficient.

[0114] During real-time monitoring, for each newly extracted acoustic feature vector, the system simultaneously inputs it into a deployed support vector machine model and a Gaussian mixture model to calculate the corresponding first and second anomaly scores, respectively. Finally, the two anomaly scores are weighted and geometrically averaged using preset weights (0.6 for the first anomaly score and 0.4 for the second anomaly score) to obtain a final comprehensive anomaly score, which is used for subsequent health status decisions.

[0115] In constructing the binary classifier for state classification, a support vector machine model is used, and its training process is achieved by solving the following optimization problem:

[0116] ;

[0117] Constraints must be met:

[0118] ;

[0119] Where w is the normal vector of the hyperplane, is a weight vector to be determined, and b is the bias term. Let w be the square of the L2 norm of w, i represent the i-th training sample, and n be the total number of training samples. y is a slack variable, a variable to be determined that is greater than or equal to 0, used to measure the degree to which the i-th sample is allowed to violate the classification margin constraint. i Let x be the true class label of the i-th training sample. i Let be the original feature vector of the i-th training sample. For kernel function mapping, T represents transpose.

[0120] in, This represents the sample feature vector mapped to the high-dimensional feature space through the kernel function. This implementation uses the radial basis function as the kernel function, defined as:

[0121] ;

[0122] In the specific implementation, the key parameters are set as follows: penalty factor C = 1.5, and the scaling parameter of the radial basis kernel function. =0.8. To address the data imbalance problem, a weight coefficient of 2 is set for outlier class samples during the model training phase. This results in a higher misclassification cost during optimization, thereby improving the detection capability for rare outlier states.

[0123] Step S50: Based on the latest acquired comprehensive anomaly score sequence and historical anomaly score data, dynamically update the threshold for judging anomaly status. The threshold can adaptively reflect the statistical distribution characteristics of the equipment operating status.

[0124] In this embodiment, the threshold for judging abnormal states is dynamically updated based on the latest acquired abnormal scoring sequence and historical abnormal scoring data. The method is as follows:

[0125] Step S51: Continuously acquire the comprehensive anomaly score generated by real-time calculation to form the latest anomaly score sequence;

[0126] Step S52: The latest abnormal score sequence is fused with the historically stored normal operating condition abnormal score data to generate an updated abnormal score dataset.

[0127] Step S53: Based on the updated abnormal score dataset, calculate the statistical mean and statistical standard deviation of the abnormal scores;

[0128] Step S54: Based on the calculated statistical mean and statistical standard deviation, combined with the current running time and load status, perform the threshold generation calculation again to update the threshold.

[0129] In step S51, the system continuously monitors and captures the comprehensive device anomaly scores generated in real time by the machine learning model, and appends these time-sequentially generated score values ​​to a predefined storage structure, thereby forming a continuously updated sequence of the latest anomaly scores that reflects the recent operating status of the device. This sequence is typically managed with a fixed-length sliding time window (e.g., containing data from the most recent N hours or M sampling points) to ensure that it is both representative of the current state and not too lengthy.

[0130] In step S52, to ensure the stability and historical representativeness of the threshold update, the system merges the latest anomaly score sequence composed of real-time data with the large-scale anomaly score data accumulated over a long period under known normal operating conditions stored in the historical database. Specifically, the fusion method can be to take the union of these two data sets to form an updated anomaly score dataset for threshold calculation. To ensure the validity of historical data, a time decay factor can be applied to the historical data, or only a subset of historical data matching the current operating conditions (such as load and speed) can be selected for fusion.

[0131] Based on the updated anomaly score dataset generated in step S52, step S53 performs the core statistical calculations. The system iterates through all anomaly score samples in the dataset and calculates their arithmetic mean and standard.

[0132] The calculated mean and standard deviation together describe the central tendency and dispersion of the abnormal scores of the equipment under normal conditions, providing an accurate statistical basis for the dynamic generation of subsequent thresholds.

[0133] Step S54: Based on the calculated statistical mean and statistical standard deviation, and combined with the current running time and load status, perform the threshold generation calculation again to update the threshold, including:

[0134] Step S541: Obtain the calculated updated statistical mean and updated statistical standard deviation;

[0135] Step S542: Obtain the load sensitivity factor corresponding to the current cumulative running time of the device, wherein the value of the load sensitivity factor decreases according to a predetermined rule as the running time increases.

[0136] Step S543: Obtain the load deviation coefficient, which characterizes the difference between the current operating load state and the baseline load state;

[0137] Step S544: Add the updated statistical mean, the updated statistical standard deviation by a preset multiple, the product of the load sensitivity factor and the load deviation coefficient to generate the updated adaptive threshold.

[0138] In this embodiment, the statistical mean and statistical standard deviation, recalculated from the previous steps based on the latest dataset, are first obtained. Then, the system retrieves the corresponding load sensitivity factor k from a predefined time-sensitivity factor lookup table or function F(T) based on the device's current cumulative operating time T. This factor is designed to reflect the decreased tolerance of the device to load fluctuations as operating time increases and components age; therefore, its value decreases with increasing operating time T according to a predetermined rule (such as linear or exponential decay). Next, the system queries the real-time monitoring system to obtain the actual operating load of the device and compares it with a preset benchmark load to calculate a load deviation coefficient. Its value is given by the formula =|L current -L base | / L base Determined to quantify the deviation of the current load state from the ideal baseline state, where L current L represents the actual operating load of the current equipment. base This is the preset baseline load.

[0139] Finally, the system substitutes all the above parameters into the following formula for synthesis calculation: Updated Adaptive Threshold The core idea of ​​this formula is that the dynamic threshold consists of the mean representing the location of the normal score set, three times the standard deviation representing the normal fluctuation range, and a compensation term for fine-tuning based on the degree of equipment aging and real-time load conditions. Together, they form a system that enables the threshold to adapt to both the device's own state evolution and external working conditions.

[0140] in, The standard deviation of abnormal scores in normal samples. This represents the mean score of abnormal samples from normal samples.

[0141] Step S60: Compare the currently obtained anomaly score with the threshold, and determine whether the device is in an abnormal state based on the comparison result;

[0142] If an abnormality is detected, an alarm signal is generated and maintenance suggestions are output.

[0143] Specifically, the comprehensive anomaly score calculated in real time is compared with the current threshold.

[0144] If the overall anomaly score is greater than the threshold, a first judgment signal indicating an abnormal health status is generated.

[0145] If the overall abnormal score is less than or equal to the threshold, a second judgment signal indicating a normal health status is generated.

[0146] Based on the generated judgment signal, the final judgment result of the real-time health status of the device is output.

[0147] If the final determination result is an abnormal health status, an alarm signal with a predetermined level will be generated.

[0148] Based on the magnitude of the comprehensive anomaly score and the degree of degradation of specific indicators in the acoustic features, corresponding maintenance recommendations are selected from a predetermined set of maintenance measures and generated.

[0149] Output the alarm signal and the maintenance recommendations.

[0150] In this embodiment, the system compares the real-time calculated comprehensive anomaly score with the currently effective adaptive dynamic threshold, and performs device status determination and decision output based on the comparison result. Specifically, the system continuously reads the latest comprehensive anomaly score output by the machine learning model and compares it in real-time with the current threshold determined by the aforementioned dynamic update process. If the comprehensive anomaly score is significantly greater than the current threshold, the system generates a first determination signal indicating an abnormal device health status; conversely, if the score is less than or equal to the threshold, a second determination signal indicating a normal health status is generated. Based on this determination signal, the system ultimately determines and outputs a conclusion regarding the real-time health status of the device.

[0151] When the final determination is "abnormal health status," the system immediately activates the alarm generation process. This process first determines the predetermined alarm level based on the specific extent to which the comprehensive anomaly score exceeds a threshold (e.g., exceeding by 20% is a "warning" level, exceeding by 50% is a "serious" level), and generates an alarm signal of the corresponding level. Simultaneously, the system initiates the maintenance suggestion generation module. This module not only considers the absolute value of the anomaly score but also deeply analyzes the degree of degradation of specific indicators in the underlying acoustic characteristics that constitute the comprehensive score. For example, a significant increase in spectral energy in a specific high-frequency band may indicate early bearing failure, while an abnormal increase in a characteristic frequency component in the envelope spectrum may point to localized damage to the gearbox. The system matches these degradation patterns of characteristic indicators with a predefined knowledge base containing various maintenance measures (e.g., "suggest lubrication within one week," "suggest immediate shutdown to check bearing condition," "suggest checking gearbox alignment"), selecting one or more maintenance suggestions that best match the current fault symptoms. Finally, the system outputs the generated alarm signal and maintenance suggestions to the human-machine interface or the upper-level monitoring system, thus completing the closed loop from status perception to decision support.

[0152] In another embodiment of the present invention, if an abnormality is determined, an alarm signal is generated and maintenance suggestions are output, and the method further includes:

[0153] If the health status is determined to be abnormal, the abnormality interpretability analysis process will be initiated.

[0154] Based on the comprehensive anomaly score and the acoustic features, the contribution of each sub-feature constituting the acoustic features to the comprehensive anomaly score is analyzed and determined.

[0155] The sub-features are sorted according to their contribution, and one or more sub-features with the highest contribution are identified as key features that cause the current abnormal health status.

[0156] The key features are mapped and matched with a pre-stored fault feature knowledge base, which records the correspondence between different key features and potential fault causes and maintenance operations.

[0157] Based on the mapping matching results, diagnostic information is generated and output to indicate the cause of potential failures.

[0158] The diagnostic information, alarm signals, and maintenance recommendations will be output together.

[0159] In implementation of this invention, if an abnormal health status is determined, the system automatically initiates an anomaly interpretability analysis process. This process first obtains the acoustic feature vector that currently generates a high anomaly score and its corresponding comprehensive anomaly score. Then, the system analyzes the score result using a mechanism based on feature perturbation sensitivity analysis. Specifically, the system sequentially simulates the changes in the value of each sub-feature in the feature vector within its possible range, keeping all other sub-feature values ​​constant in each simulation. After each simulation change, the system re-inputs the newly constructed feature vector into the trained machine learning model, thus obtaining a new simulated anomaly score. By calculating the difference between this new score and the original actual score, the system quantifies the influence of the currently perturbed sub-feature on the final comprehensive anomaly score. After all sub-features have undergone the above simulation analysis, the system normalizes the influence of all features, ultimately obtaining the percentage contribution of each sub-feature to the current anomaly event, thus completing the process from black-box model output to transparent contribution analysis.

[0160] After obtaining the contribution values ​​of all sub-features, the system sorts all sub-features from highest to lowest based on their contribution values, and identifies the top-ranked sub-features, i.e., those with the highest contribution values, as the key features causing the current health status abnormality. These key features usually directly correspond to the abnormal state of certain specific physical components inside the equipment. For example, the highest contribution value is determined to come from the sub-feature representing meshing impact intensity.

[0161] Next, the system maps and matches one or more of the identified key features to a pre-stored fault feature knowledge base. This knowledge base is a structured database pre-built using domain expert experience and historical fault case data. It details the correspondence between different key features, combinations of key features, potential fault causes, and recommended maintenance operations. For example, the knowledge base might define a rule: if both "sideband energy ratio" and "high-frequency energy percentage" are identified as key features, then the mapped potential fault cause is "early bearing wear," and the corresponding recommended maintenance operation is "lubricate the bearing and recommend a vibration re-inspection within one week."

[0162] Based on the current mapping and matching results, the system automatically generates a diagnostic description that indicates the potential cause of the fault. This description is not a simple alarm, but includes a professional interpretation of key characteristics, the inferred fault mechanism, and a detailed description of the names of potentially affected components, thus forming a diagnostic conclusion with direct guidance.

[0163] Ultimately, the system integrates this diagnostic information with the initially generated alarm signal and maintenance recommendations based on the rating level, and outputs it to maintenance personnel through channels such as the human-machine interface, the factory manufacturing execution system, or a remote operation and maintenance platform. This provides maintenance personnel with more than just an "abnormal" alarm; it provides a comprehensive report containing preliminary fault diagnosis conclusions, decision-making basis, and targeted handling recommendations, greatly improving the efficiency and accuracy of maintenance decisions.

[0164] More specifically, the method for analyzing and determining the contribution of each sub-feature constituting the acoustic features to the comprehensive anomaly score based on the comprehensive anomaly score is as follows:

[0165] Obtain the acoustic feature vector that is currently causing the abnormal health status. This vector consists of multiple feature values.

[0166] The values ​​of each component of the acoustic feature vector are simulated to change within their respective ranges in turn, while keeping the values ​​of other components constant.

[0167] After each simulated change, the new feature vector is input into the trained machine learning model to obtain its corresponding new comprehensive anomaly score.

[0168] The influence of the sub-features corresponding to this simulation on the comprehensive anomaly score is quantified by comparing the difference between the new comprehensive anomaly score and the current actual comprehensive anomaly score.

[0169] The influence levels corresponding to all sub-features are normalized to obtain the contribution of each sub-feature to the current comprehensive anomaly score.

[0170] The method for analyzing and determining the contribution of each sub-feature constituting the acoustic features to the comprehensive anomaly score, based on the comprehensive anomaly score and the acoustic features, is implemented through a systematic feature perturbation and sensitivity analysis process. The system first obtains the acoustic feature vector corresponding to the current state determined to be abnormal. This vector consists of multiple predefined sub-feature values, such as the specific values ​​of the meshing impact strength, sideband energy ratio, and high-frequency energy proportion.

[0171] The system then enters a cyclical processing procedure: for each feature item in the feature vector, it simulates a representative change within its possible numerical range. A typical simulation method is to replace the feature value with the statistical mean of that feature under historical normal conditions. While simulating the change of the current feature item, the system keeps the values ​​of all other features in the feature vector completely unchanged to isolate the influence of a single feature. After each simulation change is completed, the system uses the newly generated feature vector as a new input and feeds it back into the previously trained machine learning model to obtain a new comprehensive anomaly score output by the model for this simulated vector.

[0172] Subsequently, the system precisely quantifies the influence of the sub-feature corresponding to this simulation operation on the final comprehensive anomaly score by calculating the absolute difference or relative rate of change between the new comprehensive anomaly score and the current actual anomaly score. This influence is an absolute value characterizing its sensitivity to change. After all sub-features have completed the above simulation, re-scoring, and difference calculation processes, the system obtains the original influence value set for each feature. Finally, the system uses the softmax function or the min-max normalization method to normalize all the original influence values, so that the sum of the contributions of all features is 100%, thus finally outputting a set of clear and comparable contribution percentages. Each percentage represents the contribution of a sub-feature to triggering this anomaly score.

[0173] Although the steps in the above embodiments are described in the above order, those skilled in the art will understand that in order to achieve the effect of this embodiment, different steps do not need to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple variations are all within the protection scope of this invention.

[0174] A system for determining the operating status of a device according to a second embodiment of the present invention, based on a method for determining the operating status of a device according to a first embodiment, the system comprising:

[0175] A signal acquisition module configured to acquire raw acoustic signals from an acoustic sensor;

[0176] A preprocessing module is configured to preprocess the original acoustic signal, including filtering and noise reduction operations, to extract the target frequency band signal and suppress ambient noise;

[0177] The feature extraction module is configured to extract a set of predefined acoustic features from the preprocessed signal, the acoustic features including envelope spectrum features, spectral energy distribution features and modulation features;

[0178] An anomaly score calculation module is configured to input the acoustic features into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions.

[0179] The threshold update module is configured to dynamically update the threshold for judging abnormal states based on the latest acquired comprehensive abnormal score sequence and historical abnormal score data. The threshold can adaptively reflect the statistical distribution characteristics of the device's operating status.

[0180] The status judgment module is configured to compare the currently obtained anomaly score with the threshold and determine whether the device is in an abnormal state based on the comparison result.

[0181] The maintenance module is configured to generate an alarm signal and output maintenance suggestions if an abnormality is detected.

[0182] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the system described above can be found in the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0183] It should be noted that the system for determining the operating status of a device provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the various modules or steps and are not considered as an improper limitation of the present invention.

[0184] An electronic device according to a third embodiment of the present invention includes:

[0185] At least one processor; and

[0186] A memory communicatively connected to at least one of the processors; wherein,

[0187] The memory stores instructions that can be executed by the processor to implement the method described above for determining the operating state of a device.

[0188] A fourth embodiment of the present invention provides a computer-readable storage medium storing computer instructions, which are executed by the computer to implement the above-described method for determining the operating state of a device.

[0189] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the storage device and processing device described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0190] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps can be placed in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art 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 invention.

[0191] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0192] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0193] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for determining the operating status of equipment, characterized in that, include: Acquire raw acoustic signals from acoustic sensors; The original acoustic signal is preprocessed, including filtering and noise reduction operations, to extract the target frequency band signal and suppress environmental noise; Extract a set of predefined acoustic features from the preprocessed signal, including envelope spectrum features, spectral energy distribution features, and modulation features: Extracting a set of predefined acoustic features from the preprocessed signal, including envelope spectrum features, spectral energy distribution features, and modulation features, using the following method: Envelope analysis is performed on the preprocessed signal to obtain its envelope spectrum; Within a predetermined frequency range of the envelope spectrum, the maximum value of its amplitude is calculated to characterize the intensity of a specific mechanical impact; wherein the specific mechanical impact includes at least gear meshing impact; The ratio of the energy of the envelope spectrum within a preset sideband frequency range to the total energy of the envelope spectrum across the entire frequency band is calculated to characterize the significance of the modulation effect. The ratio of the energy of the original spectrum in a preset high-frequency band to the total energy of the original spectrum in the full frequency band is calculated to characterize the energy proportion of high-frequency friction or impact noise. The calculated maxima, sideband energy ratio, and high-frequency energy ratio are combined into a set of predefined acoustic feature vectors. The acoustic features are input into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions. Based on the latest acquired comprehensive anomaly scoring sequence and historical anomaly scoring data, the threshold for judging anomaly status is dynamically updated. The threshold can adaptively reflect the statistical distribution characteristics of the equipment's operating status. The currently obtained anomaly score is compared with the threshold, and the device is determined to be in an abnormal state based on the comparison result. If an abnormality is detected, an alarm signal is generated and maintenance suggestions are output.

2. The method for determining the operating status of equipment according to claim 1, characterized in that, The method for acquiring raw acoustic signals from acoustic sensors is as follows: A vibration sensor is installed on the surface of the device housing, and a reference microphone is deployed in the acoustic environment in which the device is located. Signals from the vibration sensor and the reference microphone are collected synchronously as the raw acoustic signals.

3. The method for determining the operating status of equipment according to claim 1, characterized in that, The acoustic features are input into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions, and the method is as follows: Acoustic feature data is continuously collected during a predetermined learning phase after the device is started and confirmed to be in a fault-free operating state; When the amount of collected fault-free acoustic feature data reaches a preset minimum requirement, the output is a dataset, which is then used to train a support vector machine model and a Gaussian mixture model. The acoustic features extracted in real time are input into the support vector machine model and the Gaussian mixture model respectively to calculate the corresponding first anomaly score and second anomaly score. A weighted fusion calculation is performed on the first abnormal score and the second abnormal score to obtain a comprehensive abnormal score.

4. The method for determining the operating status of equipment according to claim 1, characterized in that, Based on the latest acquired anomaly scoring sequence and historical anomaly scoring data, the threshold for judging anomaly states is dynamically updated. The method is as follows: Continuously acquire comprehensive anomaly scores generated by real-time calculations to form the latest anomaly score sequence; The latest anomaly score sequence is fused with the historically stored normal operating condition anomaly score data to generate an updated anomaly score dataset. Based on the updated anomaly scoring dataset, the statistical mean and statistical standard deviation of the anomaly scores are calculated; Based on the calculated statistical mean and statistical standard deviation, combined with the current running time and load status, the threshold generation calculation is performed again to update the threshold.

5. The method for determining the operating status of equipment according to claim 4, characterized in that, Based on the calculated statistical mean and standard deviation, combined with the current running time and load status, the threshold generation calculation is performed again. The method is as follows: Obtain the calculated updated statistical mean and updated statistical standard deviation; Obtain the load sensitivity factor corresponding to the current cumulative running time of the device, wherein the value of the load sensitivity factor decreases according to a predetermined rule as the running time increases; Obtain the load deviation coefficient, which characterizes the difference between the current operating load state and the baseline load state; The updated threshold is generated by adding the updated statistical mean, the updated statistical standard deviation by a preset multiple, the load sensitivity factor, and the load deviation coefficient.

6. The method for determining the operating status of equipment according to claim 1, characterized in that, The method for comparing the currently obtained anomaly score with the threshold and determining whether the device is in an abnormal state based on the comparison result is as follows: The comprehensive anomaly score calculated in real time is compared with the current threshold. If the overall anomaly score is greater than the threshold, a first judgment signal indicating an abnormal health status is generated. If the overall abnormal score is less than or equal to the threshold, a second judgment signal indicating a normal health status is generated. Based on the generated judgment signal, the final judgment result of the real-time health status of the device is output.

7. A method for determining the operating status of equipment according to claim 6, characterized in that, If an anomaly is detected, an alarm signal is generated and maintenance suggestions are output. The method is as follows: If the final determination result is an abnormal health status, an alarm signal with a predetermined level will be generated. Based on the magnitude of the comprehensive anomaly score and the degree of degradation of specific indicators in the acoustic features, corresponding maintenance recommendations are selected from a predetermined set of maintenance measures and generated. Output the alarm signal and the maintenance recommendations.

8. A system for determining the operating status of equipment, based on the method for determining the operating status of equipment according to any one of claims 1-7, characterized in that, The system includes: A signal acquisition module configured to acquire raw acoustic signals from an acoustic sensor; A preprocessing module is configured to preprocess the original acoustic signal, including filtering and noise reduction operations, to extract the target frequency band signal and suppress ambient noise; The feature extraction module is configured to extract a set of predefined acoustic features from the preprocessed signal, the acoustic features including envelope spectrum features, spectral energy distribution features and modulation features; An anomaly score calculation module is configured to input the acoustic features into a trained machine learning model to obtain a comprehensive anomaly score; the machine learning model is trained based on historical acoustic data under normal operating conditions. The threshold update module is configured to dynamically update the threshold for judging abnormal states based on the latest acquired comprehensive abnormal score sequence and historical abnormal score data. The threshold can adaptively reflect the statistical distribution characteristics of the device's operating status. The status judgment module is configured to compare the currently obtained anomaly score with the threshold and determine whether the device is in an abnormal state based on the comparison result. The maintenance module is configured to generate an alarm signal and output maintenance suggestions if an abnormality is detected.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement a method for determining the operating state of a device as described in any one of claims 1-7.

Images