Eddy current sensor signal noise reduction method and system based on frequency division detection low-pass filtering

Abstract

The invention provides an eddy current sensor signal noise reduction method based on frequency division detection low-pass filtering and a monitoring system, belongs to the technical field of sensor signal processing, and can at least partially solve the problems of unstable signals and inaccurate measurement in the prior art. The signal noise reduction method for the eddy current sensor comprises the following steps: processing an original output signal of the eddy current sensor by adopting a synchronous detection module, demodulating a dynamic vibration alternating current component of the original output signal into a direct current voltage signal by utilizing a reference signal, and separating a static displacement direct current component signal; an independent dynamic signal channel and an independent static signal channel are formed; filtering high-frequency noise in the dynamic signal channel through a dynamic channel low-pass filter with a first preset cut-off frequency, and filtering power frequency interference in the static signal channel through a static channel low-pass filter with a second cut-off frequency; and reversely injecting compensation voltage at the output end of the static channel low-pass filter so as to eliminate the direct-current offset error of the static displacement component signal.

Description

Technical Field

[0001] This invention relates to sensor signal processing technology, specifically to a noise reduction method for eddy current sensor signals based on frequency division detection and low-pass filtering. Furthermore, this invention also relates to a rotating machinery condition monitoring system. This invention is mainly applied to eddy current sensor signal processing in rotating machinery monitoring systems such as turbine generator set TSI (turbine detection) systems. Background Technology

[0002] In the power and petrochemical industries, condition monitoring of rotating machinery (such as steam turbine generators) relies on eddy current sensors to collect rotor vibration and displacement signals in real time. Eddy current sensors, with their non-contact and fast-response characteristics, are commonly used in TSI systems of equipment such as steam turbine generator sets to monitor rotor vibration and displacement. While the non-contact measurement characteristics of eddy current sensors are advantageous, the complex industrial environment makes sensor signals susceptible to interference. On the one hand, high-frequency electromagnetic noise (typically 25kHz-150kHz) generated by equipment such as frequency converters can reduce the signal-to-noise ratio and introduce waveform glitches. On the other hand, dynamic vibration and static displacement signals can overlap in traditional processing circuits, causing DC offset interference during vibration information extraction and AC ripple instability when reading static displacement values.

[0003] Existing noise reduction methods all have drawbacks: low-pass filters with fixed cutoff frequencies struggle to balance preserving wideband signals and filtering out noise, and may even cause phase distortion; software filtering is limited by processor real-time performance, introducing delays in high-speed multi-channel systems and resulting in poor suppression of transient noise. Furthermore, simple detector circuits lack sufficient accuracy in separating AC and DC signals and cannot effectively suppress residual noise and temperature drift. These problems lead to unstable sensor signals, inaccurate measurements, and impair the TSI system's ability to diagnose equipment faults early.

[0004] Specifically, existing signal processing solutions suffer from the following three defects: First, insufficient high-frequency interference suppression: low-pass filters with fixed cutoff frequencies struggle to simultaneously retain wideband signals (0.5Hz-10kHz) and filter out high-frequency noise, and are prone to phase distortion; Second, signal aliasing: dynamic vibration (AC) and static displacement (DC) components crosstalk each other in the detection circuit, causing vibration measurements to be affected by DC offset interference and displacement readings to be affected by AC ripple; Third, temperature drift control defects: simple compensation circuits cannot suppress drift over a wide temperature range of -40℃ to 85℃ (measured operational amplifier misalignment caused a ±42mV offset), affecting long-term stability.

[0005] Therefore, there is an urgent need for a signal noise reduction processing technology for eddy current sensors. Summary of the Invention

[0006] The present invention aims to at least solve one of the technical problems existing in the prior art, and provides a method for denoising eddy current sensor signals based on frequency division detection low-pass filtering and a rotating machinery condition monitoring system.

[0007] One aspect of the present invention provides a noise reduction method for eddy current sensor signals based on frequency division detection low-pass filtering, comprising: S1: processing the original output signal of the eddy current sensor using a synchronous detection module, and using a reference signal that is in phase and frequency with the excitation source of the eddy current sensor to demodulate the dynamic vibration AC component of the original output signal into a DC voltage signal, and separating the static displacement DC component signal to form independent dynamic signal channels and static signal channels; S2: filtering out high-frequency noise in the dynamic signal channel through a dynamic channel low-pass filter with a first preset cutoff frequency, and filtering out power frequency interference in the static signal channel through a static channel low-pass filter with a second preset cutoff frequency; and injecting a compensation voltage in reverse at the output of the static channel low-pass filter to eliminate the DC offset error of the static displacement component signal.

[0008] Specifically, the synchronous detection module employs an analog multiplier.

[0009] Preferably, the dynamic channel low-pass filter adopts a second-order Chebyshev low-pass filter with a first preset cutoff frequency of 20kHz, and the static channel low-pass filter adopts a second-order Butterworth low-pass filter with a second preset cutoff frequency of 5Hz.

[0010] More specifically, the second-order Chebyshev low-pass filter has an in-band ripple of 0.5dB, a roll-off slope of ≥-40dB / decibels at the first preset cutoff frequency of 20kHz, and adopts a Sallen-Key topology.

[0011] Preferably, a bias compensation circuit is provided at the output of the static channel low-pass filter, the compensation voltage is generated by an adjustable voltage source, and the temperature drift is corrected in real time based on a temperature-compensation voltage lookup table.

[0012] Optionally, the bias compensation circuit includes a low temperature drift voltage reference source with a temperature drift ≤2ppm / ℃ and a digital potentiometer, which serve as the adjustable voltage source.

[0013] More preferably, the temperature drift correction of the bias compensation circuit is achieved by the following method: within a set temperature range, the static channel zero-point drift is calibrated with a preset temperature value as the step size, the compensation voltage-temperature lookup table is generated by linear regression fitting, and the compensation voltage is generated in real time by the controller to perform temperature drift correction.

[0014] More specifically, within the set temperature range of -40°C to 85°C, the static channel zero-point drift is calibrated in steps of 10°C using the preset temperature value.

[0015] As a preferred embodiment, the reference signal is generated by synchronously shaping the excitation signal provided by the preamplifier of the eddy current sensor through a phase-locked loop, and the phase jitter is ≤0.1°.

[0016] Another aspect of the present invention provides a rotating machinery condition monitoring system, comprising an eddy current sensor, wherein the signal processing unit of the eddy current sensor processes the signal acquired by the eddy current sensor using any of the noise reduction methods described above.

[0017] This invention relates to an eddy current sensor signal noise reduction method and a rotating machinery condition monitoring system based on frequency division detection low-pass filtering technology. The method separates dynamic vibration AC signals from static displacement DC signals using a synchronous detection module. It uses a dynamic channel low-pass filter with a first preset cutoff frequency to filter out high-frequency noise, and a static channel low-pass filter with a second cutoff frequency to filter out power frequency interference. DC offset error is eliminated through compensation voltage. This eddy current sensor signal noise reduction method achieves breakthrough technical effects through collaborative innovation in hardware circuitry. Its core lies in fundamentally solving the three long-standing pain points in rotating machinery monitoring: signal aliasing, high-frequency interference, and temperature drift, significantly improving the measurement accuracy and reliability of the TSI system. At the signal separation level, based on a physical isolation mechanism, high-precision decoupling of the dynamic vibration AC component and the static displacement DC component is achieved. The AC / DC signal separation accuracy can reach ±0.5% of full scale, effectively eliminating baseline fluctuations and measurement distortion caused by crosstalk between dynamic and static parameters in traditional solutions. This lays a hardware foundation for independently monitoring rotor vibration characteristics and axial position. To suppress high-frequency noise, the dual-channel independent filtering architecture provides precise frequency management capabilities. At the same time, through temperature adaptability and compensation voltage, it effectively eliminates the DC offset error of the static displacement component signal.

[0018] In some preferred embodiments, the dynamic signal channel employs a 20kHz cutoff second-order Chebyshev filter, which, while maintaining the integrity of the 0.5Hz-10kHz mechanical fault characteristic frequency band (in-band ripple ≤ ±0.2dB), achieves attenuation of industrial frequency converter harmonics ≥25kHz by ≥-40dB. The static signal channel is supplemented with a 5Hz cutoff Butterworth filter, effectively filtering out residual power frequency interference. This two-stage synergy ensures that the effective value of the output signal ripple is stably controlled at ≤3mV rms, equivalent to improving the signal-to-noise ratio by more than 26dB, significantly enhancing the ability to capture weak fault characteristics.

[0019] In some embodiments, an innovative dynamic temperature-compensated lookup table mechanism is employed. A compensation model is established through full-temperature-range calibration from -40℃ to 85℃, combined with real-time correction using a high-resolution digital potentiometer, suppressing the temperature drift of the static displacement channel to ≤0.01%FS / ℃. This overcomes the temperature hysteresis and nonlinearity defects of traditional analog compensation circuits, ensuring long-term measurement stability under wide temperature conditions.

[0020] Actual measurements show that the output ripple RMS value is ≤3mV, the AC / DC separation accuracy reaches ±0.5%, and the temperature drift is ≤0.01%FS / ℃. All indicators were verified through spectrum analysis, full-scale calibration, and temperature chamber gradient testing, effectively improving signal stability and measurement accuracy. The final comprehensive performance indicators include: dynamic vibration signal ripple ≤3mV rms at a 200μm pp range, static displacement channel temperature drift ≤0.008%FS / ℃ at a ±2mm range, and AC / DC channel crosstalk suppression ratio ≥60dB. The all-hardware processing chain ensures signal delay <10μs, meeting the millisecond-level response requirements of the TSI system. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of an eddy current sensor signal noise reduction method based on frequency division detection low-pass filtering according to some embodiments of the present invention. Figure 2 This is a system architecture diagram of an eddy current sensor signal noise reduction system based on frequency division detection and low-pass filtering according to some embodiments of the present invention; Figure 3 This is a circuit diagram of the phase-locked loop in the phase-locked loop synchronous detection module in some embodiments of the present invention; Figure 4 This is a circuit diagram of the analog multiplier in the phase-locked loop synchronous detection module in some embodiments of the present invention; Figure 5 This is a schematic diagram of the circuit topology of the dynamic signal channel in a dual-channel low-pass filter in some embodiments of the present invention; Figure 6 This is a schematic diagram of the circuit topology of the static signal channel in a dual-channel low-pass filter in some embodiments of the present invention; Figure 7 This is a schematic diagram of the dynamic temperature compensation correction circuit in some embodiments of the present invention; Figure 8 This is a flowchart of dynamic temperature compensation correction data processing in some embodiments of the present invention; Figure 9 This is a schematic diagram illustrating the window width adaptive moving average algorithm in some embodiments of the present invention; and Figure 10 This is a schematic diagram of the digital notch spectrum of the LMS algorithm in some embodiments of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] like Figures 1 to 10 As shown, the eddy current sensor signal noise reduction method based on frequency division detection low-pass filtering in the basic embodiment of the present invention includes: Step S1: Processing the original output signal of the eddy current sensor using a synchronous detection module, and using a reference signal that is in phase and frequency with the excitation source of the eddy current sensor to demodulate the dynamic vibration AC component of the original output signal into a DC voltage signal, and separating the static displacement DC component signal to form independent dynamic signal channels and static signal channels; Step S2: Filtering out high-frequency noise in the dynamic signal channel through a dynamic channel low-pass filter with a first preset cutoff frequency, and filtering out power frequency interference in the static signal channel through a static channel low-pass filter with a second cutoff frequency; and injecting a compensation voltage in reverse at the output of the static channel low-pass filter to eliminate the DC offset error of the static displacement component signal.

[0024] In the above-described basic embodiments of the present invention, the present invention separates the dynamic vibration AC signal and the static displacement DC signal through a synchronous detection module, uses a dynamic channel low-pass filter with a first preset cutoff frequency to filter out high-frequency noise, uses a static channel low-pass filter with a second cutoff frequency to filter out power frequency interference, and eliminates DC offset error through compensation voltage. This eddy current sensor signal noise reduction method achieves a breakthrough technical effect through collaborative innovation of hardware circuits. Its core lies in fundamentally solving the three major pain points that have long existed in the field of rotating machinery monitoring: signal aliasing, high-frequency interference, and temperature drift, significantly improving the measurement accuracy and reliability of the TSI system. At the signal separation level, based on a physical isolation mechanism, high-precision decoupling of the dynamic vibration AC component and the static displacement DC component is achieved. The AC / DC signal separation accuracy can reach ±0.5% of the full scale, effectively eliminating baseline fluctuations and measurement distortions caused by crosstalk between dynamic and static parameters in traditional solutions, laying a hardware foundation for independent monitoring of rotor vibration characteristics and axial position. For high-frequency noise suppression, the dual-channel independent filtering architecture forms a precise frequency management capability, and at the same time, through temperature adaptability and compensation voltage, the DC offset error of the static displacement component signal is effectively eliminated.

[0025] Based on the above basic implementation, optionally, the synchronous detection module employs an analog multiplier. Furthermore, the dynamic channel low-pass filter can be a second-order Chebyshev low-pass filter with a first preset cutoff frequency of 20kHz, and the static channel low-pass filter can be a second-order Butterworth low-pass filter with a second preset cutoff frequency of 5Hz. Typically, the second-order Chebyshev low-pass filter has an in-band ripple of 0.5dB, a roll-off slope at the first preset cutoff frequency of 20kHz ≥ -40dB / decibels, and employs a Sallen-Key topology.

[0026] In a preferred embodiment, a bias compensation circuit is provided at the output of the static channel low-pass filter. The compensation voltage is generated by an adjustable voltage source, and temperature drift is corrected in real time based on a temperature-compensation voltage lookup table. Specifically, the bias compensation circuit includes a low-temperature drift voltage reference source with a temperature drift ≤2ppm / ℃, used as the adjustable voltage source, and a digital potentiometer.

[0027] In practical implementation, the temperature drift correction of the bias compensation circuit is achieved in the following way: within a set temperature range, the static channel zero-point drift is calibrated with a preset temperature value as a step size. The compensation voltage-temperature lookup table is generated through linear regression fitting, and the compensation voltage is generated in real time by the controller to perform temperature drift correction. For example, the static channel zero-point drift can be calibrated with a preset temperature value of 10°C within the set temperature range of -40°C to 85°C.

[0028] In some preferred embodiments, the reference signal can be generated by the excitation signal provided by the preamplifier of the eddy current sensor after synchronous shaping by a phase-locked loop, and the phase jitter is ≤0.1°.

[0029] In the various embodiments described above, the present invention actually proposes a three-level collaborative hardware noise reduction architecture: Phase-locked physical isolation layer: Physical separation of dynamic and static signals is achieved through a synchronous detection module (e.g., an analog multiplier with phase jitter ≤0.1°); Dual-channel independent filtering layer: For example, the dynamic signal channel can use a 20kHz Chebyshev filter to preserve the fault frequency band, while the static signal channel can use a 5Hz Butterworth filter to eliminate power frequency ripple; Dynamic temperature compensation layer: For example, a lookup table mechanism based on full-temperature-range calibration can drive digital potentiometers to compensate in real time.

[0030] This three-level collaborative hardware noise reduction architecture improves signal separation accuracy. For example, the analog multiplier's quadrature demodulation achieves AC / DC separation accuracy of ±0.5%FS, effectively eliminating baseline fluctuations caused by aliasing; the phase-locked loop's phase jitter is ≤0.1°, ensuring the linearity of dynamic signal demodulation. High-frequency noise depth is suppressed; for example, the 20kHz Chebyshev filter attenuates ≥25kHz noise by ≥40dB, with passband ripple ≤0.5dB; dual-channel independent filtering reduces output ripple to ≤3mV rms, improving the signal-to-noise ratio by 26dB. Furthermore, temperature drift is also significantly controlled: for example, a digital potentiometer (0.1mV resolution) combined with lookup table compensation compresses static displacement temperature drift to ≤0.01%FS / ℃; the equivalent drift at full scale ±2mm is <0.4μm / ℃.

[0031] In addition to the above-described eddy current sensor signal noise reduction method, the present invention also provides a rotating machinery condition monitoring system, which includes an eddy current sensor, wherein the signal processing unit of the eddy current sensor uses the above-described noise reduction method to process the signal collected by the eddy current sensor.

[0032] To help those skilled in the art understand the above embodiments of the present invention, the following description is provided in conjunction with... Figures 1 to 10 A more comprehensive preferred embodiment of the present invention is described in more detail to demonstrate more technical details in the specific implementation, so that those skilled in the art can fully implement the present invention.

[0033] When implementing the hardware-circuit-coordinated frequency division detection and low-pass filtering noise reduction method of this invention, it is necessary to follow a three-level processing architecture of phase-locked loop physical isolation → dual-channel independent filtering → dynamic temperature compensation correction, combined with auxiliary enhancement measures, and proceed in an orderly manner step by step. First, the original sensor signal is processed using an analog multiplier synchronized with the phase-locked loop. In rotating machinery condition monitoring systems, the raw signals output by eddy current sensors often contain AC components of dynamic vibration and DC components of static displacement, along with high-frequency interference. Traditional processing methods struggle to accurately separate these signal components. However, by using a phase-locked loop-synchronized analog multiplier to process the sensor's raw signals, and through a unique circuit design and signal processing mechanism, the aliasing of dynamic and static signals can be eliminated at the physical level, significantly improving signal quality. The phase-locked loop (PLL) plays a crucial role in frequency tracking and phase locking during this process. Its operating principle is based on feedback control theory, consisting of a phase detector (PD), a loop filter (LF), and a voltage-controlled oscillator (VCO) forming a closed-loop system. The signal from the eddy current sensor's excitation source is input to the PLL's phase detector. The phase detector compares the frequency and phase of the input signal with the VCO's output signal, outputting an error voltage reflecting the difference. This error voltage, after being filtered by the loop filter, is used to adjust the VCO's output frequency, continuously approximating the input signal frequency until they are identical and the phase difference is constant, thus achieving real-time tracking of the sensor's excitation source frequency by the PLL. In this invention, the PLL is required to track the sensor's excitation source frequency range of 1kHz to 2MHz, ensuring stable operation under various operating conditions of industrial equipment.

[0034] Analog multipliers are key execution elements for achieving signal separation; see [link to documentation]. Figure 4 As shown, the present invention can use the ADI AD834 analog multiplier. This multiplier features high precision, low noise, and wide bandwidth, with a temperature drift of ≤0.5μV / ℃, which significantly reduces the impact of temperature on signal processing compared to the 2mV / ℃ temperature drift of traditional diode detection schemes. After the phase-locked loop locks the sensor excitation source frequency, it outputs a local reference signal that is in phase and frequency with the excitation source and has a phase jitter of ≤0.1°. This reference signal and the original sensor signal are simultaneously input into the analog multiplier. According to the quadrature demodulation principle, the analog multiplier performs a multiplication operation on the two input signals. For the dynamic vibration AC component (0.5Hz - 10kHz mechanical fault characteristic frequency band) in the original signal, after multiplication with the in-phase and frequency reference signal, and subsequent low-pass filtering, it can be converted into a baseband DC voltage. The amplitude change of this DC voltage reflects the characteristics of the dynamic vibration signal; while the static displacement DC component in the original signal is directly separated during the multiplication operation. In this way, by processing with an analog multiplier, the dynamic vibration AC component and the static displacement DC component are accurately separated at the physical level, avoiding the interference of their aliasing on subsequent signal analysis. In the actual circuit construction process, careful design and debugging of the peripheral circuits of the phase-locked loop (PLL) and analog multiplier are required. For the PLL, the parameters of the phase detector, loop filter, and voltage-controlled oscillator (VCO) must be selected appropriately to ensure that it can quickly and stably lock onto the sensor excitation source frequency within the 1kHz-2MHz frequency range. For example, the bandwidth of the loop filter needs to be adjusted according to the system's response speed and stability requirements; too wide a bandwidth may lead to system instability, while too narrow a bandwidth will affect the frequency tracking speed. For the analog multiplier, attention should be paid to the stability of its power supply, using a low-noise power supply module to reduce the impact of power supply noise on signal processing; at the same time, the routing of input and reference signals should be arranged reasonably to avoid crosstalk between signals. To further optimize signal processing, the operating parameters of the analog multiplier can be calibrated. By inputting a standard signal with known frequency and amplitude, parameters such as the gain and bias of the analog multiplier can be adjusted to ensure the accuracy of its output signal. In industrial applications, factors such as ambient temperature and electromagnetic interference can affect circuit performance. Therefore, it is also necessary to regularly maintain and calibrate the signal processing circuit composed of the phase-locked loop and the analog multiplier to ensure its long-term stable operation. By processing the original sensor signal through a phase-locked loop synchronized analog multiplier, not only is the dynamic vibration and static displacement signals effectively separated, but temperature sensitivity is also reduced from the source. This lays a solid foundation for subsequent signal filtering, error correction, and other processing steps, significantly improving the accuracy and reliability of eddy current sensor signal processing and meeting the high-precision signal requirements of rotating machinery condition monitoring systems.

[0035] Second, build a dual-channel independent filtering channel. For the dynamic vibration channel, which carries mechanical fault characteristic frequency band signals of 0.5Hz - 10kHz, the main noise source is the inverter harmonic interference of ≥25kHz in industrial environments. Therefore, a second-order Chebyshev low-pass filter with a cutoff frequency of 20kHz was selected and constructed using a Sallen-Key topology. This topology features simple circuitry and ease of debugging. Furthermore, the second-order Chebyshev filter maintains a small ripple of 0.5dB within the passband and achieves a steep roll-off of -40dB / dec after the cutoff frequency, thus both preserving the fault characteristic frequency band signals and effectively suppressing high-frequency harmonics. In component selection, to ensure the stability of the filter performance, resistors with a low temperature drift of ±5ppm / ℃ were chosen. These resistors exhibit minimal resistance fluctuations with changes in ambient temperature, preventing cutoff frequency shifts due to temperature variations and ensuring effective filtering. NPO capacitors with ΔC / C ≤ ±0.3% were selected, as their capacitance changes minimally with temperature, ensuring the filter maintains an accurate cutoff frequency under different temperature conditions. Testing showed that using these components resulted in a cutoff frequency shift of <0.1%. Based on the design formula for a second-order Chebyshev low-pass filter and the requirement of a 20kHz cutoff frequency, the specific parameter values ​​for the resistors and capacitors were calculated and determined. These components were then assembled and soldered according to the Sallen-Key topology to complete the construction of the dynamic signal channel filter.

[0036] The static signal channel primarily processes DC signals but is frequently affected by residual 50Hz power frequency ripple. To address this characteristic, a second-order Butterworth filter with a cutoff frequency of 5Hz is selected. The Butterworth filter is characterized by a flat frequency response within its passband, minimizing the impact on the static displacement DC signal while effectively filtering out the 50Hz power frequency ripple. In component selection, stability and temperature characteristics are also carefully considered to ensure reliable filter performance. Following the design methodology of a second-order Butterworth filter, suitable resistors and capacitors are calculated and selected to construct the filtering circuit for the static displacement channel.

[0037] After the dual-channel filter circuit was built, overall debugging and optimization were performed. Test signals of different frequencies were generated using a signal generator and input into the dynamic vibration and static displacement channels respectively. The changes in signal waveforms before and after filtering were observed using an oscilloscope, and the frequency response curve of the filter was measured. Based on the measurement results, the parameters of components such as resistors and capacitors were fine-tuned to further optimize the filter performance, ensuring that the dynamic signal channel could effectively suppress high-frequency noise while completely preserving fault characteristic frequency signals; and that the static signal channel could completely filter out power frequency ripple, outputting a stable and pure DC signal. In addition, temperature testing was conducted on the dual-channel filter circuit to simulate different temperature environments, monitor changes in filter performance, and verify the role of low-temperature drift components in ensuring cutoff frequency stability. This ensured that the dual-channel filter maintained good filtering performance within the operating temperature range of -40℃ to 85℃, providing high-quality input signals for subsequent signal processing. Third, implement a dynamic temperature replenishment mechanism. The implementation of the dynamic temperature compensation mechanism aims to solve the signal drift problem caused by temperature changes, especially the DC offset caused by operational amplifier misalignment in the static channel. Through a series of rigorous steps, it ensures the stability and accuracy of the signal under different temperature environments. First, a thorough analysis of the error causes reveals that in the static signal channel, operational amplifier offset is the primary factor leading to DC offset, with a measured maximum offset reaching ±42mV. Furthermore, traditional processing methods exhibit significant temperature drift, severely impacting signal quality. To address this issue, a dedicated compensation circuit is required, employing a high-precision digital potentiometer AD5171 and an auto-zero operational amplifier MCP6V01 to form the core compensation structure. The AD5171 offers high-precision adjustment capabilities, with a compensation resolution of 0.1mV, enabling fine-tuning of DC offset. The MCP6V01's auto-zero characteristic effectively reduces the operational amplifier's own offset voltage, and the combination of these two components forms a reliable hardware compensation foundation. The digital potentiometer and the auto-zero operational amplifier are connected according to the circuit design requirements, constructing a feedback loop capable of adjusting the compensation voltage in real time based on signal changes.

[0038] Secondly, key data calibration was performed: To establish an accurate temperature-compensation voltage mapping relationship, a gradient temperature chamber calibration experiment was conducted within a temperature range of -40℃ to 85℃, with a step size of 10℃. The constructed static displacement signal processing circuit was placed in the temperature chamber, and at each temperature node, a high-precision voltmeter was used to collect the DC offset data of the static displacement signal. Before each acquisition, the circuit was allowed to stabilize at that temperature for a period of time to ensure that the collected data accurately reflected the signal state at the current temperature. A large amount of offset data at different temperatures was recorded to form a raw dataset. Through statistical analysis and mathematical modeling of these data, and using fitting algorithms such as the least squares method, an accurate temperature-compensation voltage mapping table was established. This table can accurately reflect the correspondence between different temperatures and the required compensation voltage. Next, a real-time monitoring and control module is built: a surface-mount PT1000 temperature sensor is installed near the static displacement signal processing circuit, utilizing its resistance value to linearly change with temperature to sense the ambient temperature in real time. The PT1000 temperature sensor is connected to a signal conditioning circuit, converting temperature changes into a voltage signal output. An STM32G4 microcontroller is selected as the control core, connected to the temperature sensor and a digital potentiometer. The microcontroller acquires the voltage signal output by the temperature sensor through its internal A / D conversion module and converts it into an actual temperature value. Based on the real-time temperature value, the microcontroller retrieves the corresponding compensation voltage value from a pre-established temperature-compensation voltage mapping table. Using communication protocols such as SPI, the compensation voltage value is converted into a control signal and sent to the AD5171 digital potentiometer to dynamically adjust its resistance value, thereby changing the output voltage of the compensation circuit and achieving real-time compensation for DC offset error. Finally, the dynamic temperature compensation mechanism can be continuously optimized and verified: Compensated signal data at different temperatures is collected periodically and compared with uncompensated data and standard signals to evaluate the compensation effect. If a deviation in compensation is found, the accuracy of the temperature-compensation voltage mapping table is rechecked, or the parameters of the compensation circuit are fine-tuned. Simultaneously, the complex and variable temperature environment of industrial sites is simulated to test the response speed and compensation accuracy of the compensation mechanism under conditions of rapid temperature changes and fluctuations. This ensures that under various operating conditions, temperature drift can be compressed to ≤0.01%FS / ℃ (equivalent drift <0.4μm / ℃ at full scale ±2mm), significantly improving the stability and reliability of eddy current sensor signal processing and providing accurate and stable data support for the condition monitoring of rotating machinery.

[0039] Fourth, implement auxiliary enhancement measures. After completing core noise reduction steps such as phase-locked loop physical isolation, dual-channel independent filtering, and dynamic temperature compensation correction, implementing auxiliary enhancement measures can further improve signal quality and meet the stringent requirements of rotating machinery condition monitoring systems. The auxiliary enhancement measures of this invention mainly include two techniques: window-width adaptive moving average processing and digital notch filtering based on the LMS algorithm. These two techniques complement each other and effectively suppress residual noise and interference. This auxiliary measure enhances real-time performance: the above-mentioned all-hardware link delay is <10μs, meeting the millisecond-level response requirements of TSI systems; simultaneously, the LMS algorithm's dynamic notch filtering achieves a power frequency interference suppression ratio ≥60dB.

[0040] For low-pass filtered dynamic signals, an adaptive moving average processing method with an adaptive window width is implemented. In industrial settings, the vibration frequency of rotating machinery varies with operating conditions. Using a fixed-window moving average algorithm makes it difficult to achieve optimal noise reduction at different frequencies. Therefore, this invention dynamically adjusts the moving average window width by monitoring the frequency characteristics of the dynamic signal in real time. Specifically, digital signal processing technology is first used to perform spectral analysis on the dynamic signal. The Fast Fourier Transform (FFT) algorithm can be used to convert the time-domain signal into a frequency-domain signal, thereby obtaining the main frequency components of the signal and calculating the vibration period. Based on the calculated vibration period, the moving average window period is set to 1.5 times the vibration period. This ratio has been verified through extensive experiments to effectively smooth the signal, reduce ripple, and retain key feature information in the signal to the greatest extent.

[0041] In other words, in the preferred embodiment of the above-described specific implementation of the present invention, a dedicated digital signal processing module is designed in terms of hardware implementation. This module can employ a high-performance digital signal processor (DSP) or a microcontroller with powerful computing capabilities (such as the STM32G4 series). The low-pass filtered dynamic signal is input into this module, which processes the signal point by point according to a set window width. Starting from the first data point, data points accumulating to 1.5 times the number of vibration cycles form an initial window, and the average value of the data within the window is calculated as the output. As new data is continuously input, the window is updated in a sliding manner, discarding the earliest data points and incorporating new data points, recalculating the average value, and continuously outputting the smoothed signal. After this processing, the effective value of the output ripple can be stably controlled within ≤3mV rms, significantly improving the stability and purity of the signal. To suppress any residual power frequency interference, a digital notch filtering technique based on the LMS (Least Mean Square) algorithm is used. In industrial environments, although 50Hz power frequency interference is significantly weakened after low-pass filtering and other processing, frequency offset or harmonic component interference signals may still exist. Traditional fixed-frequency notch filters struggle to handle frequency fluctuation interference. This invention utilizes the LMS algorithm to dynamically adjust the notch filter's center frequency. In practice, a digital notch filter is integrated into the hardware circuit. This notch filter can employ an Infinite Impulse Response (IIR) filter structure, which is suitable for real-time hardware processing due to its high computational efficiency and low resource consumption. The signal processed by low-pass filtering and moving average is used as the input to the digital notch filter, while a short segment of undisturbed, clean signal is selected as a reference signal (this can be acquired during system initialization or during periods of low interference). The LMS method compares the error between the notch filter's output signal and the reference signal, using the gradient descent principle to continuously adjust the coefficients of the digital notch filter, gradually reducing the error. When the power frequency interference frequency changes, the algorithm can detect the frequency shift in the signal in real time and automatically adjust the notch filter's center frequency to precisely align with the interference frequency, effectively suppressing the interference. In practical applications, this method can improve the noise suppression ratio to ≥60dB, ensuring the signal is unaffected by power frequency interference. To ensure the coordinated operation of the two auxiliary enhancement measures, the hardware circuitry and software algorithms need to be optimized. In terms of hardware layout, the positions of components such as the digital signal processing module and digital notch filter are rationally planned to reduce signal transmission delay and interference. At the software level, efficient program code is written to achieve parallel processing of the moving average algorithm and the LMS algorithm, ensuring the real-time performance of the entire signal processing chain. Actual testing shows that the signal delay of the entire hardware processing chain is <10μs, meeting the millisecond-level response requirements of the TSI system. Simultaneously, it further improves the reliability and accuracy of the eddy current sensor signal, providing solid data support for fault diagnosis and condition monitoring of rotating machinery.

[0042] The following are some embodiments of the present invention at the specific implementation level, for reference by those skilled in the art.

[0043] Example 1: Signal separation achieved through phase-locked loop physical isolation Circuit setup: The phase-locked loop uses the ADF4113 chip, and the phase detector bandwidth is set to 1 / 10 of the excitation frequency to ensure that the lock-in time is <100μs in the range of 1kHz-2MHz; The analog multiplier selected is AD834, whose -40dBc harmonic distortion ensures demodulation accuracy. In this Example 1, the phase jitter was 0.08° (measured), and the dynamic channel temperature drift was reduced to 0.3μV / ℃, thus reducing temperature sensitivity at the source.

[0044] Example 2: Design of a Dual-Channel Independent Filter Dynamic channel filter: Based on Sallen-Key topology, with R1=R2=8.2kΩ (±5ppm / ℃) and C1=C2=1nF (NPO material), the measured attenuation at 20kHz is -42dB / decibels; Static channel filter: second-order Butterworth structure, cutoff frequency 5Hz, filtering out 50Hz ripple attenuation >30dB; In this embodiment 2, the cutoff frequency shift is less than 0.08% within the temperature range of -40℃ to 85℃, ensuring that the fault frequency band of 0.5Hz-10kHz is completely preserved.

[0045] Example 3: Implementation of Dynamic Temperature Compensation Mechanism Calibration process: Temperature chamber gradient test: -40℃→85℃, offset voltage is collected after 30 minutes of constant temperature at every 10℃; Linear regression fitting: Temperature drift slope k=0.0075mV / ℃, table lookup resolution 0.1mV; Real-time compensation: PT1000 sensor output is sampled by STM32G4 built-in ADC (12-bit precision); SPI interface controls AD5171 digital potentiometer, compensation response time <1ms; In this embodiment 3, the temperature drift is 0.007%FS / ℃ when the full scale is ±2mm, which is 6 times better than the traditional solution.

[0046] Example 4: Auxiliary Enhancement Measures Adaptive moving average: The dynamic signal is calculated by FFT to obtain the main frequency f, and the window width is 1.5 / f (e.g., when f=100Hz, the window is 15ms); the output ripple is reduced to 2.8mV rms (measured); LMS digital notch filter: IIR filter order=4, iteration step size μ=0.01, center frequency tracking accuracy ±0.2Hz; In this embodiment 4, the power frequency interference suppression ratio is 62dB and the end-to-end delay is 9.3μs.

[0047] As described above, this invention addresses the core challenges of eddy current sensor signals in rotating machinery condition monitoring systems, including susceptibility to high-frequency interference, aliasing of dynamic and static components, and temperature drift. It proposes a hardware-circuit-based frequency division detection and low-pass filtering noise reduction method. The core technology lies in constructing a three-level processing architecture: phase-locked loop physical isolation, dual-channel independent filtering, and dynamic temperature compensation correction. Signal purification is achieved through innovative analog circuit design. The core technical approach involves: firstly, using a phase-locked loop-synchronized analog multiplier (such as the ADI AD834) to process the original sensor signal. This multiplier tracks the sensor excitation source frequency (1kHz-2MHz) in real time, generating a local reference signal with phase jitter ≤0.1°. Using the orthogonal demodulation principle, the dynamic vibration AC component (0.5Hz-10kHz mechanical fault characteristic frequency band) is converted into a baseband DC voltage, while simultaneously separating the static displacement DC component, thus eliminating aliasing of dynamic and static signals at the physical level. Compared to the 2mV / ℃ temperature drift of traditional diode detection solutions, this design uses an analog multiplier with a temperature drift of ≤0.5μV / ℃, reducing temperature sensitivity at the source. Secondly, a dual-channel filtering system is designed: for the separated dynamic vibration channel, a second-order Chebyshev low-pass filter (Sallen-Key topology) with a cutoff frequency of 20kHz is configured. Its 0.5dB in-band ripple and -40dB / dec roll-off characteristics can effectively suppress inverter harmonic interference (the main noise source in industrial sites) of ≥25kHz while fully preserving fault characteristic frequencies such as rotor imbalance and misalignment within 10kHz. A second-order Butterworth filter with a 5Hz cutoff is added to the static displacement channel to specifically filter out residual 50Hz power frequency ripple. The filter components use resistors with a temperature drift of ±5ppm / ℃ and NPO capacitors with ΔC / C ≤ ±0.3%, ensuring that the cutoff frequency shift is <0.1% when the ambient temperature changes. Secondly, a dynamic temperature compensation mechanism is employed: the DC offset caused by operational amplifier misalignment in the static channel (measured maximum ±42mV) is compensated by constructing a compensation circuit using a high-precision digital potentiometer (such as AD5171) and an auto-zeroing operational amplifier (MCP6V01), achieving a compensation resolution of 0.1mV. Based on the calibration data of the gradient temperature chamber from -40℃ to 85℃ (step size 10℃), a temperature-compensation voltage mapping table is established. The ambient temperature is monitored in real time by a surface-mount PT1000 temperature sensor, and the STM32G4 microcontroller dynamically loads the compensation value, ultimately compressing the temperature drift to ≤0.01%FS / ℃ (equivalent drift <0.4μm / ℃ at full scale ±2mm). Finally, through auxiliary enhancement measures, an adaptive sliding average with a window width is applied to the dynamic signal after low-pass filtering (window period = 1.5 times the oscillation period), stabilizing the output ripple RMS value to ≤3mV rms. For residual power frequency interference, the LMS algorithm is used to adjust the center frequency of the digital notch filter in real time, improving the noise suppression ratio to ≥60dB. The all-hardware processing link ensures signal delay of <10μs, meeting the millisecond-level response requirements of the TSI system.

[0048] The eddy current sensor signal noise reduction method provided by this invention achieves breakthrough technical results through collaborative innovation in hardware circuitry. Its core value lies in fundamentally solving the three major pain points that have long existed in the field of rotating machinery monitoring: signal aliasing, high-frequency interference, and temperature drift, significantly improving the measurement accuracy and reliability of TSI systems.

[0049] At the signal separation level, a physical isolation mechanism based on a phase-locked loop synchronized analog multiplier achieves high-precision decoupling of the dynamic vibration AC component and the static displacement DC component, with AC / DC signal separation accuracy reaching ±0.5% of full scale. This technology completely eliminates baseline fluctuations and measurement distortions caused by crosstalk between dynamic and static parameters in traditional solutions, laying a hardware foundation for independent monitoring of rotor vibration characteristics and axial position.

[0050] For high-frequency noise suppression, the dual-channel independent filtering architecture demonstrates precise frequency management capabilities. The dynamic channel employs a 20kHz cutoff second-order Chebyshev filter, which maintains the integrity of the 0.5Hz-10kHz mechanical fault characteristic frequency band (in-band ripple ≤ ±0.2dB) while achieving ≥-40dB attenuation of industrial frequency converter harmonics ≥25kHz. The static channel is supplemented by a 5Hz cutoff Butterworth filter, effectively filtering out residual power frequency interference. The two-stage synergy ensures that the effective value of the output signal ripple is stably controlled at ≤3mV rms, equivalent to improving the signal-to-noise ratio by more than 26dB, significantly enhancing the ability to capture weak fault characteristics.

[0051] Breakthrough progress has been made in temperature adaptability. An innovative dynamic temperature compensation lookup table mechanism, through full-temperature-range calibration from -40℃ to 85℃, establishes a compensation model and combines it with real-time correction using a high-resolution digital potentiometer, suppressing the temperature drift of the static displacement channel to ≤0.01%FS / ℃. This technology overcomes the temperature hysteresis and nonlinearity defects of traditional analog compensation circuits, ensuring long-term measurement stability under wide temperature conditions.

[0052] The final comprehensive performance indicators include: dynamic vibration signal ripple ≤3mVrms at a 200μm pp range, static displacement channel temperature drift ≤0.008%FS / ℃ at a ±2mm range, and AC / DC channel crosstalk suppression ratio ≥60dB. A fully hardware-based processing chain ensures signal delay <10μs, meeting the millisecond-level response requirements of the TSI system.

[0053] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for denoising eddy current sensor signals based on frequency division detection and low-pass filtering, characterized in that, include: S1: The original output signal of the eddy current sensor is processed by a synchronous detection module, and the dynamic vibration AC component of the original output signal is demodulated into a DC voltage signal by using a reference signal that is in phase and frequency with the excitation source of the eddy current sensor, and the static displacement DC component signal is separated to form independent dynamic signal channel and static signal channel. S2: High-frequency noise is filtered out in the dynamic signal channel by a dynamic channel low-pass filter with a first preset cutoff frequency, and power frequency interference is filtered out in the static signal channel by a static channel low-pass filter with a second preset cutoff frequency. And a compensation voltage is injected in reverse at the output of the static channel low-pass filter to eliminate the DC offset error of the static displacement component signal.

2. The eddy current sensor signal noise reduction method according to claim 1, characterized in that, The synchronous detection module uses an analog multiplier.

3. The eddy current sensor signal noise reduction method according to claim 1, characterized in that, The dynamic channel low-pass filter adopts a second-order Chebyshev low-pass filter with a first preset cutoff frequency of 20kHz, and the static channel low-pass filter adopts a second-order Butterworth low-pass filter with a second preset cutoff frequency of 5Hz.

4. The eddy current sensor signal noise reduction method according to claim 3, characterized in that, The second-order Chebyshev low-pass filter has an in-band ripple of 0.5dB, a roll-off slope of ≥-40dB / decibels at the first preset cutoff frequency of 20kHz, and adopts a Sallen-Key topology.

5. The eddy current sensor signal noise reduction method according to claim 1, characterized in that, A bias compensation circuit is set at the output of the static channel low-pass filter, the compensation voltage is generated by an adjustable voltage source, and the temperature drift is corrected in real time based on a temperature-compensation voltage lookup table.

6. The eddy current sensor signal noise reduction method according to claim 5, characterized in that, The bias compensation circuit includes a low temperature drift voltage reference source with a temperature drift of ≤2ppm / ℃, which serves as the adjustable voltage source, and a digital potentiometer.

7. The eddy current sensor signal noise reduction method according to claim 5, characterized in that, The temperature drift correction of the bias compensation circuit is achieved in the following way: within the set temperature range, the static channel zero-point drift is calibrated with a preset temperature value as the step size, the compensation voltage-temperature lookup table is generated by linear regression fitting, and the compensation voltage is generated in real time by the controller to perform temperature drift correction.

8. The eddy current sensor signal noise reduction method according to claim 7, characterized in that, Within the set temperature range of -40°C to 85°C, the static channel zero-point drift is calibrated in steps of 10°C using the preset temperature value.

9. The eddy current sensor signal noise reduction method according to any one of claims 1 to 7, characterized in that, The reference signal is generated by the excitation signal provided by the preamplifier of the eddy current sensor and synchronously shaped by the phase-locked loop, and the phase jitter is ≤0.1°.

10. A rotating machinery condition monitoring system, comprising an eddy current sensor, characterized in that, The signal processing unit of the eddy current sensor processes the signal acquired by the eddy current sensor using the noise reduction method according to any one of claims 1 to 9.

Images