Oil metal particle sensor micro-particle signal waveform identification method and system

Abstract

The application discloses a kind of oil metal particle sensor micro-particle signal waveform identification method and system, method includes: obtaining the differential voltage signal that oil metal particle sensor outputs, signal sequence is obtained by preprocessing;Peak detection is carried out to signal sequence, extract several candidate pulse pairs, and in signal sequence, the local signal segment containing each candidate pulse pair is intercepted;Gaussian function first derivative extension model for matching candidate pulse pair is constructed, and initial parameters are calculated based on candidate pulse pair, and iteration solution is carried out with initial parameters as starting point, and optimal fitting parameter is obtained;Based on optimal fitting parameter, generate fitting waveform, and calculate the comprehensive matching degree score between fitting waveform and local signal segment, whether matching candidate pulse is real metal particle signal based on comprehensive matching degree score and symmetry index judging is determined.The application is applied to signal processing field, can significantly improve metal particle detection precision, and anti-interference is strong, and robustness is excellent.

Description

Technical Field

[0001] This invention relates to the field of mechanical equipment condition monitoring and signal processing technology, specifically a method and system for identifying the waveform of minute particle signals from an oil-metal particle sensor. Background Technology

[0002] Online oil monitoring technology analyzes wear-prone metal particles carried in lubricating oil, effectively assessing the wear condition of mechanical equipment without disassembling the equipment. It has become a key tool for condition-based maintenance and equipment health management of complex mechanical systems such as engines and wind turbine gearboxes. Online oil particle sensors based on the principle of electromagnetic induction are the most widely used commercial sensors due to their simple structure, fast response, ability to distinguish between ferromagnetic and non-ferromagnetic particles, and strong anti-interference capabilities and adaptability to direct connection across all flow lines, often employing differential structures. However, in practical applications, these sensors still face significant signal detection challenges. When the wear-prone metal particles are small, the resulting changes in induced voltage are extremely weak, making them difficult to detect even under complex artifact interference.

[0003] These artifacts mainly originate from actual operating conditions and experimental environments, such as joints in polyurethane (PU) pipes used for calibration or connection, impurities in the oil, and inhomogeneities in the pipe wall material. These often cause significant baseline fluctuations or sudden pulses. Under such a complex background of strong artifacts, accurately extracting signals from magnetic particles with a diameter of less than 70 μm is extremely difficult, easily leading to missed detections and false alarms.

[0004] To address these issues, researchers have proposed various advanced methods for signal denoising and feature extraction, such as empirical mode decomposition and variational mode decomposition, and have attempted to utilize techniques such as fractional calculus enhancement and adaptive threshold segmentation. However, traditionally, the induced electromotive force signal waveform generated when worn metal particles pass through a sensor is generally simplified into a sine or cosine pulse, and recognition algorithms are designed based on this model.

[0005] Extensive experimental observations show that the output waveform of actual inductive metal particle sensors, due to the non-uniform magnetic field distribution and the complexity of particle motion, often resembles the first derivative waveform of a Gaussian function rather than a sine or cosine pulse shape. Existing signal processing methods for inductive metal particle sensors, which use simplified sine or cosine pulse models, struggle to distinguish between artifact signals and genuine metal particle output signals when faced with strong artifact interference from actual pipe joints, bubbles, etc. Therefore, a more accurate waveform description model and a corresponding robust waveform recognition method are urgently needed. Summary of the Invention

[0006] To address the shortcomings of the existing technology, this invention provides a method and system for identifying the waveform of minute metal particles from an oil-based metal particle sensor. Based on nonlinear fitting of the first derivative waveform of a Gaussian function and multidimensional comprehensive matching degree evaluation, it can effectively identify the output signal of minute metal particles from an inductive oil sensor.

[0007] To achieve the above objectives, the present invention provides a method for identifying the signal waveform of tiny particles using an oil-metal particle sensor, comprising the following steps: Step 1: Acquire the differential voltage signal output by the metal particle sensor and preprocess it to obtain the preprocessed signal sequence; Step 2: Perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract a local signal segment from the signal sequence that contains the complete positive and negative lobes and transition region of each candidate pulse pair. Step 3: Construct a Gaussian function first derivative extended model for matching candidate impulse pairs, calculate the initial parameters of the Gaussian function first derivative extended model based on the candidate impulse pairs, and iteratively solve the model using the initial parameters as the starting point to obtain the optimal fitting parameters; Step 4: Generate a fitted waveform based on the optimal fitting parameters, and calculate the comprehensive matching score between the fitted waveform corresponding to the candidate pulse and the local signal segment. Based on the comprehensive matching score and the symmetry index, determine whether the matched candidate pulse is a real metal particle signal.

[0008] To achieve the above objectives, the present invention also provides a system for identifying the waveform of minute particle signals from an oil-metal particle sensor, employing the above-described method. The system includes: The signal processing module is used to acquire the differential voltage signal output by the oil metal particle sensor and preprocess it to obtain the preprocessed signal sequence. The pulse extraction module is used to perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract a local signal segment from the signal sequence that contains the complete positive and negative lobes and transition regions of each candidate pulse pair. A parameter fitting model is used to construct a Gaussian function first derivative extension model for matching candidate impulse pairs. The initial parameters of the Gaussian function first derivative extension model are calculated based on the candidate impulse pairs. The optimal fitting parameters are obtained by iteratively solving the model starting from the initial parameters. A signal determination model is used to generate a fitted waveform based on the optimal fitting parameters, calculate the comprehensive matching score between the fitted waveform corresponding to the candidate pulse and the local signal segment, and determine whether the candidate pulse pair is a real metal particle signal based on the comprehensive matching score and the symmetry index.

[0009] Compared with the prior art, the present invention has the following beneficial technical effects: 1. This invention describes the actual output signal of the inductive metal particle sensor by using a Gaussian first derivative extended model with DC bias. The model parameters are automatically adjusted by nonlinear least squares fitting, which improves the waveform matching degree compared with the traditional sine / cosine pulse model, effectively enhancing the recognition capability of weak metal particle output signals and providing stronger recognition reliability. 2. This invention uses a multi-dimensional comprehensive matching degree fusion evaluation, which integrates correlation coefficient, determination coefficient, amplitude matching degree and shape score reflecting local morphology, to effectively distinguish between real Gaussian derivative type metal particle signals and asymmetric and irregular artifact pulses caused by pipeline joint fluctuations, etc., avoiding misjudgment based on a single criterion. With the help of constraint optimization fitting, it can effectively identify ferromagnetic metal particles smaller than 70μm in environments with strong interference such as joint fluctuations and bubbles, and has excellent anti-artifact interference capability. 3. In this invention, the initial parameters of the Gaussian first derivative extended model are automatically estimated by the physical characteristics of the candidate pulse pairs, ensuring the speed and stability of the fitting convergence. At the same time, the automatic backoff mechanism ensures that the algorithm can still output reasonable results under extreme noise when the optimization fails to converge. It does not require a preset fixed pulse width template and can automatically adapt to waveform scale changes caused by different flow velocities, thus exhibiting high robustness. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0011] Figure 1 This is a flowchart of the method for identifying the waveform of a tiny particle signal from an oil-metal particle sensor in an embodiment of the present invention; Figure 2 This is a schematic diagram of the original measured output signal of the sensor containing noise and artifacts in an embodiment of the present invention; Figure 3 This is a schematic diagram of the preprocessed output signal and the corresponding signal of the tiny metal particles identified by experts in an embodiment of the present invention; Figure 4This is a schematic diagram of the recognition result obtained by using the preprocessed signal in this embodiment of the invention; Figure 5 This is a comparison chart of the actual output signal of the particle and the waveform of the first derivative of the fitted Gaussian function in an embodiment of the present invention, wherein: Figure 5 (a) is a comparison chart of the identification results for the 5th group. Figure 5 (b) is a comparison chart of the identification results for the 9th group; Figure 6 This is a comparison diagram of the actual output signal of the particle and the fitted sinusoidal pulse signal waveform in an embodiment of the present invention, wherein: Figure 6 (a) is a comparison chart of the identification results for the 5th group. Figure 6 (b) is a comparison chart of the identification results for the 9th group.

[0012] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0013] 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0014] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0015] Example 1 This embodiment discloses a method for identifying the waveform of a microparticle signal from an oil-based metal particle sensor. First, the original sensor signal is acquired and preprocessed. Then, candidate pulse pairs are generated through peak detection. For each candidate pulse pair, local signal truncation, initial parameter estimation, constrained nonlinear least-squares fitting, multi-dimensional comprehensive matching degree evaluation, and validity judgment are performed sequentially. Finally, it is determined whether the candidate pulse pair represents a real metal particle signal. In application, wear status can be assessed by extracting features from the candidate pulse pairs corresponding to the metal particle signal.

[0016] refer to Figure 1 The method for identifying the waveform of minute particle signals from an oil-metal particle sensor in this embodiment mainly includes the following steps: Step 1: Acquire the differential voltage signal output by the inductive oil metal particle sensor and preprocess it to obtain the preprocessed signal sequence; Step 2: Perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract a local signal segment from the signal sequence that contains the complete positive and negative lobes and transition regions of each candidate pulse pair. Step 3: Construct a Gaussian function first derivative extended model for matching candidate impulse pairs, calculate the initial parameters of the Gaussian function first derivative extended model based on the candidate impulse pairs, and iteratively solve the model using the initial parameters as the starting point to obtain the optimal fitting parameters; Step 4: Generate a fitted waveform based on the optimal fitting parameters, and calculate the comprehensive matching score between the fitted waveform of the candidate pulse pair and the local signal segment. Based on the comprehensive matching score and the symmetry index, determine whether the candidate pulse pair is a real metal particle signal.

[0017] In the specific implementation of step 1, the output of the inductive oil metal particle sensor is a differential voltage signal, which is then converted by an analog-to-digital converter at a sampling frequency. Collected discrete sequences and the corresponding time vector ,in , This represents the number of sampling points. To suppress trend term drift caused by changes in temperature and oil quality, the original collected signal... Preprocessing is performed, including mean and trend removal, and high-frequency noise components are filtered out using a smoothing filter, resulting in the preprocessed signal sequence. .

[0018] In the specific implementation of step 2, the preprocessed signal sequence Peak detection is performed. Based on the physical characteristics of the worn metal particle signal, positive and negative peaks should alternate, and the time interval should be within a reasonable range. Based on this, each pair of positive and negative peaks that meets the time interval requirements and their time positions are extracted to form a candidate pulse pair. In this embodiment, the time interval between adjacent positive and negative peaks is set to 1ms-50ms. This time interval can cover the positive and negative peak time intervals of the waveform generated by metal particles passing through the sensor at speeds of 0.5m / s-10m / s. For each pair of positive and negative peaks that satisfies the alternation relationship and interval constraints, the time positions of the positive and negative peaks are extracted. , and peak amplitude , , forming candidate pulse pairs Among them, the candidate pulse pairs may be real metal particle signals or artifact interference. In practical applications, according to the principle of inductive metal particle sensors, for ferromagnetic metal particle signals, the sensor output waveform is positive peak first and negative peak second, while for non-ferromagnetic metal particles, it is exactly the opposite, with negative peak first and positive peak second.

[0019] After extracting all candidate pulse pairs, a local signal segment containing the complete positive and negative lobes and transition region of each candidate pulse pair can be extracted from the signal sequence. The specific implementation process is as follows: For each candidate pulse pair, first calculate the peak spacing. Then, at the peak center position Based on the baseline, expand to both sides peak spacing times An analysis window is formed. Among them, the preferred method is... This ensures that the analysis window fully encompasses the positive and negative lobes and transition region of the pulse. After determining the analysis window, the signal sequence... Extracting a local signal segment within this time interval At the same time from the time vector Intercepting local signal segments Corresponding time vector .in, , This represents the number of sampling points for the extracted pulse segment.

[0020] In the specific implementation of step 3, the first derivative of the Gaussian function is used to model the output signal waveform of the metal particle sensor. The first derivative of the Gaussian function is defined as:

[0021] in, For amplitude factor, This is a scale parameter (which determines the pulse width). For time, The center position of the waveform; In this embodiment, the baseline drift and the order of positive and negative peaks of the actual signal are taken into account, and a structure including polarity factors is constructed. With DC bias The extended model of the first derivative of the Gaussian function is:

[0022] in, The parameters to be fitted are... This is the transpose of the matrix.

[0023] In practical applications, polarity factor It can be determined based on the order of the positive and negative peaks; if the positive peak comes first, then... Otherwise take The positive and negative peaks of the extended model of the first derivative of the Gaussian function are located at... Peak amplitude is The peak-to-peak time interval is The initial parameters of the Gaussian function first derivative extended model are... , , , , , The initial parameters for calculating the first derivative of the Gaussian function based on candidate pulse pairs include: polarity factor, amplitude factor, waveform center position, scaling parameter, and DC bias. If the positive peak of the candidate pulse pair comes first, then the initial polarity factor ,otherwise ; Initial scale parameters ; Initial amplitude factor ; Initial waveform center position ; Initial DC bias , For local signal segments The mean.

[0024] Determining initial parameters After that, it is possible Starting from [the given information], solve the following nonlinear optimization problem:

[0025] Among them, the parameters are constrained by physical meaning. A >0, , , The value was determined before optimization. Unconstrained. A nonlinear least squares method is used to fit the Levenberg-Marquardt algorithm iteratively, with a maximum number of iterations of 1000 and a convergence tolerance of [value missing]. If the residual is less than the tolerance or the maximum number of iterations is reached during the iteration process, the optimal fitting parameters are output. If convergence fails (e.g., due to severe signal distortion leading to matrix singularity), then revert to the initial parameters. This means that a reasonable reference value can still be given at this time to ensure robustness.

[0026] After obtaining the optimal fitting parameters Then, the fitted waveform can be generated. Then calculate the local signal segment. With fitted waveform The comprehensive matching score includes multiple matching indicators such as correlation coefficient, determination coefficient, amplitude matching degree, and shape score, thereby evaluating the local signal segment from multiple aspects. With fitted waveform Overall similarity, goodness of fit, and local morphological consistency.

[0027] Correlation coefficient Used to measure actual local signal segments With fitted waveform The overall shape similarity between them is defined as follows: ,in, For local signal segments With fitted waveform covariance, , For local signal segments Standard deviation; Coefficient of determination Used to evaluate goodness of fit R 2 The closer to 1, the better the fit, which is defined as follows: ,in For local signal segments The mean; Amplitude matching Used for comparison with actual local signal segments With fitted waveform The consistency of peak-to-peak values ​​(the difference in amplitude between positive and negative peaks) between them. A value closer to 1 indicates better peak-to-peak consistency, which is defined as follows: ,in, For local signal segments peak value, To fit the waveform peak value; Shape Score Used for further evaluation of actual local signal segments With fitted waveform The local morphological details between them include: First-order difference sequence correlation coefficient The calculation process is as follows: first calculate the local signal segment. First-order difference and the fitted waveform First-order difference Then calculate the correlation coefficient of the first-order difference series. , for and covariance, , for , Standard deviation; Zero-crossing overlap rate The calculation process is as follows: detect the actual local signal segments respectively. With fitted waveform The set of zero-crossing positions is considered to coincide if the time difference between the corresponding zero-crossing points of the two sets is less than one sampling interval. It equals the number of coincident points divided by the total number of actual signal zero-crossing points; Therefore, the shape score can be calculated. ; Overall matching score The weighted sum of the above four indicators is:

[0028] in, , , , These are weighting coefficients. In this embodiment, the weights are... , , , In practical applications, the above weights can be adjusted according to sensor characteristics or actual operating conditions. The final comprehensive matching score is truncated to... interval, The closer the value is to 1, the higher the matching degree between the candidate pulse pair and the first derivative waveform of the Gaussian function, and the greater the probability that it is a real metal particle signal.

[0029] Theoretically, the positive and negative peaks of a truly effective metal particle output waveform are symmetrical. Therefore, in addition to setting a comprehensive matching score, a symmetry index is also set to determine the effectiveness of the metal particle output waveform, ensuring the basic symmetry of the pulse's positive and negative peaks and filtering out asymmetric artifact interference. The symmetry index for candidate pulse pairs... In this embodiment, if the symmetry index of the candidate pulse pair meets a set range, for example... Meanwhile, when the overall matching score of the candidate pulse pair is greater than or equal to a set threshold, for example... If the candidate pulse pair is selected, it is determined to be a valid metal particle output signal; otherwise, it is considered a false signal and discarded. For valid metal particle output signals, features such as peak-to-peak value, pulse width, and waveform energy can be further extracted for wear condition assessment.

[0030] It is worth noting that, although this embodiment Figure 1The steps are shown sequentially as indicated by the arrows, but they are not necessarily executed in the exact order indicated by the arrows. Unless otherwise specified in this document, there is no strict order in which these steps are performed; they can be executed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0031] The following example further illustrates the method for identifying the waveform of minute particle signals from an oil-metal particle sensor in this embodiment.

[0032] In this example, a cyclic testing platform for an oil-metal particle sensor was constructed. Ferromagnetic metal spheres with a diameter of 50 μm were encapsulated in PU tubing filled with lubricating oil. Driven by a control system, the PU tubing circulated through the metal particle sensor at a speed of 0.5 m / s. The sampling frequency was set to 10 kHz, and the original sensor output signal was acquired. The signal was then processed using both the method described in this embodiment and the sinusoidal pulse matching method. The measured output waveform of the sensor, including artifacts, is shown below. Figure 2 As shown, the output signal contains signals from more than two complete cycles, that is, signals from two metal particles. Figure 2 It is evident that the sensor output signal exhibits a clear periodic characteristic. Figure 3 for Figure 2 The preprocessed output signal after removing the mean, trend term, and smoothing filter Because the metal particles are only 50μm in diameter, the output signal is weak. The positions of the output signals corresponding to the particles were manually identified by experts and marked in the figure. In addition to the particle signals, It also contains strong artifact signals, which greatly interfere with the automatic identification of particle signals.

[0033] Will Figure 3 The preprocessed output signal shown is identified using the method of this invention, and the result is as follows: Figure 4 As shown in the figure, positive peaks are labeled with upper triangles, and negative peaks with lower triangles. A pair of upper triangles and a pair of lower triangles form a candidate pulse pair. In a candidate pulse pair, the positive peak may come first, or the negative peak may come first. Figure 4 There are 11 pairs of candidate pulses from left to right. The output of the method in this embodiment is shown in Table 1.

[0034] Table 1. Sensor output waveform recognition results for metal particles based on waveform matching of the first derivative of Gaussian function.

[0035] As shown in Table 1, the metal particle signals identified by the method in this embodiment are groups 5 and 9. Figure 4 The particles are marked with dashed boxes, with particle type 1 indicating a positive peak first and a negative peak later. The waveform matching degrees of the two sets of signals reach 0.948 and 0.949, respectively. The results show that the method in this embodiment can effectively overcome the interference of artifacts such as joints and impurities, and successfully identify the sensor output signal corresponding to a 50µm ferromagnetic metal particle.

[0036] Furthermore, the method of this embodiment is compared with the sinusoidal pulse waveform matching method. Figure 5 The image shows a comparison between the actual output signal of the particle and the waveform of the first derivative of the Gaussian function fitted by the method in this embodiment. The waveform of the first derivative of the Gaussian function is highly consistent with the output signal of the particle. Figure 6 The image shows a comparison between the actual output signal of the particle and the fitted sinusoidal pulse signal waveform. The fitted sinusoidal pulse waveform differs significantly from the actual output signal of the particle. The overall waveform matching degrees of the two pulses are only 0.5264 and 0.5608, respectively, which are far lower than those of the method of this invention. In summary, this embodiment significantly improves the ability to identify weak metal particle signals under strong artifact interference by using the first derivative waveform matching method of Gaussian function.

[0037] Example 2 Based on the method for identifying the waveform of a microparticle signal from an oil-metal particle sensor in Example 1, this example discloses a system for identifying the waveform of a microparticle signal from an oil-metal particle sensor. This system includes a signal processing module, a pulse extraction module, a parameter fitting model, and a signal determination model. Specifically: The signal processing module is used to acquire the differential voltage signal output by the oil metal particle sensor and preprocess it to obtain the preprocessed signal sequence. The pulse extraction module is used to perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract local signal segments from the signal sequence that contain the complete positive and negative lobes and transition regions of each candidate pulse pair. The parameter fitting model is used to construct the first derivative extended model of the Gaussian function for matching candidate impulse pairs. The initial parameters of the first derivative extended model of the Gaussian function are calculated based on the candidate impulse pairs. The optimal fitting parameters are obtained by iteratively solving the problem starting from the initial parameters. The signal determination model is used to generate a fitted waveform based on the optimal fitting parameters, and to calculate the comprehensive matching score between the fitted waveform corresponding to the candidate pulse and the local signal segment. Based on the comprehensive matching score and the symmetry index, it is determined whether the candidate pulse pair is a real metal particle signal.

[0038] In this embodiment, the specific working process and working principle of the signal processing module, pulse extraction module, parameter fitting model, and signal determination model are the same as those in Embodiment 1, and therefore will not be described again in this embodiment. Each module can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in the processor of a computer device in hardware form or independent of it, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to the above unit modules. Alternatively, it can be embedded in a digital signal processor or industrial control computer to achieve online real-time analysis.

[0039] The above description is only a preferred embodiment of the present invention and does not limit the scope of protection of the present invention. All equivalent structural transformations made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the scope of protection of the present invention.

Claims

1. A method for identifying the signal waveform of tiny particles from an oil-metal particle sensor, characterized in that, Includes the following steps: Step 1: Acquire the differential voltage signal output by the metal particle sensor and preprocess it to obtain the preprocessed signal sequence; Step 2: Perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract a local signal segment from the signal sequence that contains the complete positive and negative lobes and transition region of each candidate pulse pair. Step 3: Construct a Gaussian function first derivative extended model for matching candidate impulse pairs, calculate the initial parameters of the Gaussian function first derivative extended model based on the candidate impulse pairs, and iteratively solve the model using the initial parameters as the starting point to obtain the optimal fitting parameters; Step 4: Generate a fitted waveform based on the optimal fitting parameters, and calculate the comprehensive matching score between the fitted waveform of the candidate pulse pair and the local signal segment. Based on the comprehensive matching score and the symmetry index, determine whether the candidate pulse pair is a real metal particle signal.

2. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 1, characterized in that, In step 1, the preprocessing includes removing the mean, removing the trend term, and smoothing filtering.

3. The method for identifying the waveform of microparticle signals from an oil-metal particle sensor according to claim 1, characterized in that, In step 2, the process of extracting the local signal segment is as follows: The peak spacing is calculated based on the positive and negative peak positions of each candidate pulse pair. After expanding the peak spacing to both sides by a predetermined multiple based on the center position of the positive and negative peaks, a local signal segment containing the corresponding candidate pulse pair is extracted from the signal sequence.

4. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 1, 2, or 3, characterized in that, In step 3, the extended model of the first derivative of the Gaussian function is as follows: in, As a polarity factor, For amplitude factor, For scale parameters, For time, The center position of the waveform. DC bias, The parameters to be fitted are... This is the transpose of the matrix.

5. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 4, characterized in that, In step 3, the initial parameters , , , , , The initial polarity factor, amplitude factor, waveform center position, scale parameter, and DC bias are used. The initial parameters for calculating the extended model of the first derivative of the Gaussian function based on the candidate impulse pairs specifically include: If the positive peak of the candidate pulse pair comes first, then the initial polarity factor ,otherwise ; Initial scale parameters ,in, The peak spacing of the candidate pulse pairs; Initial amplitude factor ,in, , The peak amplitudes of the positive and negative peaks of the candidate pulse pair; Initial waveform center position ,in, , The time positions of the positive and negative peaks of the candidate pulse pair; Initial DC bias ,in, This refers to a local signal segment of the candidate pulse pair. For local signal segments The mean.

6. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 1, 2, or 3, characterized in that, In step 3, the Levenberg-Marquardt algorithm is used to iteratively solve the problem using a nonlinear least squares method: If the residual is less than the tolerance or the maximum number of iterations is reached during the iteration process, the optimal fitting parameters are output. If the iteration fails to converge, the initial parameters are output as the optimal fitting parameters.

7. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 1, 2, or 3, characterized in that, In step 4, the comprehensive matching score is specifically as follows: in, , , , These are the weighting coefficients. The correlation coefficient is... As the coefficient of determination, For amplitude matching degree, Score the shape; in, For local signal segments With fitted waveform covariance, , For local signal segments standard deviation For local signal segments The mean, , For local signal segments Fitted waveform peak value, For local signal segments With fitted waveform The correlation coefficient of a first-order difference sequence For local signal segments With fitted waveform The zero-crossing overlap rate.

8. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 7, characterized in that, The calculation process for the zero-crossing coincidence rate is as follows: Detect local signal segments separately With fitted waveform The set of zero-crossing positions is considered to coincide if the time difference between the corresponding zero-crossing points of the two is less than 1 sampling interval; The zero-crossing coincidence rate is the sum of the number of coincidence points and the local signal segment. The ratio of the total number of points crossing zero.

9. The method for identifying the waveform of a microparticle signal from an oil-metal particle sensor according to claim 1, 2, or 3, characterized in that, In step 4, the determination of whether the candidate pulse pair is a real metal particle signal is based on the comprehensive matching score and symmetry index, specifically including: Calculate the symmetry index of the candidate pulse pairs. , , The peak amplitudes of the positive and negative peaks of the candidate pulse pair; When the symmetry index of the candidate pulse pair meets the set range, and the comprehensive matching score of the candidate pulse pair is greater than or equal to the set threshold, the candidate pulse pair is determined to be a metal particle output signal.

10. A system for identifying the waveform of microparticle signals from an oil-based metal particle sensor, characterized in that, The system comprising the method according to any one of claims 1 to 9, wherein the system includes: The signal processing module is used to acquire the differential voltage signal output by the oil metal particle sensor and preprocess it to obtain the preprocessed signal sequence. The pulse extraction module is used to perform peak detection on the signal sequence, extract the positive and negative peak amplitudes and positions of all pulse pairs that meet the time interval requirements, form several candidate pulse pairs, and extract a local signal segment from the signal sequence that contains the complete positive and negative lobes and transition regions of each candidate pulse pair. A parameter fitting model is used to construct a Gaussian function first derivative extension model for matching candidate impulse pairs. The initial parameters of the Gaussian function first derivative extension model are calculated based on the candidate impulse pairs. The optimal fitting parameters are obtained by iteratively solving the model starting from the initial parameters. A signal determination model is used to generate a fitted waveform based on the optimal fitting parameters, calculate the comprehensive matching score between the fitted waveform corresponding to the candidate pulse and the local signal segment, and determine whether the candidate pulse pair is a real metal particle signal based on the comprehensive matching score and the symmetry index.

Images