Contact type rotating machinery abnormal sound alarm device and alarm method

Abstract

The invention discloses a contact-type rotating machinery abnormal sound alarm device and alarm method. The contact-type rotating machinery abnormal sound alarm device comprises a contact-type sound acquisition unit, a signal processing unit, a feature extraction and analysis unit and an alarm decision and execution unit. A high-sensitivity piezoelectric microphone array is in direct contact with key parts (such as a bearing seat and a machine shell) of a rotating machine, internal operation sound signals are collected in real time, time domain, frequency domain and time-frequency domain joint features are extracted after amplification and adaptive filtering processing, and abnormal sound recognition is carried out in combination with an intelligent diagnosis model based on a convolutional neural network (CNN). And if the judgment result is abnormal, alarm information is pushed to a site and a remote terminal through an audible and visual alarm and a wireless communication module (5G / Wi-Fi). According to the invention, the problems of large environmental noise interference and weak weak signal capturing capability in the prior art are solved, high-precision and low-delay monitoring of the abnormal sound of the rotating machinery is realized, and the operation safety of equipment is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of rotating machinery fault detection, specifically a contact-based real-time monitoring device for abnormal sounds of rotating machinery and a corresponding alarm method, which is particularly suitable for real-time abnormal sound monitoring and early warning of rotating equipment such as motors, fans, and compressors in industrial scenarios. Background Art

[0002] In many production scenarios such as machinery, automobiles, aerospace, ships, high-speed rails, chemical industries, oil and natural gas, water conservancy and hydropower, construction, mines, and terminal freight yards, rotating machinery (such as various motors, fans, compressors, internal combustion engines, conveyors, etc.) plays a key role, and its normal operation is crucial for the continuity and safety of the entire production process. However, with the increase in operating time and the change of operating conditions, problems such as poor lubrication, excessive wear, looseness, and fatigue may occur in the internal components of rotating machinery, and these problems will first cause abnormal sounds. Timely detection and handling of these abnormal sounds are of great significance for preventing serious equipment failures, reducing maintenance costs, and avoiding safety accidents. Traditional monitoring methods have the following limitations:

[0003] Manual inspection method: relying on experience for judgment, with poor real-time performance and unable to cover continuous operation scenarios;

[0004] Vibration analysis method: vulnerable to mechanical structure resonance interference and insufficient sensitivity to early slight abnormal sounds;

[0005] Non-contact acoustic monitoring: vulnerable to environmental noise (such as workshop background noise and airflow noise), resulting in an increased false alarm rate.

[0006] The existing patent CN201010263896 (diagnostic system for non-contact rotating machinery faults) uses a non-contact microphone but does not optimize for the characteristics of internal sound sources of rotating machinery. Therefore, there is an urgent need for an abnormal sound alarm technology that is directly contact-based, has strong anti-interference ability, and supports intelligent diagnosis. Summary of the Invention

[0007] The present invention aims to provide a contact-based abnormal sound alarm device and alarm method for rotating machinery. The alarm method includes: contact-based sound acquisition, using a piezoelectric microphone array in close contact with mechanical components to suppress environmental noise and improve the signal-to-noise ratio; adaptive filtering technology: dynamically adjusting the cut-off frequency to match the sound spectrum characteristics at different speeds; deep learning diagnostic model: a time-frequency feature fusion algorithm based on CNN to improve the accuracy of abnormal sound recognition; multi-level alarm mechanism: combining local acoustic-optical alarm and remote information push to ensure timely response to faults.

[0008] The structure of the alarm device includes:

[0009] 1. Contact-based sound acquisition unit:

[0010] An array is composed of 4 - 8 high - sensitivity piezoelectric microphones (frequency response range 20Hz - 20kHz, sensitivity ≥50mV / Pa), encapsulated in a high - temperature - resistant silicone sleeve, and fixed to key parts such as the bearing housing and gearbox shell of the rotating machinery through a magnetic base or bolts.

[0011] Microphone layout strategy: According to the symmetry of the mechanical structure, adopt circular or linear distribution to cover the areas with high incidence of abnormal sounds.

[0012] 2. Signal processing unit:

[0013] Amplification circuit: Use an AD620 instrumentation amplifier with adjustable gain (10 - 1000 times) and input noise <1μV.

[0014] Adaptive filter: Based on the LMS algorithm, update the filter coefficients in real - time to filter out non - mechanical sound source interference (such as electromagnetic noise and air flow sound).

[0015] 3. Feature extraction and analysis unit:

[0016] Time - domain analysis: Calculate the signal peak value, root - mean - square (RMS), and kurtosis index;

[0017] Frequency - domain analysis: Use FFT transform to extract the main frequency and the proportion of harmonic energy;

[0018] Joint time - frequency domain analysis: Use wavelet transform (Daubechies 4 basis function) to extract energy entropy features.

[0019] Intelligent diagnosis model: A CNN model trained based on the PyTorch framework, with an input of a 128×128 time - frequency diagram and an output of the normal / abnormal classification probability (threshold set to 0.85).

[0020] 4. Alarm decision - making and execution unit:

[0021] Local alarm: Use an STC89C52 single - chip microcomputer to control an audible and visual alarm (LED strobing frequency 2Hz, buzzer volume ≥90dB);

[0022] Remote communication: Integrate the ESP32 module, support the MQTT protocol, push the alarm information to the cloud platform, and synchronously send it to the maintenance personnel's mobile APP.

[0023] The contact - type abnormal sound alarm method for rotating machinery of the present invention is specifically implemented according to the following steps:

[0024] Contact - type sound signal acquisition step: Through the contact - type sound acquisition unit, closely contact with the key parts of the rotating machinery, and the microphone array continuously acquires sound signals at a sampling rate of 10kHz.

[0025] Signal conditioning step: Use the signal conditioning module to perform conditioning operations such as amplification and filtering on the collected original sound vibration signal to obtain a high-quality and distinctively featured sound signal.

[0026] Feature extraction and analysis step: Use the feature extraction and analysis module to extract features from the conditioned sound signal, generate a time-frequency diagram of the extracted feature data, and input it into the CNN model for analysis to determine whether there are abnormal sound features.

[0027] Alarm trigger step: If the abnormal probability ≥ 0.85, the corresponding alarm mechanism is triggered by the alarm decision and execution module, and an alarm is issued through methods such as audible and visual alarms and pushing information to the remote terminal.

[0028] Beneficial effects

[0029] Compared with the prior art, the present invention has the following remarkable beneficial effects:

[0030] High precision: By means of the contact-type sound acquisition method, the sound signal inside the rotating machinery is directly obtained, effectively avoiding external environmental noise interference, and being able to accurately focus on the key internal parts. Combining with the intelligent diagnosis model to analyze rich signal features, the false alarm and missed alarm phenomena are greatly reduced.

[0031] Strong pertinence: The sound acquisition unit directly contacts the key structural body of the rotating machinery for monitoring, so that the collected sound signal can better reflect the actual internal operating state. For rotating machinery of different types and under different working conditions, this device can accurately capture abnormal sounds caused by internal faults, thus significantly improving the reliability of fault diagnosis.

[0032] Timeliness advantage: The entire system monitors in real time and online, and the process from sound signal acquisition to alarm triggering responds quickly. Once an abnormal sound appears, the alarm can be completed in a very short time, enabling the operation and maintenance personnel to intervene and handle in a timely manner, effectively avoiding the further deterioration of equipment failures and ensuring the continuity of production.

[0033] Wide application range: The device and method of the present invention can be applied to rotating machinery of various different powers, sizes, and application fields. Whether it is the key rotating equipment and internal combustion engines in large-scale industrial production or small-scale civilian rotating machinery, it can play a good abnormal sound alarm function and has good versatility. Description of the drawings

[0034] Figure 1 is the structural schematic diagram of the present invention;

[0035] Figure 2 is the flow chart of the present invention;

[0036] The marks in the figure are:

[0037] In the figure: 1 - motor, 2 - rotating machinery, 3 - signal processing unit, 4 - feature extraction and analysis unit, 5 - alarm decision-making and execution unit, 6 - contact-type sound acquisition unit. Specific implementation manner

[0038] The following further elaborates on the technical solution of the present invention in conjunction with the accompanying drawings to help those skilled in the art have a more complete, accurate, and in-depth understanding of the concept and technical solution of the present invention.

[0039] As Figure 1 is a schematic structural diagram of an embodiment of the present invention. 1 is a motor, and there are rotating bearings at both ends of the motor. The high-sensitivity piezoelectric microphone in the contact-type sound acquisition unit 6 is fixed to key parts such as the bearing housing and gearbox housing of the rotating machinery through a magnetic base or bolts, and is arranged in a circular or linear distribution to cover the areas where abnormal sounds are likely to occur. 2 is the rotating machinery, and the contact-type sound acquisition unit 6 is fixedly installed at the stator of the rotating part of the rotating machinery, and the sound acquisition unit 6 is connected by wires to transmit the signal of the sound acquisition unit to the signal processing unit 3. The signal processing unit 3 is responsible for conditioning the collected original sound vibration signal. It has the following functions:

[0040] Amplification function: The AD620 instrumentation amplifier is used to amplify the weak sound vibration signal according to a preset gain multiple (10 - 1000 times) to make it reach the appropriate amplitude range required by the subsequent signal processing module, ensuring that the signal does not lose key information due to too low an amplitude.

[0041] Filtering function: According to the spectral characteristics and frequency distribution of the sound signal during the normal operation of the rotating machinery, the filtering parameters are adjusted in real time and dynamically based on the LMS algorithm to filter out irrelevant high-frequency clutter and low-frequency interference, extract the effective sound frequency components, and further improve the signal quality.

[0042] The sound signal processed by the signal processing unit 3 is input into the feature extraction and analysis unit 4. The extracted features cover the time-domain features (such as peak value, effective value, pulse width, etc.), frequency-domain features (such as main frequency, frequencies of each harmonic, spectral distribution, etc.) of the sound signal, as well as the joint time-frequency domain features. Then, the extracted feature data is input into a pre-constructed and trained CNN intelligent diagnosis model. This model is based on a deep learning algorithm and accurately determines whether there are abnormal sound features in the current sound signal through training and learning with normal and abnormal sound sample data.

[0043] When the feature extraction and analysis module determines that the probability of abnormal sound in the sound signal is ≥ 0.85, the alarm decision-making and execution unit 5 immediately starts working. It can give prompts through various alarm methods. For example, audible and visual alarms are set at the site of the rotating machinery and in the central control room, emitting eye-catching flashing lights and high-decibel sound alarms to attract the attention of on-site staff. At the same time, with the help of a wireless communication module (such as Wi-Fi, Bluetooth, 5G, etc.), the alarm information is pushed to the remote operation and maintenance management system and the mobile terminal devices of relevant operation and maintenance personnel in real time to ensure that relevant personnel can obtain the abnormal situation of the equipment in the first time and take corresponding countermeasures.

[0044] Such as Figure 2 For example, "collecting the sound signal of the rotating machinery" is the front-end device of the signal processing system, usually composed of a high-sensitivity microphone array. These microphones are distributed at different positions to collect the sound signal in the rotating machinery in real time; "amplifying, filtering and conditioning the signal" means that the collected original sound signal usually contains noise and interference components. In order to improve the quality and processability of the signal, the signal processing unit will first preprocess the signal; "conditioning and diagnosing the signal features" means extracting the features of the processed signal and transmitting them to the data analysis unit, and the data analysis unit will analyze and process these features according to the specific functions of the system; "alarm decision-making and execution" means that if the data analysis unit detects an abnormal situation or meets the preset alarm conditions, the alarm unit will be triggered, and the alarm unit will take corresponding alarm measures according to the functions and settings of the system.

[0045] The present invention has been described exemplarily above in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above-mentioned manner. As long as various non-substantive improvements are made by adopting the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A contact type rotating machinery alarm device, characterized in that, It consists of a contact-type sound acquisition unit, a signal processing unit, a feature extraction and analysis unit, and an alarm decision-making and execution unit; The contact-type sound acquisition unit consists of a high-sensitivity piezoelectric microphone encapsulated in a high-temperature-resistant silica gel sleeve and fixed to the bearing housing or gearbox housing of the rotating machinery through a magnetic base or bolts; the microphone layout strategy adopts a circular or linear distribution according to the symmetry of the mechanical structure to cover the areas with high incidence of abnormal sounds; The amplifier circuit of the signal processing unit uses an instrumentation amplifier with an adjustable gain of 10 - 1000 times and an input noise <1 μV; The adaptive filter updates the filter coefficients in real time based on the LMS algorithm to filter out the interference of non-mechanical sound sources; The feature extraction and analysis unit includes time-domain analysis: calculating the signal peak value, root mean square (RMS), and kurtosis index; frequency-domain analysis: extracting the main frequency and the proportion of harmonic energy through FFT transformation; Joint time-frequency domain analysis: extracting the energy entropy feature using the Daubechies 4 basis function of wavelet transform, and intelligent diagnosis model: a CNN model trained based on the PyTorch framework; The alarm decision-making and execution unit includes local alarm, using a single-chip microcomputer to control the sound and light alarm; remote communication, using an integrated module, supporting the MQTT protocol, pushing the alarm information to the cloud platform and synchronously sending it to the mobile APP of the operation and maintenance personnel.

2. The alarm device according to claim 1, characterized in that, There are 8 high-sensitivity piezoelectric microphones with a frequency response range of 20 Hz - 20 kHz and a sensitivity ≥50 mV / Pa, forming an array.

3. The alarm device according to claim 1, wherein The instrumentation amplifier uses AD620.

4. The alarm device according to claim 1, wherein The input of the CNN model is a 128×128 time-frequency diagram, and the output is the classification probability of normal / abnormal, with the threshold set at 0.

85.

5. The alarm device according to claim 1, characterized in that The local alarm uses an STC89C52 single-chip microcomputer to control the sound and light alarm, with the LED strobe frequency of 2 Hz and the buzzer volume ≥90 dB; the remote communication uses an integrated ESP32 module, supporting the MQTT protocol.

6. The alarm method of the alarm device according to claim 1, characterized in that, The steps are as follows: (1) Contact-type sound signal acquisition: Closely contact with the key parts of the rotating machinery through the contact-type sound acquisition unit, and the microphone array continuously acquires sound signals at a sampling rate of 10 kHz; (2) Signal conditioning: Use the signal conditioning module to perform amplification and filtering conditioning operations on the collected original sound vibration signals to obtain high-quality and distinct-feature sound signals; (3) Feature extraction and analysis: Use the feature extraction and analysis module to extract features from the conditioned sound signals, generate a time-frequency diagram from the extracted feature data, and input it into the CNN model for analysis to determine whether there are abnormal sound features; (4) Alarm triggering step: If the abnormal probability ≥0.85, the corresponding alarm mechanism is triggered by the alarm decision-making and execution module, and an alarm is issued by means of sound and light alarm and pushing information to the remote terminal.

Images