A variable working condition rotating machinery fault diagnosis method and embedded device based on instantaneous speed self-adaption

Abstract

The application discloses a variable working condition rotating machinery fault diagnosis method and embedded device based on instantaneous rotating speed self-adaption, which comprises synchronous signal acquisition, instantaneous rotating speed self-adaption extraction, self-adaption analysis parameter dynamic calculation, angular domain resampling and feature extraction of vibration signals and fault diagnosis based on a light intelligent model, and rotating speed information is directly obtained from an existing electric control system or a simple sensor, instantaneous rotating speed is robustly estimated under strong noise through an advanced signal processing algorithm, an expensive encoder is saved, and deployment cost and threshold are reduced. Without relying on additional rotating speed measurement hardware, a self-adaption pulse processing mechanism guarantees system stability under imperfect rotating speed signals.

Description

Technical Field

[0001] This invention relates to the field of mechanical fault diagnosis technology, specifically a fault diagnosis method and embedded device for rotating machinery under varying operating conditions based on instantaneous speed adaptation. Background Technology

[0002] Rotating machinery, such as motors, fans, gearboxes, and pumps, are key equipment in industrial production. Vibration signal analysis is currently the mainstream fault diagnosis method. Under stable speeds, the fault characteristic frequency has a fixed multiple relationship with the rotational frequency, making it easy to identify. However, in actual industrial scenarios, equipment often operates under variable conditions, with fluctuations or gradual changes in rotational speed. This causes changes in the frequency components of the vibration signal, resulting in "frequency ambiguity" in the conventional equal-time-interval FFT spectrum. The fault characteristic peaks are broadened or even submerged, making it difficult to accurately extract key indicators such as rotational frequency and bearing throughput frequency. Order analysis methods can resample non-stationary time-domain signals to the angular domain, highlighting the periodic components related to rotational speed as constant-order spectral lines, which is an effective means of solving variable-speed analysis.

[0003] Traditional order analysis relies on external key phase sensors to obtain instantaneous rotational speed, increasing hardware costs and installation complexity. Under drastic operating conditions, strong noise, or signal interference, pulse loss or distortion will directly lead to angular domain resampling failure, causing the diagnostic process to collapse. It also lacks an adaptive adjustment mechanism when the rotational speed signal deteriorates. Secondly, high-precision order analysis and complex intelligent diagnostic models require enormous computational resources, making real-time processing difficult on resource-constrained embedded edge devices. Data often needs to be uploaded to the cloud or industrial control computer, resulting in high latency and strong network dependence. Furthermore, parameters such as analysis bandwidth, resampling rate, and feature extraction range of the diagnostic system are usually fixed values. When the operating speed range of the equipment differs significantly from the design conditions, fixed analysis parameters cannot adaptively capture changing fault characteristics, leading to missed or false diagnoses. Therefore, this invention provides a method and embedded device for fault diagnosis of rotating machinery under varying operating conditions based on instantaneous speed adaptation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a fault diagnosis method and embedded device for rotating machinery under varying operating conditions based on instantaneous speed adaptation, in order to solve the aforementioned problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a fault diagnosis method for rotating machinery under varying operating conditions based on instantaneous speed adaptation, comprising the following steps: S1. Synchronously acquire vibration signals from rotating machinery. and the raw speed signal originating from the equipment's electronic control system. ; S2, regarding the original rotation speed signal The process involves adaptively extracting the instantaneous rotational speed sequence synchronized with the vibration signal. ; S3, based on the instantaneous speed sequence Dynamically calculate the analysis parameters of the vibration signal at the current moment, including the range of analysis orders. Equal angle resampling interval ,in S4, based on the instantaneous rotational speed sequence and dynamically calculated resampling interval The vibration signal is resampled from the time domain to the angular domain to obtain the angular domain signal. and to Perform order analysis to extract the order feature vector F, which is independent of rotational speed; S5. Input the order feature vector F into the pre-trained lightweight intelligent diagnostic model to obtain the fault classification results and equipment health status.

[0006] Preferably, the specific process for the instantaneous rotational speed adaptive extraction in step S2 is as follows: S2.1, targeting The pulse signal in the calculation uses an adaptive threshold to detect the pulse leading edge and calculate the pulse interval. And convert it to instantaneous speed When an abnormal pulse interval is detected, pulse loss interpolation compensation or interference pulse elimination logic is initiated. At the same time, periodic harmonic pulses are identified and suppressed, the autocorrelation function of the pulse interval sequence is calculated, and by finding the peak value of the autocorrelation sequence on the non-zero hysteresis, the true fundamental frequency period can be inferred and verified. The correct pulse interval sequence is obtained by eliminating harmonic interference pulses, thus completing the initial correction of the abnormal signal. S2.2, targeting The three-phase motor current signal is then subjected to Clarke-Park transform to extract the rotational frequency component, thereby estimating the instantaneous speed. ; S2.3. Smooth the estimated instantaneous speed sequence to obtain a continuous and smooth instantaneous speed curve. and instantaneous frequency switching .

[0007] Preferably, the specific method for dynamically calculating the analysis parameters in step S3 is as follows: Let the current instantaneous frequency be... The order of the largest characteristic of the fault of interest is Then the equal angle resampling interval Based on Shannon's sampling theorem, it is adaptively determined as follows: ,in The oversampling coefficients are used; the upper limit of the analysis frequency is adaptively set to... .

[0008] Preferably, step S4 specifically includes: S4.1, For instantaneous frequency conversion By integrating, we can obtain the relationship between the rotation angle and time. ; S4.2, as determined in step S3 Generate equiangular sequences for intervals ,according to With time The mapping relationship is found by interpolation algorithm for each... corresponding time The vibration amplitude at that moment is obtained by interpolation, forming an angular domain equally spaced sampling signal. ; S4.3, Diagonal domain signal Spectral analysis was performed to obtain the order spectrum. And extract the amplitude, harmonic energy, sideband modulation energy or spectral entropy at a predetermined order from the order spectrum as the feature vector F.

[0009] Preferably, the lightweight intelligent diagnostic model in step S5 is a gradient boosting decision tree model; after the model is trained in the cloud or on a host computer, it is embedded into the edge computing device with fixed parameters for operation.

[0010] An embedded device for the above method includes a sensor interface module, a signal conditioning and synchronous acquisition module, a main control and computing module, and an interaction and communication module. The sensor interface module is used to connect the vibration acceleration sensor and the sensor or signal line that provides the raw rotation speed signal; The signal conditioning and synchronous acquisition module includes circuits for conditioning the vibration signal and the rotation speed signal respectively, and a synchronous analog-to-digital converter for synchronously converting the two analog signals into digital signals. The main control and computing module employs a field-programmable gate array (FPGA). and microprocessor unit (MCU) The heterogeneous architecture includes an FPGA configured to perform instantaneous rotational speed estimation and angular domain resampling interpolation calculation in real time, and a microprocessor unit configured to perform dynamic calculation of analysis parameters, order analysis, feature extraction, and intelligent diagnostic model inference. The interaction and communication module is used to provide a human-computer interaction interface and external data communication.

[0011] Preferably, in the signal conditioning and synchronous acquisition module, the vibration signal conditioning circuit provides DC bias and drive for the MEMS accelerometer and includes a voltage buffer, an anti-aliasing filter, and a programmable gain amplifier; the rotational speed signal conditioning circuit includes a Schmitt trigger for shaping the pulse signal and an isolation amplifier for isolating the analog signal; the synchronous analog-to-digital converter synchronously samples the two signals at a sampling rate of not less than 20 kHz.

[0012] Preferably, the field-programmable gate array (FPGA) communicates with the microprocessor unit via a high-speed parallel bus or a serial peripheral interface; the FPGA outputs equiangular vibration data blocks and corresponding average rotational speed values ​​to the buffer of the microprocessor unit in real time using a stream processing method.

[0013] Preferably, the interaction and communication module includes a local display screen, status indicator lights, an alarm buzzer, buttons, and at least one industrial communication interface, which includes RS-485, Ethernet, or wireless cellular network interface.

[0014] Preferably, the microprocessor unit also runs a self-learning and optimization module, which periodically collects statistics on the device's operating characteristics to fine-tune diagnostic thresholds or securely receives and updates local diagnostic model parameters from the cloud.

[0015] Beneficial effects Compared with the prior art, the present invention has the following advantages: 1. This invention directly acquires rotational speed information from existing electronic control systems or simple sensors, and robustly estimates instantaneous rotational speed under strong noise conditions through advanced signal processing algorithms, eliminating the need for expensive encoders and reducing deployment costs and barriers to entry. It does not rely on additional rotational speed measurement hardware, and the adaptive pulse processing mechanism ensures system stability even under imperfect rotational speed signals.

[0016] 2. This invention dynamically drives all subsequent analysis parameters through instantaneous rotational speed. This enables the diagnostic system to automatically adapt to all operating conditions from low speed to high speed, always focusing the analysis on the fault characteristic frequency band related to the current rotational speed, thus solving the problem of fixed-parameter systems failing under varying operating conditions.

[0017] 3. This invention employs a hardware-software co-design, where a Field-Programmable Gate Array (FPGA) processes the underlying signals and a Microprocessor Unit (MCU) runs the intelligent algorithm. This offloads computationally intensive preprocessing to the FPGA, ensuring real-time data preprocessing. The lightweight diagnostic model running on the microprocessor unit maintains high accuracy while minimizing computational and storage overhead, enabling the complete intelligent diagnostic process to run in real-time on resource-constrained embedded terminals with faster response times, achieving true edge-side intelligence.

[0018] 4. The device of this invention is highly integrated and easy to install. It only requires connection of a vibration sensor and 1-2 signal lines, provides local display and alarm, and supports remote communication. It is very suitable for digital transformation and predictive health management of a large number of in-service equipment. Attached Figure Description

[0019] Figure 1 This is a flowchart of the process of the method of the present invention; Figure 2 This is a block diagram of the embedded device of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Please see Figure 1-2 A fault diagnosis method for rotating machinery under varying operating conditions based on instantaneous speed adaptation includes the following steps: S1. Synchronous signal acquisition: Synchronously acquire the vibration acceleration signal of the rotating machinery bearing housing. And one of the following, the original speed signal, which contains noise and fluctuations, and the three-phase current signal, obtained from the motor drive system, are used as the original speed signal. .

[0022] The pulse signal is taken from the encoder frequency division output, zero speed pulse (Z pulse) or ready signal of the motor driver. For older motors without frequency division output, a low-cost Hall / photoelectric proximity switch can be directly installed on the shaft. It does not require a high-precision encoder to meet the instantaneous speed calculation requirements, which greatly reduces the cost of adding additional hardware and the complexity of installation.

[0023] Three-phase current signal = obtained from any phase of the motor power line through a current transformer (CT) or Hall current sensor, using a non-invasive ground clamp.

[0024] S2, Instantaneous speed adaptive extraction: S2.1, For the original speed signal Perform filtering and noise reduction preprocessing.

[0025] S2.2, targeting The pulse signal in the calculation employs an adaptive threshold zero-crossing detection combined with a pulse interval calibration algorithm to determine the instantaneous rotational speed. Specifically, the pulse interval is first dynamically estimated. Instantaneous speed When a pulse is lost, interpolation prediction is performed using the trend of the preceding rotational speed change; when interference pulses appear, they are removed through interval rationality checks to suppress periodic frequency-doubled pulses, as detailed below: Calculate the autocorrelation function of the pulse interval sequence. The true rotational frequency corresponds to the longest stable period, and the harmonics will produce high correlation peaks at 1 / 2 and 1 / 3 of this period. By finding the peak value of the autocorrelation sequence on the non-zero hysteresis, the true fundamental frequency period can be inferred and verified. Utilize physical constraints—the rate of change of rotational speed (acceleration) of large rotating machinery is finite. Within a short time, the calculated instantaneous rotational speed should not undergo step jumps on the order of multiples. Set a reasonable threshold for the rate of change of rotational speed, and backtrack to check the rotational speed points that occur at integer multiple jumps. When both pulse and current signals are present, the rotational speed estimated by the current can be used as an auxiliary reference to verify the rationality of the rotational speed calculated by the pulse, and to identify and eliminate persistent harmonic errors.

[0026] S2.3, targeting The three-phase current signal is subjected to Clarke-Park transform to obtain the direct and quadrature axis currents. Then, the rotational frequency component of the current signal is extracted by sliding window Hilbert transform to estimate the instantaneous rotational speed. .

[0027] S2.4. Use the Savitzky-Golay filter to smooth the estimated instantaneous speed sequence to obtain a smooth and continuous instantaneous speed curve. (Unit: revolutions per minute) and its corresponding instantaneous frequency (Unit: Hz).

[0028] S3. Adaptive analysis parameter dynamic calculation: based on the real-time instantaneous frequency... Dynamically calculate the analysis parameters at the current moment: Analysis order range : Set according to the target fault type. Specifically, bearing faults usually focus on For frequency doubling, gear faults focus on the order near the meshing frequency. The corresponding instantaneous analysis bandwidth is calculated as follows: .

[0029] Resampling angle increment Based on the Nyquist sampling theorem and the maximum order of interest Adaptive determination. (Radians). Therefore, the equal-angle resampling interval is determined.

[0030] Envelope analysis carrier frequency adaptation: If envelope analysis is performed to detect impulse faults, the center frequency of its bandpass filter is... Based on instantaneous frequency and fault characteristic order Adaptive adjustment, i.e. .

[0031] S4. Angular domain resampling and feature extraction of vibration signals: S4.1, Angle-Time Relationship Construction: For Instantaneous Frequency Integrate to obtain the instantaneous phase angle. .

[0032] S4.2, Isoangular resampling: Based on the method determined in step S3 As intervals, generate equally spaced angle sequences in the angle domain. .according to and The correspondence is established using a cubic spline interpolation algorithm in the vibration signal. The amplitude at the corresponding moment is found and interpolated to obtain the angular domain vibration signal. Only after completing the instantaneous speed sequence Extraction and smoothing, and based on Dynamically calculate the required resampling angle interval. Only after that can the vibration signal be processed. Equal-angle resampling operation.

[0033] S4.3, Order Feature Extraction: Diagonal Domain Signal Perform a discrete Fourier transform to obtain the order spectrum. The amplitude, harmonic family energy, sideband modulation energy, and spectral entropy of a pre-defined order of interest are extracted from the order spectrum as feature vectors. .

[0034] S5. Fault diagnosis based on lightweight intelligent models: S5.1 Input the extracted feature vector F into a pre-trained Lightweight Gradient Boosting Machine (LightGBM) model.

[0035] S5.2 Probability distribution of fault categories output by the model These include: normal operation, rotor imbalance, misalignment, bearing inner ring failure, bearing outer ring failure, and broken gear teeth.

[0036] S5.3, Based on the maximum probability Provide a diagnostic conclusion and calculate the current health index.

[0037] An embedded device for implementing the above method includes a sensor interface module, a signal conditioning and acquisition module, a main control and computing module, and an interaction and communication module; the device, through embedded design, independently completes the entire process from signal acquisition to fault diagnosis at the device edge.

[0038] Sensor interface module: Includes a vibration acceleration sensor interface and a rotation speed signal interface. The vibration interface is IEPE type and is used to connect to a piezoelectric accelerometer. The rotation speed signal interface includes a digital input channel for pulse signals and an analog input channel for current signals, compatible with digital pulse inputs from photoelectric switches and Hall sensors and analog voltage / current inputs from current transformers.

[0039] Signal conditioning and acquisition module: Contains two parallel conditioning circuits. The vibration signal conditioning circuit includes a constant current source, an anti-aliasing filter, and a programmable gain amplifier. The speed signal conditioning circuit includes a Schmitt trigger for pulse shaping and an isolation amplifier for analog signals. Both signals are acquired by a high-performance synchronous sampling analog-to-digital converter (ADC) to ensure strict time base synchronization.

[0040] Main control and computing module: This is the core module, which adopts a heterogeneous architecture of "microprocessor unit (MCU) + field programmable gate array (FPGA)".

[0041] The Field Programmable Gate Array (FPGA) is responsible for implementing high-speed, deterministic low-level signal processing tasks, including the instantaneous rotational speed extraction algorithm in step S2 and the equal-angle resampling interpolation calculation in step S4.2. Its parallel computing characteristics ensure the real-time performance of rotational speed estimation and resampling.

[0042] The microprocessor unit (MCU) runs an embedded real-time operating system (RTOS), responsible for process control, adaptive parameter calculation in step S3, order analysis in step S4.3, and feature extraction and lightweight diagnostic model inference in step S5. Diagnostic algorithm software is embedded within the microprocessor unit, receiving aligned corner domain data blocks from the field-programmable gate array for processing.

[0043] Interaction and communication module: including local display screen, indicator lights, buttons and industrial communication interface. The local display screen and indicator lights are used to display speed, waveform, diagnostic results and alarm status. The industrial communication interface includes but is not limited to RS-485, Ethernet and 4G, which are used to upload diagnostic results and key data to the cloud or host computer system.

[0044] Furthermore, the device also includes a self-learning and optimization module, which runs in software within the microprocessor unit. This module periodically collects the operating data of the device in different speed ranges, automatically fine-tunes the decision threshold of the diagnostic model, and securely synchronizes the updated model to the edge device through incremental learning.

[0045] This invention directly acquires rotational speed information from existing electronic control systems or simple sensors, and uses advanced signal processing algorithms to robustly estimate instantaneous rotational speed even under strong noise, eliminating the need for expensive encoders and reducing deployment costs and barriers to entry. It does not rely on additional rotational speed measurement hardware, and its adaptive pulse processing mechanism ensures system stability even when the rotational speed signal is imperfect.

[0046] By dynamically driving all subsequent analysis parameters with instantaneous rotational speed, the diagnostic system can automatically adapt to all operating conditions from low to high speed, always focusing the analysis on the fault characteristic frequency band related to the current rotational speed, thus solving the problem of fixed-parameter systems failing under varying operating conditions.

[0047] By employing a hardware-software co-design that utilizes a Field-Programmable Gate Array (FPGA) to process low-level signals and a Microprocessor Unit (MCU) to run intelligent algorithms, computationally intensive preprocessing is offloaded to the FPGA, ensuring real-time data preprocessing. The lightweight diagnostic model running on the MCU maintains high accuracy while incurring minimal computational and storage overhead, enabling the complete intelligent diagnostic process to run in real-time on resource-constrained embedded terminals with faster response times, achieving true edge-side intelligence.

[0048] The device is highly integrated and easy to install, requiring only a vibration sensor and 1-2 signal lines. It provides local display and alarm functions and supports remote communication, making it ideal for digital transformation and predictive health management of a large number of in-service equipment.

[0049] Example 2: Specific Calculation Flow and Formulas for Diagnostic Methods S1. Synchronous signal acquisition: Assume the acquired discrete sequence of vibration signals is as follows: The sampling frequency is The original speed pulse signal discrete sequence is: .

[0050] S2, Instantaneous speed adaptive extraction: 1. Regarding Perform digital low-pass filtering to remove high-frequency noise.

[0051] 2. Adaptive threshold zero-crossing detection: Set a dynamic threshold. ,in =0.6 is an empirical coefficient. The rising edge of the detection signal exceeds... The moment is denoted as the pulse arrival time. .

[0052] 3. Instantaneous speed calculation and calibration: Pulse interval .

[0053] Instantaneous rotational speed (rpm): .in, Let be the instantaneous rotational speed at the i-th pulse moment. It represents the time interval between the i-th and (i-1)-th pulses.

[0054] Pulse loss handling: If If so, it is assumed that a pulse may have been lost in the middle. At this time, and Insert a virtual pulse time point between And use this to calculate the rotational speed of the interpolation point.

[0055] Interference pulse rejection: If If the current pulse is considered to be interference, it will be removed and not included in the speed calculation sequence.

[0056] 4. The discrete instantaneous rotational speed points Through cubic spline interpolation, the signal was resampled to match the vibration signal. Synchronous instantaneous rotational speed sequences with the same sampling rate .

[0057] 5. Regarding Savitzky-Golay smoothing filtering with a window length of 101 and a polynomial order of 3 is performed to obtain the final smoothed instantaneous speed used for control. and .

[0058] S3. Adaptive analysis parameter dynamic calculation: Assuming the current smoothed instantaneous frequency is... Focus on the order of fault characteristics in the bearing outer ring. .

[0059] The current center frequency is: .

[0060] Perform envelope analysis and set the passband of the bandpass filter to 0. .

[0061] Focus on order range The analysis bandwidth is .

[0062] Resampling angle interval set to According to the highest level of attention The angle increment for equal-angle resampling is dynamically determined to ensure that the Nyquist sampling theorem is satisfied.

[0063] S4. Angular domain resampling and feature extraction of vibration signals: 1. Angle calculation: For Numerical integration, .

[0064] 2. Isoangular resampling: Generate isoangular sequences For each Find satisfaction index Then the resampling time Obtained through linear interpolation: in, Let θk be the k-th angle value in the target's equal angle sequence, where θk = k ⋅ Δθ. To the angle of the target The corresponding time point, Enclosing the original time series Two time points, These are the two angle values ​​that enclose θk in the original angle sequence.

[0065] vibration signal amplitude By analyzing the original signal exist The value is obtained by performing cubic spline interpolation at any given time.

[0066] 3. Feature Calculation: For Perform FFT to obtain the order spectrum Feature extraction: 1X, 2X, 3X order amplitudes: .

[0067] Bearing Failure Characteristic Order Amplitude and its harmonics at the point The amplitude and sum: .

[0068] by The spectral entropy of the five spectral lines on each side of the center represents the periodicity of the impact: The spectral entropy value is... Let be the normalized magnitude (probability) of the i-th order component. Let be the amplitude of the i-th order. The total number of order components involved in the calculation. Used to calculate the entropy value of the spectrum within a specific order range, quantifying the complexity of the spectrum or the periodicity of the shock. A smaller value indicates stronger periodicity.

[0069] Feature vector ,in This represents the average rotational speed of the current data segment.

[0070] S5. Fault diagnosis based on lightweight intelligent models: Using the LightGBM model. Feature vectors. The model is trained on historical data in the cloud and exported in a format callable by embedded C language. In the MCU, the diagnostic process is a forward inference step, outputting the probabilities of each category. The fault probabilities are then set. The threshold is set to trigger a local alarm.

[0071] Example 3: Embedded Device Workflow 1. Power-on initialization: When the device is powered on, the microprocessor unit and field-programmable gate array are initialized, and diagnostic algorithms and model parameters are loaded.

[0072] 2. Synchronous Acquisition: Vibration sensor and speed signal line are connected. The ADC, under the control of a field-programmable gate array, operates at a fixed frequency. =25.6 kHz synchronous sampling of two signals.

[0073] 3. Real-time preprocessing by the Field-Programmable Gate Array (FPGA): The ADC data stream directly enters the FPGA. The FPGA's logic units perform pulse detection, calibration, interpolation, and smoothing filtering on the speed signal stream in real time, as described in step S2, outputting a smooth instantaneous speed sequence. At the same time, according to and preset Field-programmable gate array (FPGA) calculates angles in real time. It also performs real-time interpolation on the vibration signal stream and outputs vibration data blocks at equal angles. .

[0074] 4. Intelligent Analysis by Microprocessor Unit: The Field Programmable Gate Array (FPGA) transmits packaged data via a high-speed bus. The corresponding average rotational speed is sent to the microprocessor unit. After receiving a complete data block, the microprocessor unit initiates a diagnostic task: performing order analysis, feature extraction, and invoking the LightGBM engine for inference.

[0075] 5. Results Output and Interaction: Diagnostic results and key indicators are displayed on the local screen and uploaded to the SCADA system via RS-485 interface using the Modbus RTU protocol. If a fault is diagnosed, a buzzer alarm is triggered and a red indicator light flashes.

[0076] The scope of protection of this invention is not limited to the above embodiments. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this invention should be included within the scope of protection of this invention.

[0077] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0078] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0079] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fault diagnosis method for rotating machinery under varying operating conditions based on instantaneous speed adaptation, characterized in that, Includes the following steps: S1. Synchronously acquire vibration signals from rotating machinery. and the raw speed signal originating from the equipment's electronic control system. ; S2, regarding the original rotation speed signal The process involves adaptively extracting the instantaneous rotational speed sequence synchronized with the vibration signal. ; S3, based on the instantaneous speed sequence Dynamically calculate the analysis parameters of the vibration signal at the current moment, including the range of analysis orders. Equal angle resampling interval ,in S4, based on the instantaneous rotational speed sequence and dynamically calculated resampling interval The vibration signal is resampled from the time domain to the angular domain to obtain the angular domain signal. and to Perform order analysis to extract the order feature vector F, which is independent of rotational speed; S5. Input the order feature vector F into the pre-trained lightweight intelligent diagnostic model to obtain the fault classification results and equipment health status.

2. The method for fault diagnosis of rotating machinery under varying operating conditions based on instantaneous speed adaptation according to claim 1, characterized in that, The specific process for the instantaneous rotational speed adaptive extraction in step S2 is as follows: S2.1, targeting The pulse signal in the calculation uses an adaptive threshold to detect the pulse leading edge and calculate the pulse interval. And convert it to instantaneous speed ; When an abnormal pulse interval is detected, pulse loss interpolation compensation or interference pulse elimination logic is initiated. At the same time, periodic harmonic pulses are identified and suppressed, the autocorrelation function of the pulse interval sequence is calculated, and by finding the peak value of the autocorrelation sequence on the non-zero hysteresis, the true fundamental frequency period can be inferred and verified. The correct pulse interval sequence is obtained by eliminating harmonic interference pulses, thus completing the initial correction of the abnormal signal. S2.2, targeting The three-phase motor current signal is then subjected to Clarke-Park transform to extract the rotational frequency component, thereby estimating the instantaneous speed. ; S2.

3. Smooth the estimated instantaneous speed sequence to obtain a continuous and smooth instantaneous speed curve. and instantaneous frequency switching .

3. The method for fault diagnosis of rotating machinery under varying operating conditions based on instantaneous speed adaptation according to claim 1, characterized in that, The specific method for dynamically calculating the analysis parameters in step S3 is as follows: Let the current instantaneous frequency be... The order of the largest characteristic of the fault of interest is Then the equal angle resampling interval Based on Shannon's sampling theorem, it is adaptively determined as follows: ,in The oversampling coefficients are used; the upper limit of the analysis frequency is adaptively set to... .

4. The method for fault diagnosis of rotating machinery under varying operating conditions based on instantaneous speed adaptation according to claim 1, characterized in that, Step S4 specifically includes: S4.1, For instantaneous frequency conversion By integrating, we can obtain the relationship between the rotation angle and time. ; S4.2, as determined in step S3 Generate equiangular sequences for intervals ,according to With time The mapping relationship is found by interpolation algorithm for each... corresponding time The vibration amplitude at that moment is obtained by interpolation, forming an angular domain equally spaced sampling signal. ; S4.3, Diagonal domain signal Spectral analysis was performed to obtain the order spectrum. And extract the amplitude, harmonic energy, sideband modulation energy or spectral entropy at a predetermined order from the order spectrum as the feature vector F.

5. The method for fault diagnosis of rotating machinery under varying operating conditions based on instantaneous speed adaptation according to claim 1, characterized in that, The lightweight intelligent diagnostic model in step S5 is a gradient boosting decision tree model; after the model is trained in the cloud or on a host computer, it is embedded into an edge computing device with fixed parameters to run.

6. An embedded device for implementing the method of any one of claims 1-5, characterized in that, It includes a sensor interface module, a signal conditioning and synchronous acquisition module, a main control and computing module, and an interaction and communication module; The sensor interface module is used to connect the vibration acceleration sensor and the sensor or signal line that provides the raw rotation speed signal; The signal conditioning and synchronous acquisition module includes circuits for conditioning the vibration signal and the rotation speed signal respectively, and a synchronous analog-to-digital converter for synchronously converting the two analog signals into digital signals. The main control and computing module employs a field-programmable gate array (FPGA). and microprocessor unit (MCU) The heterogeneous architecture includes an FPGA configured to perform instantaneous rotational speed estimation and angular domain resampling interpolation calculation in real time, and a microprocessor unit configured to perform dynamic calculation of analysis parameters, order analysis, feature extraction, and intelligent diagnostic model inference. The interaction and communication module is used to provide a human-computer interaction interface and external data communication.

7. The embedded device according to claim 6, characterized in that, In the signal conditioning and synchronous acquisition module, the vibration signal conditioning circuit provides DC bias and drive for the MEMS accelerometer, and includes a voltage buffer, an anti-aliasing filter, and a programmable gain amplifier; the rotation speed signal conditioning circuit includes a Schmitt trigger for shaping the pulse signal and an isolation amplifier for isolating the analog signal; the synchronous analog-to-digital converter synchronously samples the two signals at a sampling rate of not less than 20 kHz.

8. The embedded device according to claim 6, characterized in that, The field-programmable gate array (FPGA) communicates with the microprocessor unit via a high-speed parallel bus or a serial peripheral interface; the FPGA outputs vibration data blocks at equal angles and the corresponding average rotational speed values ​​to the buffer of the microprocessor unit in real time using a stream processing method.

9. The embedded device according to claim 6, characterized in that, The interaction and communication module includes a local display screen, status indicator lights, an alarm buzzer, buttons, and at least one industrial communication interface, which includes RS-485, Ethernet, or wireless cellular network interface.

10. The embedded device according to claim 6, characterized in that, The microprocessor unit also runs a self-learning and optimization module, which periodically collects data on the device's operating characteristics to fine-tune diagnostic thresholds or securely receives and updates local diagnostic model parameters from the cloud.

Images