A device fault detection method and inspection system based on vibration acoustic method

Abstract

The present application relates to the technical field of equipment fault detection, and in particular to an equipment fault detection method and inspection system based on a vibroacoustic method, wherein the method includes a data acquisition step, a preprocessing step, a classification step, a first judgment step, a fault determination step, a vibration intervention step, a second judgment step, a risk warning step, and a normal detection step; wherein the inspection system includes a data acquisition module, a preprocessing module, a classification module, a first judgment module, a fault determination module, a vibration intervention module, a second judgment module, a risk warning module, and a normal detection module. The present application has the effect of improving the accuracy of fault detection results.

Description

Technical Field

[0001] The present application relates to the technical field of equipment fault detection, and in particular to an equipment fault detection method and inspection system based on a vibroacoustic method. Background Art

[0002] At present, the vibration signal generated by the equipment during operation contains a large amount of status information. When the equipment has defects or failures, its vibration will change to varying degrees according to the degree of wear. These characteristics can be used as important parameters for diagnosing equipment defects and failures.

[0003] In the prior art, vibration-based fault detection technology mainly relies on vibration sensors, which obtain vibration data of the component to be detected by installing the vibration sensor on the component to be detected or as close to the detection component as possible. However, in actual operation, a situation is often encountered: the vibration frequency of the equipment has not changed significantly, but an abnormal sound has appeared inside the equipment. This abnormal sound may be caused by slight wear and tear of mechanical parts, looseness, electrical problems or material problems. The above method cannot detect and deal with these potential faults in a timely manner, resulting in inaccurate fault detection results of the equipment. Summary of the invention

[0004] In order to improve the accuracy of fault detection results, the present application provides an equipment fault detection method and inspection system based on a vibroacoustic method.

[0005] In the first aspect, the present application provides a device fault detection method based on a vibroacoustic method, which adopts the following technical solution:

[0006] A method for detecting equipment faults based on a vibroacoustic method comprises the following steps:

[0007] Data collection: Collect equipment vibration signals and obtain the duration of the collection period. The vibration frequency in the monitoring device is converted into a frequency distribution diagram to obtain vibration voiceprint information; the acoustic signal is collected, where the acoustic signal includes environmental sound, the sound of the monitoring device itself and the sound of the detected device, and the voiceprint information of the acoustic signal is obtained, which is recorded as the first voiceprint information, the second voiceprint information and the third voiceprint information respectively;

[0008] Preprocessing: removing the first voiceprint information and the second voiceprint information contained in the third voiceprint information to obtain the fourth voiceprint information, and then extracting the information matching the fourth voiceprint information and the vibration voiceprint information to obtain the valid voiceprint information emitted by the detected device;

[0009] Grading: Set voiceprint restriction information, determine whether the valid voiceprint information meets the voiceprint restriction information, classify the valid voiceprint information that exceeds the restriction information, and record it as the initial classification. When classifying, the closer it is to the voiceprint restriction information, the lower its grade;

[0010] First judgment: judging whether the vibration frequency is greater than or equal to the preset vibration limit value, if so, executing the fault determination step, otherwise, executing the vibration intervention step;

[0011] Fault determination: Determine if the equipment is faulty and remind staff to shut down for maintenance;

[0012] Vibration intervention: adjust and intervene the initial classification of valid voiceprint information according to the vibration limit and vibration frequency to obtain the intervention classification; the closer the vibration frequency is to the vibration limit, the greater the increase in the classification;

[0013] Second judgment: judge whether the intervention classification exceeds the critical level. If so, execute the risk warning step; otherwise, execute the normal detection step;

[0014] Risk warning: setting key monitoring cycles , focus on monitoring the equipment;

[0015] Normal detection: Perform normal periodic monitoring on the equipment.

[0016] By adopting the above technical solution, combining vibration signals and sound signals, the operating status of the equipment can be captured more comprehensively. The vibration signal can reflect the mechanical structure problems of the equipment, while the acoustic signal can provide more information about the internal operation of the equipment and the external environment. The combination of the two can more accurately analyze the location of the fault and fault diagnosis monitoring, which is convenient for improving the accuracy of fault diagnosis, and can quickly and accurately identify equipment faults, shortening the time for fault discovery. Accurate fault warning and positioning help reduce unnecessary downtime and maintenance costs, and improve equipment utilization and economic benefits.

[0017] Through the preprocessing steps, the influence of environmental sound and the sound of the monitoring equipment itself is effectively eliminated, making the analysis of the sound of the detected equipment more accurate, which helps to reduce false alarms and missed alarms and improve the accuracy of fault detection.

[0018] Voiceprint limitation information is set, and valid voiceprint information is graded, so that potential faults can be discovered in time and corresponding measures can be taken; in addition, the initial classification is adjusted through the vibration intervention step, which further improves the sensitivity and accuracy of the early warning.

[0019] This method achieves automatic detection and early warning of equipment failures by setting parameters such as preset values ​​and critical levels, which reduces the need for human intervention, improves detection efficiency, and reduces labor costs.

[0020] Optionally, a frequency estimation step is further provided between the second judgment step and the normal detection step;

[0021] Frequency estimation: The growth of the vibration frequency is predicted through initial classification to obtain the predicted frequency, and then the first judgment step is performed again based on the predicted frequency.

[0022] By adopting the above technical solution, the frequency estimation step can predict the future vibration frequency based on the current vibration frequency and initial classification, which allows the system to issue early warning and intervention based on the predicted frequency before the fault actually occurs, thereby greatly improving the advance amount of fault detection; through frequency estimation, the system can adjust its monitoring and early warning strategies in real time and dynamically; if it is predicted that the vibration frequency will approach or exceed the preset vibration limit, the system can take immediate measures, such as adjusting the monitoring frequency, triggering a higher level of early warning, etc.; the frequency estimation step combines the current status and historical data of the equipment (reflected by the initial classification) to make the prediction results more accurate; this helps to reduce false alarms and missed alarms and improve the accuracy of fault detection; after introducing frequency estimation, the system can better cope with changes in equipment status and the impact of the external environment. Even if the equipment status fluctuates or the external environment changes, the system can make timely adjustments through predictions to maintain its stability and reliability.

[0023] Optionally, the frequency estimation step further includes acquiring the temperature, determining the temperature rising trend, and adjusting the predicted frequency according to the temperature rising trend.

[0024] By adopting the above technical solution, by introducing temperature monitoring and adjusting the prediction frequency according to the temperature rising trend, the vibration characteristics of the equipment under different temperature conditions can be more accurately reflected, thereby improving the accuracy of the prediction; the vibration characteristics of the equipment under different temperature environments may be different. By obtaining temperature information in real time and dynamically adjusting the prediction frequency according to the temperature trend, the prediction model can be better adapted to different temperature conditions and the generalization ability of the model can be improved.

[0025] Temperature rise is often associated with potential failures such as equipment overheating and increased wear. By monitoring temperature trends and taking this factor into account in the predicted frequency, the risk of possible equipment failure can be discovered in advance, so that preventive measures can be taken to avoid failures.

[0026] Optionally, a voiceprint analysis step is further provided between the second judgment step and the risk warning step;

[0027] Voiceprint analysis: including trend analysis steps and warning steps;

[0028] Trend analysis: obtain the voiceprint change characteristics within a period of time, analyze the voiceprint feature change trend, perform time series analysis on the effective voiceprint information, and predict the change trend of the voiceprint features. If the trend is greater than the set growth threshold, execute the warning step, otherwise, execute the risk warning step;

[0029] Warning: Send a signal that the equipment is about to fail, reminding staff to check.

[0030] By adopting the above technical solution, through the trend analysis step, the system can obtain the voiceprint change characteristics within a period of time, and conduct in-depth analysis of the change trend of the voiceprint characteristics; this enables the system to issue an early warning before equipment failure occurs, providing staff with sufficient troubleshooting time; at the same time, through the time series analysis of voiceprint information, the system can accurately locate the part or link where the failure may occur, thereby improving the efficiency and accuracy of troubleshooting.

[0031] The risk warning in the existing technology is often based on a single threshold judgment, which is easily affected by factors such as environmental noise and equipment status fluctuations, resulting in false alarms or missed alarms. After adding a trend analysis step, the system can comprehensively consider the changing trend and growth threshold of voiceprint features to improve the accuracy and reliability of the warning.

[0032] The introduction of trend analysis steps makes the entire risk warning system more intelligent and automated; the system can automatically obtain voiceprint information, analyze changing trends, and automatically execute warning or risk warning steps based on the analysis results, reducing the need for manual intervention and judgment, and improving work efficiency and accuracy.

[0033] Optionally, a period adjustment step is provided between the trend analysis step and the risk warning step;

[0034] Cycle adjustment: According to the change trend, determine the change range, and then attenuate the key monitoring cycle according to the change range. As the change range increases, the key monitoring cycle Narrow down, and then perform risk warning steps based on the updated key monitoring cycle.

[0035] By adopting the above technical solution and through the periodic adjustment step, the system can dynamically adjust the risk warning cycle according to the changing trend and amplitude of the voiceprint features. This dynamic adaptability enables the system to more accurately capture subtle changes in the equipment status, thereby issuing warnings earlier before equipment failure occurs.

[0036] The introduction of the periodic adjustment step means that risk warnings no longer rely solely on a single threshold judgment. The system can comprehensively consider the changing trend, change amplitude and updated cycle of the voiceprint features, so as to more accurately judge whether the device status is abnormal and reduce false alarms and missed alarms.

[0037] This technical solution has good flexibility. The system can adjust the cycle adjustment algorithm and parameters according to actual needs to adapt to the needs of different equipment and scenarios. This flexibility enables the system to be more widely used in various industrial fields to improve the reliability and safety of equipment. After adding the cycle adjustment step, the entire risk warning system is more intelligent. The system can automatically obtain voiceprint information, analyze changing trends, adjust cycles and execute risk warning steps, reducing the need for manual intervention and judgment, and improving work efficiency and accuracy.

[0038] Optionally, a cycle determination step is further provided between the cycle adjustment step and the risk warning step;

[0039] Cycle judgment: judge the key monitoring cycle Is it less than or equal to the minimum value of the key monitoring period? , if yes, then execute the warning step, otherwise, execute the risk warning step.

[0040] By adopting the above technical solution, through the cycle judgment step, the system can further judge the equipment status according to the change of the cycle on the basis of risk warning, which is equivalent to adding an additional safety barrier to the warning system and improving the reliability and accuracy of the warning; different equipment failures may have different cycle characteristics, and through the cycle judgment step, the system can respond to different types of failures more flexibly. For example, for some periodic failures, the system can issue a warning in time when the cycle reaches the maximum value, while for non-periodic failures, it can be monitored and warned through the risk warning step.

[0041] The introduction of the cycle judgment step enables the system to more accurately determine whether the equipment status is abnormal; by comprehensively considering the cycle's changing trend, change range, and whether it has reached the maximum value, the system can more accurately formulate early warning strategies, reduce false alarms, and improve the accuracy and effectiveness of early warnings.

[0042] After adding the cycle judgment step, the entire risk warning system becomes more intelligent. The system can automatically obtain voiceprint information, analyze changing trends, adjust the cycle, determine whether the cycle has reached the maximum value, and execute corresponding warning or alert steps based on the judgment results. This highly intelligent warning system can significantly improve work efficiency and accuracy.

[0043] Optionally, after the frequency estimation step, a first risk classification step is further provided;

[0044] First risk classification: including risk estimation step and damage level classification step;

[0045] Risk estimation: Set the frequency danger value and determine whether the predicted frequency exceeds the frequency danger value. If so, perform the damage level classification step;

[0046] Damage level classification: Damage levels are classified according to the interval between the frequency danger value and the vibration limit value. Each damage level corresponds to the urgency of the investigation, and then the damage level is determined based on the estimated frequency.

[0047] By adopting the above technical solution, through the risk prediction step, the system can set the frequency danger value and determine whether the predicted frequency exceeds the value. Once the predicted frequency exceeds the danger value, the system can execute the damage level classification step, thereby preventing potential equipment damage or failure in advance and effectively reducing risks.

[0048] The damage level classification procedure divides the degree of damage into different levels according to the interval between the frequency danger value and the vibration limit value. Each damage level corresponds to a different degree of urgency for investigation, which enables staff to quickly determine the damage level based on the estimated frequency and take corresponding investigation and maintenance measures.

[0049] This technical solution combines frequency estimation, risk estimation and damage level classification to form a complete early warning system. Through this system, the system can more accurately judge the equipment status and improve the accuracy and efficiency of the early warning. At the same time, the damage level classification also provides clear guidance for the staff, making the investigation and maintenance work more orderly and efficient.

[0050] After adding the risk estimation and damage classification steps, the whole system is more intelligent. The system can automatically obtain frequency information, perform frequency estimation, judge risks and classify damage levels, reducing the need for manual intervention and judgment, and improving work efficiency and accuracy.

[0051] Optionally, after the vibration intervention step, a second risk classification step is further provided;

[0052] The second risk classification: including the fault prediction step and the fault possible level classification step;

[0053] Fault prediction: Set the risk level and determine whether the intervention level exceeds the risk level. If so, perform the possible fault level classification step.

[0054] Fault possibility level classification: Fault possibility levels are classified according to hazard level and critical level. Each fault possibility level corresponds to the urgency of troubleshooting, and then the fault possibility level is determined according to the intervention level.

[0055] By adopting the above technical solution, through the fault prediction step, the system can set the danger level and determine whether the current intervention level exceeds this danger threshold. Once exceeded, the system will immediately execute the possible fault level classification step, thereby identifying potential equipment failure risks in advance and providing an important basis for subsequent fault prevention work.

[0056] The fault possibility classification step divides the fault possibility into different levels according to the hazard classification and criticality level. Each level corresponds to a different degree of urgency for investigation. This enables maintenance personnel to make maintenance decisions quickly based on the fault possibility level, reasonably allocate maintenance resources, and improve maintenance efficiency.

[0057] After adding the fault prediction and fault probability level classification steps, the entire detection process is further optimized. The system can quickly screen out equipment that may have faults through the fault prediction step before normal detection, and then conduct targeted detection based on the fault probability level, avoiding the blindness and inefficiency of traditional detection methods.

[0058] Optionally, a fault classification step is also included;

[0059] Fault classification: Determine the estimated possible level of fault based on the damage level and possible level of fault, combined with the risk matrix.

[0060] By adopting the above technical solution, the fault classification step combines the damage level and the possible level of fault, and introduces a risk matrix for comprehensive evaluation; this comprehensive evaluation method can more comprehensively consider various factors of the fault, thereby more accurately determining the possible level of fault estimation.

[0061] Through fault classification, the system can quickly provide maintenance personnel with clear information about the severity and urgency of the fault. This helps maintenance personnel make decisions quickly, allocate maintenance resources reasonably, and improve maintenance efficiency.

[0062] The fault classification step enables companies to have a clearer understanding of the risk status of equipment failures. This helps companies develop more effective risk management strategies and reduce the impact of failures on production operations.

[0063] In the second aspect, the present application provides a patrol inspection system, which adopts the following technical solution:

[0064] A patrol inspection system includes the following modules:

[0065] Data acquisition module: used to collect equipment vibration signals and acoustic signals;

[0066] Preprocessing module: The input end is connected to the output end of the data acquisition module, and is used to preprocess the collected information and then extract the effective voiceprint information emitted by the detected device;

[0067] Classification module: The input end is connected to the pre-processing module and is used to classify the effective voiceprint information;

[0068] The first judgment module: the input end is connected to the output end of the preprocessing module, and is used to judge whether the vibration frequency is greater than or equal to the preset vibration limit;

[0069] Fault determination module: the input end is connected to the output end of the first judgment module, and is used to determine that the equipment has a fault when the vibration frequency is greater than or equal to the preset vibration limit, and remind the staff to shut down the equipment for maintenance;

[0070] Vibration intervention module: the input end is connected to the output end of the first judgment module and the output end of the classification module, and is used to adjust and intervene the initial classification of the effective voiceprint information according to the vibration limit and the vibration frequency when the vibration frequency is less than the preset vibration limit, so as to obtain the intervention classification;

[0071] The second judgment module: the input end is connected to the output end of the vibration intervention module, and is used to judge whether the intervention classification exceeds the critical level;

[0072] Risk warning module: The input end is connected to the output end of the second judgment module, and is used to perform periodic Set and monitor the equipment in key periods;

[0073] Normal detection module: The input end is connected to the output end of the second judgment module, and is used to perform normal periodic monitoring on the equipment when the intervention classification does not exceed the critical level.

[0074] By adopting the above technical solution, the data acquisition module of the inspection system starts working first to collect the vibration signal and acoustic signal of the equipment; the preprocessing module receives the output of the data acquisition module, preprocesses the collected information to eliminate noise and interference, and extracts effective voiceprint information; the grading module grades the effective voiceprint information extracted by the preprocessing module, and the first judgment module determines whether the vibration frequency is greater than or equal to the limit value according to the preset vibration limit value; if the vibration frequency is greater than or equal to the preset vibration limit value, the fault determination module will determine that the equipment has a fault and remind the staff to shut down for maintenance; if the vibration frequency is less than the preset vibration limit value, the vibration intervention module will adjust the initial grading of the effective voiceprint information according to the vibration limit value and the vibration frequency to obtain the intervention grading; the second judgment module receives the output of the vibration intervention module to determine whether the intervention grading exceeds the critical level; if the intervention grading exceeds the critical level, the risk warning module will set a period , and focus on monitoring the equipment according to this cycle; if the intervention classification does not exceed the critical level, the normal detection module will perform normal periodic monitoring on the equipment.

[0075] By collecting the vibration and acoustic signals of the equipment, and performing preprocessing and extracting effective voiceprint information, the system can more accurately determine the operating status and potential faults of the equipment; based on the judgment results of the vibration frequency and intervention classification, the system can automatically determine whether it is necessary to shut down for maintenance or conduct key monitoring, thereby optimizing maintenance decisions and improving maintenance efficiency; by setting a period for key monitoring or normal periodic monitoring of the equipment, the system can promptly detect and handle potential equipment failures, thereby preventing the occurrence of failures and extending the service life of the equipment; the inspection system has a high degree of automation, which can reduce the frequency and intensity of manual inspections, improve inspection efficiency, and reduce inspection costs; by promptly detecting and handling equipment failures, the system can maintain the stable operation of the equipment, improve the reliability of the equipment, and ensure the continuity and stability of the production process.

[0076] In summary, the present application includes at least one of the following beneficial technical effects:

[0077] 1. Combining vibration signals and sound signals can more comprehensively capture the operating status of the equipment. Vibration signals can reflect the mechanical structure problems of the equipment, while acoustic signals can provide more information about the internal operation of the equipment and the external environment. The combination of the two can more accurately analyze the location of the fault and fault diagnosis monitoring, which is convenient for improving the accuracy of fault diagnosis, and can quickly and accurately identify equipment faults, shortening the time of fault discovery;

[0078] 2. Through the trend analysis step, the system can obtain the voiceprint change characteristics within a period of time and conduct in-depth analysis of the change trend of the voiceprint characteristics; this enables the system to issue early warnings before equipment failures occur, providing staff with sufficient time for troubleshooting; at the same time, through time series analysis of voiceprint information, the system can accurately locate the parts or links where failures may occur, thereby improving the efficiency and accuracy of troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a flow chart of the fault detection method in Example 1 of the present application;

[0080] Figure 2 is a detailed flow chart of the voiceprint analysis step in Example 2 of the present application;

[0081] Figure 3 is a flow chart of the period adjustment step in Example 2 of the present application;

[0082] Figure 4 It is a detailed flowchart of the first risk classification step and the second risk classification step in Example 2 of the present application. DETAILED DESCRIPTION

[0083] The following combination Figures 1 to 4This application is described in further detail.

[0084] This embodiment discloses a device fault detection method based on a vibroacoustic method.

[0085] Example 1: Reference Figure 1 , a device fault detection method based on a vibroacoustic method, comprising the following steps:

[0086] Data collection: Collect equipment vibration signals and obtain the duration of the collection period. The vibration frequency in the monitoring device is converted into a frequency distribution diagram to obtain vibration voiceprint information; the acoustic signal is collected, where the acoustic signal includes environmental sound, the sound of the monitoring device itself and the sound of the detected device, and the voiceprint information of the acoustic signal is obtained, which is recorded as the first voiceprint information, the second voiceprint information and the third voiceprint information respectively;

[0087] Specifically, use professional vibration sensors (such as accelerometers, speedometers or displacement sensors) to install them on key parts of the equipment (such as bearings, gearboxes or mechanical structures); set the acquisition duration cycle , usually an operating cycle of the equipment or a specific time period; the sensor converts the collected vibration signal into an electrical signal and records it through the data acquisition system; the signal processing technology is used to convert the vibration signal into a frequency distribution diagram to show the distribution of different frequency components; the vibration soundprint information is extracted from the frequency distribution diagram as a unique identifier of the vibration characteristics of the equipment; a microphone array or a single microphone is used to collect acoustic signals around the equipment, and the acoustic signals include ambient sound (such as background noise), the monitoring equipment's own sound (such as the sound of motor operation) and the sound of the detected equipment (such as the friction sound of internal parts of the equipment); the voiceprint recognition technology is used to extract the first voiceprint information (ambient sound), the second voiceprint information (the monitoring equipment's own sound) and the third voiceprint information (the sound of the detected equipment) respectively.

[0088] Preprocessing: removing the first voiceprint information and the second voiceprint information contained in the third voiceprint information to obtain the fourth voiceprint information, and then extracting the information matching the fourth voiceprint information and the vibration voiceprint information to obtain the valid voiceprint information emitted by the detected device;

[0089] Specifically, a signal processing algorithm (such as filtering, denoising, etc.) is used to remove the first voiceprint information and the second voiceprint information from the third voiceprint information, and the pure voiceprint information emitted by the detected device is retained and recorded as the fourth voiceprint information; the fourth voiceprint information is matched with the vibration voiceprint information to find out the similar parts between the two. These similar parts are the valid voiceprint information emitted by the detected device, which reflects the actual operating status of the device.

[0090] Grading: Set voiceprint restriction information, determine whether the valid voiceprint information meets the voiceprint restriction information, classify the valid voiceprint information that exceeds the restriction information, and record it as the initial classification. When classifying, the closer it is to the voiceprint restriction information, the lower its grade;

[0091] Specifically, according to the normal operating status and historical data of the equipment, the voiceprint limitation information is set. The voiceprint limitation information includes parameters such as frequency range and amplitude range, which are used to determine whether the valid voiceprint information is normal. The valid voiceprint information is compared with the voiceprint limitation information. If the valid voiceprint information exceeds the voiceprint limitation information, the level is divided; the level is divided according to the degree to which the valid voiceprint information exceeds the voiceprint limitation information. The closer it is to the voiceprint limitation information, the lower the level; otherwise, the higher the level.

[0092] For example, suppose we have a running mechanical device whose normal operating status has been recorded and analyzed through historical data, and voiceprint qualification information is set accordingly. Voiceprint qualification information may include specific frequency ranges and amplitude ranges that represent the voiceprint characteristics generated by the device during normal operation.

[0093] Set voiceprint limitation information: frequency range: for example, set to 50Hz to 100Hz; amplitude range: for example, set to 0.1 to 0.5mm; these ranges are based on the normal operating status and historical data of the device, and are used to determine whether the current voiceprint information is normal.

[0094] We compare the collected valid voiceprint information with the voiceprint qualification information. Assuming that the frequency of the valid voiceprint information is 60Hz and the amplitude is 0.3mm, these values ​​are within the set range, so it can be considered that the device is currently in normal operation.

[0095] However, if the frequency or amplitude of the effective voiceprint information exceeds the set range, such as a frequency of 120 Hz or an amplitude of 0.6 mm, then classification is required.

[0096] If the valid voiceprint information exceeds the limited voiceprint information, we will classify it into levels according to the degree of excess.

[0097] For example, we can set several levels, such as slight abnormality, moderate abnormality and severe abnormality, where slight abnormality corresponds to level 1, moderate abnormality corresponds to level 2, and severe abnormality corresponds to level 3, or adopt a more detailed level division with one decimal place;

[0098] A minor abnormality may mean that the voiceprint information slightly exceeds the set range, but does not cause a significant impact on the operation of the device; a moderate abnormality may mean that the voiceprint information obviously exceeds the set range, and the operation of the device has been affected to a certain extent; a serious abnormality may mean that the voiceprint information seriously exceeds the set range, and the operation of the device has been seriously affected, and may even cause the device to shut down or be damaged.

[0099] In this example, if the frequency of the valid voiceprint information is 120Hz (exceeding the set upper limit of 100Hz), but the amplitude is still within the set range, we can classify it as a slight abnormality level. If both the frequency and amplitude exceed the set range, then it may need to be classified as a higher level.

[0100] First judgment: judging whether the vibration frequency is greater than or equal to the preset vibration limit value, if so, executing the fault determination step, otherwise, executing the vibration intervention step;

[0101] Specifically, the preset vibration limit is set according to the normal operating status and historical data of the equipment. If the vibration frequency is greater than or equal to the preset vibration limit, it is considered that the equipment may have a fault and the fault determination step is executed; otherwise, the vibration intervention step is executed.

[0102] Fault determination: Determine if the equipment is faulty and remind staff to shut down for maintenance;

[0103] Vibration intervention: adjust and intervene the initial classification of valid voiceprint information according to the vibration limit and vibration frequency to obtain the intervention classification; the closer the vibration frequency is to the vibration limit, the greater the increase in the classification;

[0104] Specifically, the vibration limit is preset : The upper limit of vibration frequency set according to equipment characteristics and experience; current vibration frequency : The vibration frequency of the device during current operation; initial classification : The classification is preliminarily determined based on the valid voiceprint information.

[0105] Calculate the adjustment factor: Adjustment factor It can be calculated based on the difference between the vibration frequency and the preset vibration limit. We can use a linear function or other function forms. In this embodiment, a linear function is selected; for example, a linear function can be defined as: ,in is a set threshold value used to control the range of change of the adjustment coefficient, which is close to .when near hour, should be close to 0 (indicating a need for significant adjustment); keep away hour, Should be close to 1 (indicating a smaller adjustment).

[0106] In order to avoid the situation where the denominator is 0, Should be less than And greater than the lowest possible vibration frequency.

[0107] Calculating the Intervention Grade: Intervention Grade It can be calculated by the following formula: ,in is an offset used to further adjust the classification based on the adjustment coefficient. This offset can be set according to actual conditions to ensure that the intervention classification is within a reasonable range; is the minimum value of the adjustment factor (when equal ), among which, is a number greater than 0 but less than 1 to avoid the denominator being 0.

[0108] Output results: Output the calculated intervention classification as the final classification result.

[0109] For example, assume the following parameters: Preset vibration limit :100Hz; Current vibration frequency ( : 95Hz; initial classification : Level 2 (assuming the grading range is 1-5, the higher the level, the greater the risk); set threshold : 90Hz (used to control the range of adjustment coefficient); offset : 0.5 level (used to further adjust the classification based on the adjustment coefficient); minimum value of the adjustment coefficient : 0.1 (when equal hour, will approach 0.1, which is determined based on historical data and experience);

[0110] Calculate the adjustment factor according to the algorithm: ;

[0111] Calculate intervention classification:

[0112] Level (Since the level is usually an integer, it can be rounded up or rounded up according to actual needs, for example, rounded to 3 levels; the level can also be rounded to one decimal place).

[0113] Second judgment: judge whether the intervention classification exceeds the critical level. If so, execute the risk warning step; otherwise, execute the normal detection step;

[0114] Risk warning: setting key monitoring cycles , focus on monitoring the equipment;

[0115] Normal detection: Perform normal periodic monitoring on the equipment.

[0116] Specifically, a critical level is set as a threshold for determining whether a device has potential risks. If the intervention level exceeds the critical level, the device is considered to have potential risks and the risk warning step is executed; otherwise, the normal detection step is executed. The critical level is determined by analyzing the historical operating data and fault records of the device and the level of the voiceprint information before a fault occurs or is about to occur.

[0117] When the intervention classification exceeds the critical level, the risk warning system is triggered and a key monitoring cycle is set (n is a natural number, indicating the number of monitoring times or cycles; i represents a specific monitoring item or indicator), and the equipment is monitored in a focused manner; during the monitoring period, the monitoring frequency can be increased, the monitoring parameters can be adjusted, or other measures can be taken to reduce the risk of equipment failure; when the intervention level does not exceed the critical level, the equipment is monitored in a normal periodic manner, and regular monitoring and analysis are performed according to the set monitoring cycle and monitoring parameters; if any abnormal situation is found or the monitoring data exceeds the normal range, timely measures are taken to deal with it.

[0118] Among them, the key monitoring period in the risk warning step is shorter than the ordinary period of normal detection, and the normal detection period is a fixed value. The key monitoring period in the risk warning step can be calculated and adjusted or directly set with the ordinary period as the initial period value.

[0119] Embodiment 1 of the present application can more accurately analyze the location of the fault and perform fault diagnosis monitoring by combining vibration signals and sound signals, thereby improving the accuracy of fault diagnosis, and can quickly and accurately identify equipment faults, thereby shortening the time for fault discovery.

[0120] Example 2: Reference Figure 2 , the difference between this embodiment and embodiment 1 is that a frequency estimation step is provided between the second judgment step and the normal detection step;

[0121] Frequency prediction: obtain the temperature and determine the temperature rising trend, predict the growth of the vibration frequency through the temperature rising trend and the initial classification, obtain the predicted frequency, and then perform the first judgment step again according to the predicted frequency.

[0122] Specifically, according to the initial classification and temperature data, wherein the temperature data is collected in real time from the temperature sensor of the equipment, a corresponding prediction model is selected, and the prediction model can be a statistical model based on historical data or a machine learning model; wherein the statistical model based on historical data can be adopted as follows: Linear regression: If the growth trend of the vibration frequency of the equipment is linear, then the linear regression model is a good choice; Time series analysis: If the change of the vibration frequency of the equipment has time dependence, that is, the current value is related to the value at the previous time point, then the time series analysis model (such as ARIMA, GARCH, etc.) may be more appropriate; Exponential smoothing: For data with trend and seasonal changes, the exponential smoothing model can provide a smooth prediction value; wherein the machine learning model is as follows: Decision tree: If the growth of the vibration frequency of the equipment is related to multiple feature variables, and there are complex nonlinear relationships between these feature variables, then the decision tree model can capture these relationships and provide accurate predictions; Random forest: Random forest is a collection of decision trees, which improves the accuracy and robustness of the model by integrating the prediction results of multiple decision trees; Neural network: For highly nonlinear and complex data sets, neural networks (such as multilayer perceptron, convolutional neural network, etc.) can provide powerful modeling capabilities.

[0123] In this embodiment, random forest is preferably used as the prediction model, and historical data is used to train the random forest model, and the parameters of the model (such as the number of trees, maximum depth, etc.) are adjusted to optimize the performance. Then, feature information related to the initial classification and temperature rising trend information are input into the prediction model. The model outputs the predicted frequency, that is, the possible value of the vibration frequency of the equipment in the future period of time. Then, the predicted frequency is input and the first judgment step is re-executed. According to the predicted frequency, it is determined whether immediate intervention is needed. Otherwise, the normal detection step is continued.

[0124] Specifically, historical data is collected and organized to form a data set, ensuring that each feature has a clear physical meaning and that the data has been cleaned without missing values ​​or outliers; for example, for vibration frequency data, noise is removed or smoothing is performed; for sound signals, filtering, denoising and feature extraction are performed; and then the features are standardized or normalized to ensure that the dimensions of different features are consistent.

[0125] Use the holdout method or cross-validation method to divide the dataset into training and testing sets. For example, we can use 70% of the data as the training set to train the random forest model and 30% of the data as the testing set to evaluate the performance of the model.

[0126] Use the training set to train the random forest model and adjust the model parameters; during the training process, the random forest will randomly select features and samples to build multiple decision trees. Each decision tree will be trained independently and output its own prediction results.

[0127] The parameters to be adjusted are: n_estimators (the number of decision trees):

[0128] Initial setting: You can start with a moderate value, such as 100 or 200.

[0129] Adjustment strategy: Observe the changes in model performance through cross-validation. If the performance improves significantly after increasing the number of decision trees, continue to increase it. If the performance improvement is not obvious or fluctuates, the optimal or near-optimal number of decision trees may have been reached.

[0130] max_depth (maximum depth of decision tree):

[0131] Initial setting: You can set a relatively large value, such as 10 or 20, to allow the decision tree to deeply capture the data characteristics.

[0132] Adjustment strategy: gradually increase or decrease the maximum depth, observe the performance changes of the model on the training set and validation set, and avoid setting it too deep to cause overfitting, or too shallow to cause underfitting.

[0133] min_samples_split (minimum number of samples required for internal node re-division):

[0134] Initial setting: Set according to the size of the data set, such as 2 or 5.

[0135] Adjustment strategy: By adjusting this parameter, you can control the growth rate of the decision tree. A smaller value will make the decision tree grow faster, but may increase the risk of overfitting; a larger value will make the decision tree grow slower, but may improve the generalization ability of the model.

[0136] min_samples_leaf (minimum number of samples for leaf nodes):

[0137] Initial setting: Usually it can be set to 1 or a smaller value;

[0138] Adjustment strategy: Increasing the value of this parameter can reduce the risk of overfitting, but it may also cause the model to underfit. You need to make a trade-off based on the performance of the model on the training set and the validation set.

[0139] max_features (maximum number of features to consider when splitting):

[0140] Initial setting: Can be set to "auto" (i.e. sqrt(n_features)) or "log2" (i.e. log2(n_features)).

[0141] Tuning strategy: By adjusting this parameter, you can control the number of features that each tree considers when splitting. Smaller values ​​can increase the diversity of the model, but may also cause the model to underfit; larger values ​​may improve the performance of the model, but may also increase the risk of overfitting.

[0142] Training process example: Collect vibration frequency data of equipment at different temperatures and different initial classifications; data format example: temperature (°C), initial classification (1-5 levels), vibration frequency (Hz); clean the data to remove outliers and duplicate data; take the temperature rising trend as one of the features, which can be obtained by calculating the rate of change of continuous temperature data;

[0143] The preprocessed data was divided into a training set and a test set. The random forest model was trained with the training set. The parameters of the model were adjusted, such as the number of trees was 100 and the maximum depth was 10 layers. The performance of the model was evaluated using methods such as cross-validation. The average prediction error of the model was less than 5 Hz, indicating that the model had high accuracy.

[0144] Example of forecasting frequency data:

[0145] Assume that the current temperature of the equipment is 40°C, the temperature rise trend is 0.5°C / hour, and the initial classification is level 4. This information is input into the trained random forest model, and the model outputs the predicted frequency.

[0146] Input features: temperature = 40°C, temperature rising trend = 0.5°C / hour, initial classification = level 4; model prediction: the vibration frequency is predicted to be 90Hz in the next 24 hours.

[0147] Reference Figure 2 In other embodiments, a voiceprint analysis step is further provided between the second judgment step and the risk warning step;

[0148] Voiceprint analysis: including trend analysis steps and warning steps;

[0149] Trend analysis: obtain the voiceprint change characteristics within a period of time, analyze the voiceprint feature change trend, perform time series analysis on the effective voiceprint information, and predict the change trend of the voiceprint features. If the trend is greater than the set growth threshold, execute the warning step, otherwise, execute the risk warning step;

[0150] Warning: Send a signal that the equipment is about to fail, reminding staff to check.

[0151] Specifically, voiceprint features are extracted from the sound signal emitted by the device. These features can be frequency, amplitude, timbre, etc.; time series analysis is performed on the extracted voiceprint features to identify their changing trends. The methods that can be used include ARIMA model, exponential smoothing, machine learning algorithms (such as LSTM), etc.; based on the results of time series analysis, the ARIMA model is used to predict the changing trend of voiceprint features in the future; a growth threshold is set to determine whether the predicted trend has reached the level that requires a warning; when the predicted trend is greater than the set growth threshold, a warning signal is sent to remind staff to check the equipment.

[0152] The sending of warning signals: when the predicted trend exceeds the threshold, the warning system is automatically triggered and a warning signal is sent to relevant personnel. The warning information should include the device name, predicted trend value, current time, recommended troubleshooting measures, etc. The warning information can be sent via SMS, email, system notification, etc. to ensure that relevant personnel can receive it in time.

[0153] Among them, the setting of the growth threshold can be dynamically adjusted according to historical data analysis, business needs and security, expert experience and knowledge. The historical data analysis includes analyzing the historical data of the variable over a period of time in the past, understanding its fluctuation range, growth trend and outliers, and determining a reasonable growth threshold through statistical methods (such as mean, standard deviation, percentile, etc.); Business needs and security: Determine the tolerance for the growth rate of the variable based on the operating characteristics and business needs of the equipment, and set a relatively conservative but not overly sensitive growth threshold taking into account the safety risks and economic losses that may be caused by equipment failure.

[0154] Expert experience and knowledge: Draw on the experience and knowledge of domain experts to understand the abnormal situations that may occur during the operation of similar equipment. Based on the experts' suggestions, adjust the growth threshold to better suit the actual situation of the equipment.

[0155] Reference Figure 3 In other implementations, a period adjustment step is provided between the trend analysis step and the risk warning step;

[0156] Cycle adjustment: Determine the change range based on the change trend, and then attenuate the cycle based on the change range. As the change range increases, focus on monitoring the cycle. Narrow down, and then perform risk warning steps based on the updated cycle.

[0157] Specifically, voiceprint features, such as frequency and amplitude, are extracted from the sound signal emitted by the device, and time series analysis is performed on the extracted voiceprint features to identify their changing trends and predict possible changes in the voiceprint features in the future.

[0158] Calculate the difference between the current voiceprint feature and the previous cycle or a certain benchmark value to obtain the change amplitude;

[0159] Cycle attenuation: The key monitoring cycle is attenuated according to the magnitude of the change. The larger the change, the shorter the key monitoring cycle is, so that the device status can be monitored more frequently.

[0160] Update cycle: Use the updated cycle to perform subsequent risk warning steps.

[0161] The algorithm formula for period adjustment is: ; In the formula, is the initial warning cycle, is the amplitude of the voiceprint feature change in the i-th cycle, is the preset maximum change amplitude threshold, which is used to normalize the change amplitude. is the adjustment factor used to control the sensitivity of period decay. According to the determination of the historical data collected, the probability or risk level of equipment failure under different change amplitudes is calculated by statistical analysis; according to the statistical analysis results, a reasonable α value range is determined so that within this range, the sensitivity of periodic attenuation can accurately reflect the changes in equipment status.

[0162] Then, based on the updated early warning cycle and the changing trend of the voiceprint features, determine whether the device is in a risky state; in each updated key monitoring cycle, monitor the changes in the voiceprint features; if the changing trend of the voiceprint features exceeds the preset threshold or reaches a certain abnormal pattern, a risk warning is triggered and a warning signal is sent to relevant personnel to remind them to conduct equipment inspections or maintenance.

[0163] Reference Figure 3 In other implementations, after the cycle adjustment step and the risk warning step, a cycle determination step is also provided;

[0164] Cycle judgment: judge the key monitoring cycle Is it less than or equal to the minimum value of the key monitoring period? , if yes, then execute the warning step, otherwise, execute the risk warning step.

[0165] Specifically, set a minimum value for the key monitoring period , this value can be preset according to the actual situation of the equipment and the monitoring requirements, and after the period adjustment step, the updated key monitoring period is obtained , then and , for comparison, if , then execute the warning step; otherwise, execute the risk warning step.

[0166] A warning signal is issued when an alarm is triggered, which may include sound alarm, flashing lights, text message or email notification, etc. The warning content should include the equipment name, warning time, possible fault type and recommended emergency treatment measures.

[0167] Reference Figure 4 In other implementations, after the frequency estimation step, a first risk classification step is further provided;

[0168] First risk classification: including risk estimation step and damage level classification step;

[0169] Risk estimation: Set the frequency danger value and determine whether the predicted frequency exceeds the frequency danger value. If so, perform the damage level classification step;

[0170] Damage level classification: Damage levels are classified according to the interval between the frequency danger value and the vibration limit value. Each damage level corresponds to the urgency of the investigation, and then the damage level is determined based on the estimated frequency.

[0171] Specifically, set frequency danger values: according to the type of equipment, operating conditions, historical fault data, etc., set one or more frequency danger values. These values ​​can be fixed thresholds or thresholds that are dynamically adjusted based on the equipment status. Obtain the predicted frequency of the equipment from the frequency estimation step, and compare the predicted frequency with the set frequency danger value.

[0172] If the predicted frequency exceeds the frequency danger value, the damage classification step is performed; otherwise, monitoring can continue or other preventive measures can be taken.

[0173] According to the interval between the frequency danger value and the vibration limit value, a damage level classification standard is formulated. For example, the interval between the frequency danger value and the vibration limit value can be evenly divided into several sub-intervals, each of which corresponds to a damage level, where the closer to the vibration limit value, the higher the damage level; this embodiment preferably has three sub-intervals, corresponding to the levels of increasing , , There are three levels, among which the predicted frequency is compared with the damage level classification standard to determine the current damage level of the equipment. The damage level classification table is shown in Table 1 below:

[0174] Table 1 Damage level classification table

[0175]

[0176] Develop an inspection plan: Develop an inspection plan for the equipment based on the damage level and the corresponding inspection urgency.

[0177] Reference Figure 4In other embodiments, a second risk classification step is provided between the second judgment step and the normal detection step;

[0178] The second risk classification: including the fault prediction step and the fault possible level classification step;

[0179] Fault prediction: Set the risk level and determine whether the intervention level exceeds the risk level. If so, perform the possible fault level classification step.

[0180] Fault possibility level classification: Fault possibility levels are classified according to hazard level and critical level. Each fault possibility level corresponds to the urgency of troubleshooting, and then the fault possibility level is determined according to the intervention level.

[0181] Specifically, set the risk classification: determine the voiceprint information when the fault occurs based on the type of equipment, operating conditions, historical fault data, etc., and set a risk classification based on the voiceprint information; obtain the current intervention classification from the equipment monitoring system, and compare the intervention classification with the set risk classification.

[0182] If the intervention level exceeds the danger level, the fault possibility classification step is performed; otherwise, normal detection can continue or other preventive measures can be taken.

[0183] Based on the relationship between intervention classification, hazard classification and criticality classification, the possible fault levels of the equipment are divided to determine the urgency of troubleshooting.

[0184] According to the relationship between the risk classification and the critical level, the fault possible level classification standard is formulated. For example, the interval between the intervention classification, the risk classification and the critical level can be divided into three sub-intervals, each sub-interval corresponds to a possible fault level, and the present embodiment preferably has three sub-intervals, corresponding to the levels in ascending order. , , Three levels, compare the intervention level with the possible fault level classification standard to determine the current possible fault level of the equipment. For example, the intervention level is 1-2, which can correspond to , 2.1-4 levels, can be used , 4.1-6 levels, can be used , where the possible fault classification table is as follows Table 2:

[0185] Table 2 Fault possible level classification table

[0186]

[0187] Develop a troubleshooting plan: Develop a troubleshooting plan for the equipment based on the possible level of the fault and the corresponding degree of urgency of the troubleshooting. Equipment with a high probability of failure needs to be shut down for inspection. Equipment with a medium failure potential level requires enhanced monitoring and regular inspections. Equipment with a low failure probability level can continue to be monitored and appropriate preventive measures can be taken.

[0188] Reference Figure 4 , in other embodiments, further comprising a fault classification step;

[0189] Fault classification: Determine the estimated possible level of fault based on the damage level and possible level of fault, combined with the risk matrix.

[0190] Specifically, the possible levels of fault prediction are divided into three levels for hierarchical management of risks. The risk matrix is ​​shown in Table 3 below:

[0191] Table 3 Risk Matrix

[0192]

[0193] The risk levels in Table 3 are low (I), medium (II), and high (III);

[0194] The recommended measures for different fault prediction levels are different, and the urgency of the implementation measures is different, see Table 4:

[0195] Table 4 Risk levels and corresponding measures

[0196]

[0197] The damage level is higher than the possible fault level, and its recommended measures are implemented first.

[0198] The embodiment of the present application also discloses a patrol inspection system.

[0199] Inspection system, including the following modules:

[0200] Data acquisition module: used to collect equipment vibration signals and acoustic signals;

[0201] Specifically, sensors (such as accelerometers, microphones, etc.) are used to collect vibration signals and acoustic signals of the equipment. The sensors transmit the data to the data acquisition module via wired or wireless means. The data acquisition module performs preliminary processing on the original signals, such as filtering and amplification, to improve data quality.

[0202] Preprocessing module: The input end is connected to the output end of the data acquisition module, and is used to preprocess the collected information and then extract the effective voiceprint information emitted by the detected device;

[0203] Specifically, the collected vibration signals and acoustic signals are preprocessed, such as denoising, filtering, feature extraction, etc., and signal processing techniques (such as Fourier transform, wavelet transform, etc.) are used to extract effective voiceprint information, and the preprocessed data is stored in a database for use by subsequent modules.

[0204] Classification module: The input end is connected to the pre-processing module and is used to classify the effective voiceprint information;

[0205] The first judgment module: the input end is connected to the output end of the preprocessing module, and is used to judge whether the vibration frequency is greater than or equal to the preset vibration limit;

[0206] Fault determination module: the input end is connected to the output end of the first judgment module, and is used to determine that the equipment has a fault when the vibration frequency is greater than or equal to the preset vibration limit, and remind the staff to shut down the equipment for maintenance;

[0207] Vibration intervention module: the input end is connected to the output end of the first judgment module and the output end of the classification module, and is used to adjust and intervene the initial classification of the effective voiceprint information according to the vibration limit and the vibration frequency when the vibration frequency is less than the preset vibration limit, so as to obtain the intervention classification;

[0208] The second judgment module: the input end is connected to the output end of the vibration intervention module, and is used to judge whether the intervention classification exceeds the critical level;

[0209] Risk warning module: The input end is connected to the output end of the second judgment module, and is used to perform periodic Set and monitor the equipment in key periods;

[0210] Voiceprint analysis module: The input end and the output end of the risk warning module are used to analyze the voiceprint change trend, determine whether the trend is greater than the set growth threshold, and adjust the key monitoring period according to the trend change;

[0211] Frequency estimation module: The input end is connected to the output end of the second judgment module, and the growth of the vibration frequency is predicted through the initial classification and the temperature rising trend to obtain the predicted frequency, and then the first judgment module, the second judgment module and the normal detection module are executed again according to the predicted frequency.

[0212] Normal detection module: the input end is connected to the output end of the second judgment module and the output end of the frequency estimation module, and is used to perform normal periodic monitoring on the equipment when the intervention classification does not exceed the critical level;

[0213] The first risk classification module: the input end is connected to the output end of the frequency estimation module, and is used to classify the damage level and determine the damage level according to the estimated frequency;

[0214] The second risk classification module: the input end is connected to the output end of the second judgment module, and is used to classify the possible level of the fault, and determine the possible level of the fault according to the intervention classification;

[0215] Fault classification module: The input end is connected to the output end of the first risk classification module and the output end of the second risk classification module, and is used to determine the possible level of fault estimation based on the damage level and the possible level of fault in combination with the risk matrix.

[0216] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A device fault detection method based on a vibroacoustic method, characterized in that: The following steps are involved: Data collection: collecting equipment vibration signals, obtaining the vibration frequency within the collection duration period, and converting it into a frequency distribution diagram to obtain vibration voiceprint information; collecting acoustic signals, where the acoustic signals include environmental sounds, the monitoring equipment's own sounds, and the detected equipment's sounds, obtaining the voiceprint information of the acoustic signals, and recording them as the first voiceprint information, the second voiceprint information, and the third voiceprint information respectively; Preprocessing: removing the first voiceprint information and the second voiceprint information contained in the third voiceprint information to obtain the fourth voiceprint information, and then extracting the information matching the fourth voiceprint information and the vibration voiceprint information to obtain the valid voiceprint information emitted by the detected device; Grading: Set voiceprint restriction information, determine whether the valid voiceprint information meets the voiceprint restriction information, classify the valid voiceprint information that exceeds the restriction information, and record it as the initial classification. When classifying, the closer it is to the voiceprint restriction information, the lower its grade; First judgment: judging whether the vibration frequency is greater than or equal to the preset vibration limit value, if so, executing the fault determination step, otherwise, executing the vibration intervention step; Fault determination: Determine if the equipment is faulty and remind staff to shut down for maintenance; Vibration intervention: adjust and intervene the initial classification of valid voiceprint information according to the vibration limit and vibration frequency to obtain the intervention classification; the closer the vibration frequency is to the vibration limit, the greater the increase in the classification; Second judgment: judge whether the intervention classification exceeds the critical level. If so, execute the risk warning step; otherwise, execute the normal detection step; Risk warning: set key monitoring cycles and conduct key monitoring of equipment; Normal detection: Perform normal periodic monitoring on the equipment.

2. The equipment fault detection method based on the vibroacoustic method according to claim 1, characterized in that: A frequency estimation step is also provided between the second judgment step and the normal detection step; Frequency estimation: The growth of the vibration frequency is predicted through initial classification to obtain the predicted frequency, and then the first judgment step is performed again based on the predicted frequency.

3. The equipment fault detection method based on the vibroacoustic method according to claim 2 is characterized in that: The frequency estimation step also includes obtaining the temperature, determining the temperature rising trend, and adjusting the predicted frequency according to the temperature rising trend.

4. The equipment fault detection method based on the vibroacoustic method according to claim 3 is characterized in that: Between the second judgment step and the risk warning step, a voiceprint analysis step is also provided; Voiceprint analysis: including trend analysis steps and warning steps; Trend analysis: obtain the voiceprint change characteristics within a period of time, analyze the voiceprint feature change trend, perform time series analysis on the effective voiceprint information, and predict the change trend of the voiceprint features. If the trend is greater than the set growth threshold, execute the warning step, otherwise, execute the risk warning step; Warning: Send a signal that the equipment is about to fail, reminding staff to check.

5. The equipment fault detection method based on the vibroacoustic method according to claim 4 is characterized in that: Between the trend analysis step and the risk warning step, there is also a period adjustment step; Cycle adjustment: Determine the magnitude of change based on the trend of change, and then attenuate the key monitoring cycle based on the magnitude of change. As the magnitude of change increases, the key monitoring cycle is shortened, and then the risk warning steps are executed based on the updated key monitoring cycle.

6. The equipment fault detection method based on the vibroacoustic method according to claim 5 is characterized in that: Between the cycle adjustment step and the risk warning step, there is also a cycle judgment step; Cycle judgment: Determine whether the key monitoring cycle is less than or equal to the minimum value of the key monitoring cycle. If so, execute the warning step; otherwise, execute the risk warning step.

7. The equipment fault detection method based on the vibroacoustic method according to claim 2 is characterized in that: After the frequency estimation step, a first risk classification step is also provided; First risk classification: including risk estimation step and damage level classification step; Risk estimation: Set the frequency danger value and determine whether the predicted frequency exceeds the frequency danger value. If so, perform the damage level classification step; Damage level classification: Damage levels are classified according to the interval between the frequency danger value and the vibration limit value. Each damage level corresponds to the urgency of the investigation, and then the damage level is determined based on the estimated frequency.

8. The equipment fault detection method based on the vibroacoustic method according to claim 7 is characterized in that: After the vibration intervention step, a second risk classification step is also provided; The second risk classification: including the fault estimation step and the fault possible level classification step; Fault prediction: Set the risk level and determine whether the intervention level exceeds the risk level. If so, perform the possible fault level classification step. Fault possibility level classification: Fault possibility levels are classified according to hazard level and critical level. Each fault possibility level corresponds to the urgency of troubleshooting, and then the fault possibility level is determined according to the intervention level.

9. The equipment fault detection method based on the vibroacoustic method according to claim 8, characterized in that: It also includes a fault rating step; Fault classification: Determine the estimated possible level of fault based on the damage level and possible level of fault, combined with the risk matrix.

10. An inspection system for implementing the equipment fault detection method based on the vibroacoustic method according to any one of claims 1 to 9, characterized in that: Includes the following modules: Data acquisition module: used to collect equipment vibration signals and acoustic signals; Preprocessing module: The input end is connected to the output end of the data acquisition module, and is used to preprocess the collected information and then extract the effective voiceprint information emitted by the detected device; Classification module: The input end is connected to the pre-processing module and is used to classify the effective voiceprint information; The first judgment module: the input end is connected to the output end of the preprocessing module, and is used to judge whether the vibration frequency is greater than or equal to the preset vibration limit; Fault determination module: the input end is connected to the output end of the first judgment module, and is used to determine that the equipment has a fault when the vibration frequency is greater than or equal to the preset vibration limit, and remind the staff to shut down the equipment for maintenance; Vibration intervention module: the input end is connected to the output end of the first judgment module and the output end of the classification module, and is used to adjust and intervene the initial classification of the effective voiceprint information according to the vibration limit and the vibration frequency when the vibration frequency is less than the preset vibration limit, so as to obtain the intervention classification; The second judgment module: the input end is connected to the output end of the vibration intervention module, and is used to judge whether the intervention classification exceeds the critical level; Risk warning module: the input end is connected to the output end of the second judgment module, and is used to set a key monitoring cycle after the intervention classification exceeds the critical level, and to perform key monitoring on the equipment according to the key monitoring cycle; Normal detection module: The input end is connected to the output end of the second judgment module, and is used to perform normal ordinary periodic monitoring on the equipment when the intervention classification does not exceed the critical level.

Images