Data processing method and system for phase difference method thickness gauge

Abstract

The invention discloses a data processing method and system for a phase difference method coating thickness gauge, and belongs to the technical field of nondestructive testing. The method aims at solving the problems that due to Gaussian random noise, periodic electromagnetic interference and outlier accidental noise, frequency-phase difference data are scattered, the linear fitting degree is low, and measurement repeatability is poor. The method comprises the following steps: acquiring original frequency-phase difference data; preprocessing is carried out through filtering and adaptive outlier detection; performing core straight line fitting on the preprocessed data by using robust regression algorithms such as random sample consensus (RANSAC), and extracting a stable linear relationship; according to the method, three types of typical noise can be restrained in a targeted mode, after treatment is conducted through the method, the standard deviation of repeated thickness measurement of the same standard part can be reduced by about 60%-80% compared with a direct linear fitting method, the repeatability and reliability of measurement are greatly improved, and an effective solution is provided for accurate and stable measurement of the coating thickness.

Description

I. Technical Field This invention relates to the field of nondestructive testing technology, specifically to a data processing method and system for improving the measurement accuracy of a phase difference method coating thickness gauge. II. Background Technology Phase difference thickness gauges (such as the CoatPro series) calculate thickness by measuring the phase difference under different frequency excitations. Ideally, frequency and phase difference are linearly related, and thickness can be obtained from the slope of a linear fit. However, in actual measurements, due to the combined effects of Gaussian random noise (such as sensor thermal noise), periodic electromagnetic interference, and occasional outliers (such as transient electromagnetic pulses or surface defect reflections), the acquired frequency-phase difference data points are often severely scattered, leading to a decrease in the coefficient of determination (R²) of direct linear fitting. 2 The low value of the coefficient of variation (COP) leads to an unstable slope of the fitted straight line, ultimately resulting in large errors in thickness calculation and poor measurement repeatability. Currently, the industry commonly uses general-purpose digital filters (such as low-pass filters) to smooth the original signal, but these methods are mainly for time-domain waveforms and are insensitive to non-Gaussian, local outliers. The linear optimization effect on the frequency-phase difference data, which exists as a two-dimensional scatter set, is limited, making it difficult to robustly extract the linear trend reflecting the true thickness information while suppressing various composite noises. III. Summary of the Invention This invention aims to overcome the shortcomings of the prior art and provide a data processing method and system for phase difference thickness gauges. This method effectively suppresses various types of noise, including Gaussian random noise, periodic interference, and occasional outliers, significantly improving the goodness of fit (R0) of the linear fitting of frequency-phase difference data. 2 This improves the accuracy, repeatability, and reliability of thickness measurements by enhancing their stability and stability.

[0001] Technical solution To achieve the above objectives, the present invention provides a data processing method for a phase difference thickness gauge, which addresses random noise and outlier interference, and includes the following steps: S1. Data Acquisition: Acquire the raw frequency-phase difference dataset obtained by measuring the same sample at multiple detection frequencies using a thickness gauge; S2. Data preprocessing: Frequency domain filtering is performed on the original dataset to suppress periodic interference, and an adaptive method based on moving window statistics is used to identify and correct outliers; S3. Robust Core Fitting: A robust regression algorithm such as Random Sample Consensus (RANSAC) is used to fit the preprocessed data points to a straight line to obtain the slope parameter that reflects the thickness information; S4. Model Optimization and Output: Calculate and output the thickness value based on the slope parameter.

[0002] Furthermore, the method of this invention can also be applied to measurement scenarios with localized systematic frequency response anomalies. In this application mode, by performing initial fitting and residual analysis on the data, continuous frequency response anomaly regions are intelligently identified, and data from anomaly regions are excluded during robust fitting, ultimately outputting thickness values ​​and related diagnostic information for abnormal frequency bands.

[0003] Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: i. Highly targeted and comprehensive noise suppression: The "preprocessing + robust fitting" combination scheme proposed in this invention is specifically designed with frequency domain filtering, adaptive outlier processing and robust regression algorithms. It can systematically handle the three main noise sources that cause data scattering: Gaussian random noise, periodic interference and occasional outliers. It overcomes the shortcomings of general filters in solving this type of two-dimensional scatter plot linearization problem.

[0004] ii. Significant improvement in linearity and accuracy: Experiments have verified that after applying the method of this invention, the linear fitting goodness (R0) of the frequency-phase difference data is significantly improved. 2 The value can be improved from approximately 0.70 to approximately 0.97 by traditional direct fitting. The single-measurement inversion error for standard thickness samples is significantly reduced.

[0005] iii. Significant Improvement in Measurement Repeatability and Stability: Most importantly, this invention greatly improves measurement repeatability by extracting robust linear trends from the data. When performing multiple (e.g., 20) repeated measurements on the same standard part, the standard deviation of the thickness calculation can be reduced by 60%-80% (e.g., from ±1.68). μm Decreased to ±0.54 μm This lays a solid foundation for high-precision, high-reliability online detection and quality control of coating thickness.

[0006] iv. Robust and widely applicable: Even under harsh conditions with significant electromagnetic interference or uneven sample surface conditions, the method of this invention can still maintain stable measurement performance, effectively expanding the applicability of phase difference thickness gauges in complex environments. IV. Description of the attached drawings Figure 1 This is a flowchart of the data processing method provided in Embodiment 1 of the present invention.

[0007] Figure 2 This is a schematic diagram comparing the distribution of frequency-phase difference data points and the fitted straight line before and after applying the method of the present invention in Example 1. It intuitively demonstrates the linearity (R²) of the data after processing by the present invention. 2 Significant improvement.

[0008] Figure 3This is a bar chart comparing the repeatability of multiple thickness measurements on the same standard part after fitting with the traditional direct linear fitting method and the method of this invention, used to demonstrate the outstanding effect of this invention in reducing measurement standard deviation.

[0009] Figure 4 This is a flowchart of the data processing method provided in Embodiment 2 of the present invention.

[0010] Figure 5 This is a schematic diagram comparing the distribution of frequency-phase difference data points and the fitted straight line before and after applying the method of the present invention in Example 2. It visually demonstrates the linearity (R0) of the data after processing by the present invention. 2 Significant improvement.

[0011] Figure 6 This is a comparison chart of the performance effects of the traditional method and the method of the present invention in Example 2. V. Detailed Implementation Methods To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0012] Example 1 This embodiment corresponds to a specific implementation of the method described in claim 1, demonstrating the process of applying the method of the present invention under random noise and outlier interference scenarios. The sample to be tested is a copper-based chromium-plated sample with a chromium layer thickness of 3 mm. μm See also Figure 1 This embodiment includes the following steps (wherein steps S101-S104 correspond to steps S1-S4 of claim 1 respectively): a. Step S101, Data Acquisition. A CoatPro thickness gauge is used to measure the copper-based chromium-plated single-layer coating sample. The scanning frequency is set to 100 points at equal intervals within the range of 100Hz to 200Hz. The data at each frequency point is recorded. f i Corresponding raw phase difference data φ i This forms the original dataset {( f 1 ,φ 1 ), ( f 2 ,φ 2 ), ..., ( f 100 ,φ 100 )}.

[0013] b. Step S102, Data Preprocessing. First, the original phase difference sequence is low-pass filtered to suppress high-frequency periodic noise. A 4th-order Butterworth filter is used, with the cutoff frequency set to one-tenth of the sampling frequency. Then, an adaptive outlier detection and correction algorithm based on a moving window is employed. The window width is set to 9 data points and slides along the frequency axis. For each window, the median (Med) and median absolute deviation (MAD) of the phase difference are calculated. If the phase difference value at a point deviates from the median by more than 4 times the MAD (i.e., threshold = 4 × 1.4826 × MAD), it is identified as an outlier. For outliers, the weighted average of other valid data within the window is used for replacement; the weight is inversely proportional to the distance of the data point from the median. This step effectively identifies and corrects artificially added outliers in the simulated data.

[0014] c. Step S103, Robust Core Fitting. For the preprocessed data, the Random Sample Consensus (RANSAC) algorithm is used for line fitting. Key parameters of the algorithm are set as follows: three points are randomly selected in each iteration to calculate the linear model; the inlier distance threshold is set to 0.15; the maximum number of iterations is 3000; and the confidence level is set to 99.99%. The RANSAC algorithm can find the optimal set of inliers (i.e., data points conforming to the linear model) and its corresponding best-fit linear model from data containing residual noise. φ = k_ransac * f + b_ransac ).

[0015] d. Step S104, Model Analysis and Optimization. To obtain a more accurate slope, iterative reweighted least squares (IRLS) fitting is performed on the set of interior points identified by the RANSAC algorithm to further reduce the influence of random noise. The optimized slope is then obtained. k_final The thickness value (T) is calculated according to the calibration formula: T = S * k_final ,in S The instrument sensitivity coefficient (set to 1 in this example) μm / (rad / Hz ) The final output is the thickness measurement value.

[0016] Effect verification and comparative analysis: In order to quantitatively evaluate the effect of the method of the present invention, it is compared with the result of directly fitting the original data by the traditional least squares method (OLS).

[0017] (1) Comparison of linear fit goodness Traditional OLS fitting of straight lines is severely affected by outliers, resulting in a low goodness of fit R0. 2 The original value was only 0.6986. However, after processing with the method of this invention, the linearity of the data is fundamentally improved, and the fitted straight line accurately penetrates the core distribution area of ​​the data, resulting in R0. 2 The value was significantly improved to 0.9755, an increase of 39.6%. See Table 1 for a comparison of the two fitting results. Figure 2.

[0018] Table 1. Comparison of fitting performance between the traditional method and the method of the present invention in Example 1.

[0019] (2) Comparison of measurement repeatability and accuracy The traditional method was tested by simulating 20 repeated measurements. Due to varying random noise and outlier interference in each measurement, the measured thickness values ​​fluctuated wildly, with a standard deviation as high as ±1.68. μm The method of this invention robustly extracts the stable linear trend of the data, resulting in highly consistent results across 20 measurements, with the standard deviation significantly reduced to ±0.54. μm Repeatability improved by 67.9%. Thickness measurements are shown in Table 2; distribution comparisons are shown in [reference needed]. Figure 3 .

[0020] Table 2 Comparison of results from 20 measurements using the conventional method and the method of this invention.

[0021] Table 2 (continued) Comparison of results from 20 measurements using the traditional method and the method of this invention

[0022] Conclusion: This embodiment demonstrates that the data processing method provided by the present invention can effectively resist multiple noise interferences, improve data linearity and measurement repeatability to the level required for application, and solve the problem of inaccuracy in phase difference thickness measurement under environmental interference.

[0023] Example 2: Verification of Coating Thickness Measurement Method for Regions with Abnormal Frequency Response In coating measurements based on the phase difference method, besides random noise and impulse interference, there exists a more complex and challenging systematic error: frequency response anomaly regions. When there is an acoustic impedance mismatch between the coating material and the substrate in a specific measurement frequency range, or when the coating itself possesses viscoelastic relaxation characteristics, the phase difference data acquired by the instrument in that frequency band will exhibit continuous and systematic deviations, rather than randomly distributed outliers. These clusters of "abnormal frequency bands" severely distort the overall linear trend of the frequency-phase relationship, causing traditional fitting methods to completely fail. To address this practical problem, embodiments of the present invention provide, for example... Figure 4 The innovative processing flow, which is a specific application of the method described in claim 1 in a scenario with an abnormal frequency response region, includes the following steps: a. Step S201, data preprocessing and initial robust fitting. First, wavelet transform threshold denoising is performed on the original frequency-phase difference data (including outlier regions, random noise, and impulse noise) to separate noise from useful signals; then, two-step outlier processing is performed to eliminate obvious impulse interference.

[0024] b. Step S202: Intelligent identification of abnormal frequency response regions. Based on the initially fitted residual sequence, a density-based sliding window statistical method is used for analysis. By detecting intervals where the residuals continuously deviate from the normal range on the frequency axis, the algorithm can automatically locate one or more "abnormal frequency response regions." This method overcomes the limitation of traditional methods that can only handle isolated outliers.

[0025] c. Step S203: Segmented Robust Fitting and Core Information Extraction. The data is intelligently divided into "credible normal frequency bands" and "questionable abnormal frequency bands." Only data points from the "normal frequency bands" are used for high-confidence robust regression (such as RANSAC) to extract the core slope K_core, which reflects the overall thickness of the coating and is unaffected by local systematic errors.

[0026] d. Step S204: Result Output and Intelligent Diagnosis. The final thickness value is calculated using K_core. Simultaneously, the system outputs a diagnostic report, clearly indicating the identified abnormal frequency ranges (e.g., "Systematic deviation exists in the 150-250 Hz range"), providing users with clues for assessing measurement reliability and improving processes.

[0027] Beneficial Effects and Comparative Conclusions: Experimental results show that under simulated strong noise and local anomalous frequency band interference, Traditional least squares method is affected by systematic bias, resulting in a low goodness of fit R0. 2 If the value is below 0.4, the thickness inversion error exceeds 20%.

[0028] The method of this invention accurately identifies and isolates abnormal frequency bands, R 2 The accuracy has been steadily improved to above 0.78, with a thickness inversion error of less than 10%.

[0029] Table 3. Comparison of fitting performance between the traditional method and the method of the present invention in Example 2.

[0030] Conclusion: This invention breaks through the limitations of traditional algorithms that only optimize "measured values," achieving an assessment of "measurement data quality itself" and providing location diagnostic information for "abnormal frequency bands." This means that the method has been upgraded from a single "measurement tool" to an "integrated measurement-diagnosis system," which has outstanding practical value for ensuring measurement reliability in complex environments and guiding production process optimization.

Claims

1. A data processing method for a phase difference thickness gauge, characterized in that, Includes the following steps: S1: Acquire the raw phase difference data sequence collected by the thickness gauge at multiple detection frequencies; S2: Preprocess the original data sequence to reduce the influence of Gaussian random noise, periodic interference noise and outlier noise; S3: A robust regression algorithm is used to fit a straight line to the preprocessed data sequence to obtain the slope parameter used to calculate the thickness; S4: Output the thickness measurement value based on the fitting result. After processing by the method, the standard deviation of repeated thickness measurement of the same standard part is significantly reduced compared with the direct linear fitting method.

2. The method according to claim 1, characterized in that, The preprocessing steps include: first, performing frequency domain filtering on the original data sequence to suppress periodic interference noise, and then using an adaptive method based on moving window statistics to identify and correct outliers.

3. The method according to claim 2, characterized in that, The adaptive method based on moving window statistics specifically involves: calculating the median and absolute deviation of the median of the data within the window; identifying points whose deviation from the median exceeds a preset threshold as outliers; and replacing them with the weighted average of the valid data within the window.

4. The method according to claim 1, characterized in that, The robust regression algorithm is a random sampling consensus algorithm.

5. The method according to claim 1, characterized in that, After obtaining the slope parameter, the model optimization step is also included: calculating the residuals of the data points relative to the fitted line. If the residual distribution shows a regularity, the model is optimized by introducing a compensation term or by piecewise fitting.

6. A data processing system for a phase difference thickness gauge, characterized in that, include: The data acquisition module is used to acquire the raw frequency-phase difference data sequence; Noise processing module, configured to perform the preprocessing steps as described in any one of claims 1 to 3; The robust fitting module is used to perform line fitting using a robust regression algorithm and obtain slope parameters. The thickness calculation and output module is used to calculate and output the thickness value based on the slope parameter.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 5.

Images