A drilling tool precision monitoring method and system based on artificial intelligence

Abstract

The present invention relates to the field of data processing technology, and in particular to a drilling tool precision monitoring method and system based on artificial intelligence. The method comprises: collecting a feed speed data sequence of a drilling tool during a drilling process, screening out all feature points in the feed speed data sequence, and determining a drilling cycle, calculating the independence of the data point and data points on both sides thereof, and taking a larger independence as the noise performance degree of the data point, correcting the noise performance degree according to a deviation value between a value of the data point and an average value of corresponding data points in all other drilling cycles, estimating an expected signal of the data point in an algorithm through the corrected noise performance degree, denoising the feed speed data sequence based on the expected signal, and monitoring the drilling tool according to the denoised feed speed data sequence, thereby improving the denoising effect and ensuring the monitoring accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a drilling tool precision monitoring method and system based on artificial intelligence. Background Art

[0002] With the rapid advancement of electronic industry technology, multi-layer printed circuit board (PCB) manufacturing technology is constantly moving towards higher density and refinement. In order to ensure high-quality processing of multi-layer PCB board holes, real-time monitoring and precise control of drilling tool processing parameters have become key links. For example, the patent document with authorization announcement number CN110647106B discloses a tool performance monitoring and evaluation method and system, which determines the performance parameters of each target tool during the production process; according to the performance parameters of each target tool, determines the quality parameters of the type of tool to which the target tool belongs, so as to better reflect the actual processing performance of the tool.

[0003] However, in the actual processing environment, signal acquisition during the drilling process is easily interfered by various external factors, resulting in noise mixed into the processing parameter data, affecting the processing accuracy. Therefore, it is necessary to effectively remove these noises in order to accurately evaluate and control the working state of the drilling tool.

[0004] The existing technology often uses the LMS (least mean square) algorithm to remove noise. The core goal of the LMS algorithm is indeed to adjust the weight of the filter so that the output signal (i.e. the filtered signal) is as close to the expected signal (or target signal) as possible. For example, the patent document with application number CN116524946A provides a method for blind source separation of vibration signals in a strong noise environment, which mainly imports audio signal data and aliases the audio signal; denoises through the LMS algorithm based on wavelet preprocessing, and performs dimension reduction and whitening preprocessing on the audio signal data to obtain a zero-mean whitened signal vector; and enhances the voice signal of the aliased signal. This method can effectively suppress noise mainly through the LMS algorithm based on wavelet preprocessing.

[0005] However, during the PCB drilling process, when the above solution is used to denoise the drilling signal data, due to the complex actual operating environment, the accurate expected signal of the LMS algorithm cannot be obtained, which affects the denoising effect of the algorithm, and thus the drilling tool cannot be accurately monitored, affecting the processing accuracy. Summary of the invention

[0006] In order to solve the problem that when the LMS algorithm is used to denoise the parameter data of the drilling tool, the denoising result has low accuracy and cannot accurately monitor the drilling tool, thereby affecting the processing accuracy, the present invention proposes a drilling tool accuracy monitoring method and system based on artificial intelligence. The specific technical scheme is as follows:

[0007] On the one hand, the present invention provides a drilling tool accuracy monitoring method based on artificial intelligence, comprising:

[0008] Collecting a feed speed data sequence of the drilling tool during the drilling process, and calculating the possibility that any data point in the feed speed data sequence is a feature point according to the numerical difference between the data points on both sides thereof, wherein the possibility is positively correlated with the numerical difference;

[0009] According to the possibility, all characteristic points in the feed speed data sequence are screened out, and every two adjacent characteristic points are regarded as a drilling cycle;

[0010] Performing data fitting on the feed speed data sequence, and taking the ratio of the fitting loss of the data point to the fitting loss of the data points on both sides thereof as the independence of the data point from the data points on both sides thereof, and taking the larger independence as the noise performance degree of the data point;

[0011] According to the deviation value between the value of the data point and the average value of the corresponding data points in all other drilling cycles, the noise performance degree of the data point is corrected, and the corrected noise performance degree is negatively correlated with the deviation value;

[0012] The noise performance degree of the data point after correction is used to estimate the data point An expected signal in the algorithm is used to denoise the feed speed data sequence based on the expected signal, and the drilling tool is monitored according to the denoised feed speed data sequence.

[0013] The above technical scheme improves the monitoring accuracy and efficiency during the drilling process. Specifically, firstly, by analyzing the numerical differences of data points in the feed speed data sequence, the characteristic points are identified and the drilling cycle is defined. This method can keenly capture the discontinuous changes in the processing process, ensure the accurate division of the drilling cycle, and provide an accurate time frame for subsequent precision analysis. Then, the independence of the data points is measured by the fitting loss ratio, and the degree of noise performance is determined. This strategy innovatively utilizes the local characteristics of the data points, effectively distinguishes signal changes from noise interference, improves the accuracy of identifying noise, and helps to more precisely locate and quantify the impact of noise. Then, the corrected noise performance degree and the expected signal estimation are corrected by combining the deviation value of the data point, so that the evaluation of the noise performance degree is closer to reality. The expected signal estimated in this way is more accurate, which enhances the pertinence and effectiveness of the denoising effect and provides a reliable reference standard for subsequent signal denoising. Finally, denoising and drilling tool monitoring optimization, based on the denoising processing of the corrected noise performance degree and the estimated expected signal, effectively filters out the noise interference that affects the accuracy and improves the purity of the feed speed data sequence. Through this series of operations, not only the accuracy of data analysis is improved, but also a high-quality data foundation is provided for the real-time monitoring and control of drilling tools, which helps to timely discover and adjust abnormalities in processing, avoid tool wear or degradation of processing quality, and ensure the stability and efficiency of the drilling process.

[0014] Preferably, the corrected noise performance level satisfies the following relationship:

[0015]

[0016] In the formula, is the noise performance degree of the data point after correction, is the noise performance of the data point, is the value of the data point, is the average value of the corresponding data points of the data point in other drilling cycles, is the natural exponential function, is the linear normalization function, To find the absolute value sign.

[0017] The above technical solution, by introducing the natural exponential decay relationship between the noise performance level of the data point itself and the deviation from the average value of the corresponding data points in other drilling cycles, realizes the accurate calculation of the noise performance level of the corrected data point. This not only takes into account the independent noise level of individual data points, but also effectively incorporates the data consistency correction across cycles, thereby improving the accuracy and specificity of denoising.

[0018] Preferably, the independence of the data point from the data points on both sides thereof is obtained by:

[0019] The ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its left is taken as the independence of the data point from the data points on the left; the ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its right is taken as the independence of the data point from the data points on the right.

[0020] The above technical solution quantifies the independence of a data point from the left and right data by comparing the fitting loss of a data point with the cumulative fitting loss of adjacent data points. This strategy not only finely characterizes the independence of a single data point in the sequence, but also effectively distinguishes the relationship between local data changes and the overall trend, thereby enhancing the sensitivity of outlier identification.

[0021] Preferably, all characteristic points in the feed speed data sequence are screened out according to the possibility, including:

[0022] Data fitting obtains a fitting curve of the possibility that all data points in the feed speed data sequence are feature points, wherein the abscissa of the fitting curve is the acquisition time of the data point, and the ordinate is the possibility that the data point at the acquisition time is the feature point;

[0023] Extract all peaks in the fitting curve;

[0024] Based on the peak values ​​of all the peaks, all the peaks are divided into two categories with large inter-class differences and small intra-class differences. The classification method uses clustering or maximum inter-class variance method;

[0025] The data points corresponding to the peaks with larger peak mean are taken as feature points.

[0026] The above technical solution, by fitting and analyzing the feed speed data sequence and combining precise peak point detection with classification technology, can efficiently identify feature points and significantly improve the sensitivity of data analysis.

[0027] Preferably, the noise performance degree of the data point after correction is used to estimate the data point Expected signals in the algorithm include:

[0028]

[0029] In the formula, For the Data points in The expected signal in the algorithm, For the The value of the data point, For the The noise performance after correction of the data points is is the natural exponential function, For the The value of the data point, For the The value of the data point.

[0030] The above technical solution dynamically adjusts the calculated expected signal by combining the actual value of the data point, the mean of the neighboring points and the corrected noise performance level, thereby achieving an estimation of each data point that is more in line with the actual working conditions and enhancing the adaptability of the LMS algorithm in complex environments.

[0031] Preferably, the data point and the data points on both sides thereof are centered on the data point, and w data points are taken on the left and right sides of the data point in sequence as the data points on both sides of the data point; if the number of data points on one side is less than w, the number of data points selected on the other side is the same as that on that side.

[0032] Preferably, the possibility that the data point is a feature point satisfies the following relationship:

[0033] ,in, is the possibility that the data point is a feature point, is the average value of all data points to the left of the data point, is the average value of all data points to the right of the data point, To find the absolute value sign.

[0034] The above technical solution evaluates the possibility of a data point as a feature point by calculating the absolute value of the difference between the average values ​​on the left and right sides of the data point, and can intuitively and efficiently identify data points with significant changes on both sides in the data sequence.

[0035] Preferably, extracting all peaks in the fitting curve includes:

[0036] In the fitting curve, if the first-order derivative of the possibility that a data point is a feature point is zero and the second-order derivative is less than zero, the coordinate corresponding to the data point is a peak.

[0037] Preferably, monitoring the drilling tool according to the denoised feed speed data sequence comprises:

[0038] The fluctuation range of the feed speed is set according to the historical feed speed data. When it is identified that the value of a data point in the denoised feed speed data sequence exceeds the fluctuation range, an early warning is issued; otherwise, no early warning is issued.

[0039] The above technical solution realizes real-time monitoring and anomaly identification of denoised data by setting the feed speed fluctuation range based on historical data. This method can not only timely warn of potential processing anomalies or tool wear, but also effectively filter out normal fluctuations, reduce false alarms, improve the stability of the production process and the targeted maintenance response, which is of great significance to optimizing the drilling process and ensuring product quality.

[0040] On the other hand, the present invention also provides an artificial intelligence-based drilling tool precision monitoring system, the monitoring system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the monitoring method described above.

[0041] The present invention has the following effects:

[0042] The present invention significantly enhances the timeliness and accuracy of data series analysis through intelligent recognition of feature points and precise period division, innovatively distinguishes noise from signals using local fitting loss, improves the accuracy of noise evaluation, makes the estimation of expected signals more realistic, and optimizes the denoising effect. It not only ensures the quality of feed speed data, but also improves the accuracy of drilling tool monitoring, which is beneficial to the stability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0044] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0046] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0047] Reference Figure 1 , a drilling tool accuracy monitoring method based on artificial intelligence, including steps S1-S3, specifically as follows:

[0048] S1: Collecting a feed speed data sequence, and calculating the possibility of each data point being a feature point, screening out all feature points according to the possibility, and determining a drilling cycle based on all feature points.

[0049] Drilling tool in execution The plate drilling operation shows a clear stage-by-stage speed change pattern, which is as follows:

[0050] At the initial stage of the operation, the feed speed is maintained at a preset constant value until the tool touches the board surface and starts drilling, at which point the speed will drop slightly. If a hard area inside the board is encountered (such as a dense copper foil area), the drilling speed will drop briefly. When the feed speed reaches the other side of the plate, a brief speed surge will occur, and then the feed speed will return to the initial setting value, marking the completion of a drilling cycle. It can be seen that a drilling cycle includes two instantaneous feed speed data.

[0051] This series of dynamic changes forms a unique fluctuation pattern in the feed speed data sequence, and each complete fluctuation cycle corresponds to a complete drilling process. Based on the above observations, the drilling cycle can be defined by identifying the data points that have changed significantly in the feed speed data sequence, that is, the characteristic points of the feed speed data. Every two adjacent characteristic points are a drilling cycle. These characteristic points mark the key stage transitions of the drilling process, from non-contact state to contact and drilling, through the board layer, and then back to non-contact state again, thereby revealing the periodic characteristics of the drilling operation. By systematically traversing and analyzing these characteristic points, the drilling periodic pattern can be efficiently and accurately extracted from the continuous feed speed data, providing a solid data foundation for subsequent precision monitoring and process optimization.

[0052] Among them, the method for obtaining the characteristic points in the feed speed data sequence is:

[0053] (1) According to the numerical difference between any data point in the feed speed data sequence and the data points on both sides thereof, the possibility of the data point being a feature point is calculated. The possibility is positively correlated with the numerical difference. The specific operation is as follows:

[0054] For any data point, take the data point as the center and take the data points as the data points on both sides of the data point. If there are insufficient data points on one side, , then the number of data points selected on the other side is the same as that on this side. In this embodiment Take 10, that is, take the 10 data points on the left and the 10 data points on the right of each data point as the two side data points of the data point. If there are only 5 data points on the left side of a data point, then only 5 data points are selected on the right side of the data point to obtain the two side data points of the data point. In the specific operation, a sliding window can be used to achieve the same purpose, that is, to establish a length of 1 in the feed speed data sequence. The sliding window is then used to slide through all data points. When each data point is located at the center of the sliding window, the data points on both sides of the data point are obtained.

[0055] The possibility of each data point in the feed speed data sequence being a feature point is calculated according to the following formula:

[0056]

[0057] in, is the possibility that the data point is a feature point, is the average value of all data points to the left of the data point, that is, The mean value of the data points, is the average value of all data points on the right side of the data point, that is, The mean value of the data points, To find the absolute value sign.

[0058] The possibility of each data point being a feature point is obtained by calculating the absolute value of the difference in the average values ​​of the areas on both sides of the data point. Based on the possibility of each data point being a feature point, it is possible to keenly capture data points in the feed speed data sequence that undergo significant changes, namely feature points. This calculation method is very effective in identifying key state transitions in the processing process (such as the drill bit contacting the PCB board, penetrating the board layer, etc.), which improves the sensitivity and response speed of data monitoring.

[0059] (2) According to the possibility that each data point in the feed speed data sequence is a feature point, all feature points in the feed speed data sequence are screened out, including:

[0060] A fitting curve of the possibility that all data points in the feed speed data sequence are feature points is obtained by data fitting, wherein the abscissa of the fitting curve is the acquisition time of the data point, and the ordinate is the possibility that the data point at the acquisition time is the feature point. The data fitting method may be fitted by the least square method;

[0061] The purpose of curve fitting is to identify potential feature points in the feed data sequence in the fitting curve. Feature points are often manifested as data points corresponding to the peaks in the fitting curve. The method for determining the peaks in the fitting curve is as follows: in the fitting curve, if the first-order derivative of the possibility that a data point is a feature point is zero, and the second-order derivative is less than zero, then the possibility of the data point being a feature point (the horizontal axis is the acquisition time of the data point, and the vertical axis is the possibility of the data point being a feature point) is the peak. The data point corresponding to the peak marks the moment when the feed speed changes significantly. However, larger noise points in actual data may also generate pseudo peaks. Larger noise points are random, while feature points are periodic and regular. The pseudo peaks corresponding to larger noise points are weak, which are significantly different from the peaks caused by feature points.

[0062] Therefore, in order to accurately distinguish the pseudo peaks induced by noise from the significant peaks caused by real feature points, a strategy based on maximizing the between-class variance was adopted. This strategy involves dividing all identified peaks into two categories according to their peak values, and achieving optimal classification by maximizing the variance between these two categories. In this way, the weak peaks generated by noise will be distinguished from the strong peaks generated by the real feature points, and the peaks with larger mean values ​​will be retained. The data points corresponding to these peaks are the feature points representing the real changes in the processing state.

[0063] Furthermore, the data points corresponding to a type of peak with a larger peak mean are taken as feature points for subsequent processing, thereby effectively eliminating noise interference and improving the accuracy and reliability of feature point recognition. In addition, the feature points can also be divided into two categories through a clustering algorithm to achieve the same result.

[0064] Among them, the drilling cycle is to identify the drilling periodic pattern based on the feature points in the feed speed data sequence, that is, every two adjacent, non-repeating feature points are regarded as a drilling cycle, for example: the first feature point and the second point are the first drilling cycle, and the third point and the fourth point are the second drilling cycle. In particular, if there are an odd number of points in the feature point set, then the last feature point is the beginning of the drilling cycle.

[0065] S2: Calculate the noise performance of each data point and make corrections. Estimate the noise performance of each data point based on the corrected noise performance. The expected signal in the algorithm denoises the feed speed data series based on the expected signal of all data points.

[0066] When evaluating the noise characteristics of the feed speed data sequence during the drilling process, the previous steps have eliminated the larger noise, but there may still be some other noise distributed in the feed speed data sequence. For this reason, a method is designed to measure the noise intensity of each data point based on its local data correlation, and then distinguish these noise points from feature points.

[0067] Among them, the noise level of each data point is calculated as follows:

[0068] (1) Analyze the independence of the data point and the data points on both sides:

[0069] If a data point is noise, then the data point has a low correlation with the surrounding data points in the feed speed data sequence, or even no correlation at all. Therefore, the data point and all the data points on both sides are fitted to obtain two fitting curves, namely, the fitting curve of the data point and all the data points on the left, and the fitting curve of the data point and all the data points on the right, which are respectively recorded as the left fitting curve and the right fitting curve of the data point. Regardless of the left fitting curve or the right fitting curve, the horizontal axis is the acquisition time of the data point, and the vertical axis is the value of the data point.

[0070] In the left fitting curve, the ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its left is taken as the independence of the data point from the left data points; in the right fitting curve, the ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its right is taken as the independence of the data point from the right data points; the fitting loss of each data point is equal to the absolute value of the difference between the fitting value of the data point (the ordinate value corresponding to the data point on the fitting curve) and the actual value.

[0071] In addition, if the number of data points on any side of the data point is less than or equal to 2, the side does not meet the curve fitting condition, and no curve fitting is performed, and the independence of the data point with the side is 0.

[0072] (2) The greater of the independence of the data point with the data point on the left and the independence of the data point with the data point on the right is used as the noise representation degree of the data point. For example, if the independence of the data point with the data point on the left is 0.5 and the independence of the data point with the data point on the right is 0.8, then 0.8 is used as the noise representation degree of the data point.

[0073] The noise performance degree of each data point is corrected, specifically, the noise performance degree of the data point is corrected according to the deviation value between the value of the data point and the average value of the corresponding data points in all other drilling cycles, and the corrected noise performance degree is negatively correlated with the deviation value.

[0074] The corrected noise performance satisfies the following relationship:

[0075]

[0076] In the formula, is the noise performance degree of the data point after correction, is the noise performance of the data point, is the value of the data point, is the average value of the corresponding data points of the data point in other drilling cycles, is the natural exponential function, is the linear normalization function, To find the absolute value sign.

[0077] The correction formula introduces a dynamic adjustment mechanism by combining the noise performance of the data point itself and the deviation of the data point relative to the average value of its corresponding data points in different drilling cycles. and The difference is weighted by the natural exponential function, which can be adjusted according to the specificity of each data point in the actual processing process. This means that for those data points that fluctuate greatly in a specific period but whose overall trend is consistent with other periods, the adjustment of the noise performance level will be more gentle, which improves the accuracy of identifying abnormal situations of individual data points. The corrected noise performance level can better distinguish between real signal changes and noise interference by considering the influence of the global average value, especially when the processing conditions fluctuate greatly. This adaptive adjustment helps to filter out more non-periodic noise caused by external factors, improve the denoising effect, and make subsequent monitoring of drilling tools more accurate.

[0078] Furthermore, data points outside the drilling cycle are processed. Specifically, these data points include all data before the first feature point, data between the second feature point and the third feature point, data points between the fourth feature point and the fifth feature point, and so on. They also include all data points after the last feature point when the total number of feature points is an even number.

[0079] Given that the drilling tool does not touch the PCB surface during the period corresponding to these data points outside the drilling cycle, theoretically its feed speed should remain relatively stable, so the calculation of the noise performance level at these data points should be more direct and accurate without considering the impact of changes within the cycle. Therefore, we directly perform linear normalization on the original noise performance level of these data points, that is, ,in, It is a linear normalized function and does not introduce interference from intra-cycle changes, thus ensuring the accuracy and reliability of data point analysis during non-drilling cycles.

[0080] Among them, each data point is estimated according to the corrected noise performance The expected signal in the algorithm denoises the feed speed data sequence based on the expected signal of all data points, including:

[0081] (1) Estimation of expected signal:

[0082] Usually use (least mean square) algorithm for noise removal, The core goal of the algorithm is to adjust the weights of the filter so that the output signal (i.e., the filtered signal) is as close to the expected signal (or target signal) as possible. This process is achieved by minimizing the mean square error between the output signal and the expected signal. When the algorithm has been iterated many times and the mean square error between the output signal and the expected signal is minimized, it can be considered that the output signal at this time has the best denoising effect because it is closest to the desired signal among all possible outputs.

[0083] The noise performance degree of the data point after correction is used to estimate the data point Expected signals in the algorithm include:

[0084] ;

[0085] In the formula, For the Data points in The expected signal in the algorithm, For the The value of the data point, For the The noise performance after correction of the data points is is the natural exponential function, For the The value of the data point, For the The value of the data point.

[0086] In this way, an adaptive expected signal estimation method is implemented by integrating the noise performance of the feed speed data points and the linear relationship of the local data points to optimize The denoising performance of the algorithm is specifically reflected in the following aspects: Dynamically adjust the denoising intensity. The formula dynamically adjusts the adjustment range for each data point according to the degree of noise. When the noise level is large, it means that the data point is seriously affected by the noise. At this time, the adjustment range of the data point is larger, and vice versa. In this way, personalized denoising can be performed more accurately according to the actual situation of each data point, which improves the pertinence and accuracy of denoising; maintain data continuity and authenticity, and compare the first The formula can identify the position of the data point in the local trend (concave or convex) and decide whether to adjust it upward or downward accordingly. This helps to maintain the physical meaning and trend continuity of the data series and avoid signal distortion caused by excessive denoising; combining local features with global noise levels not only considers the local linear trend of each data point, but also combines its noise performance in the entire data series, thus achieving an effective combination of local features and global noise evaluation. This comprehensive consideration method makes the expected signal The estimation of is more consistent with the physical changes in the actual processing process, which improves the rationality of the denoising process.

[0087] (2) Utilization The algorithm denoises the feed speed data sequence based on the expected signals of all data points to obtain the denoised feed speed data sequence. When processing signals such as drilling tool feed speed that may contain complex noise and dynamically changing signals, the algorithm can adjust its own parameters more flexibly, which improves the robustness of the algorithm and its adaptability to different working conditions. It helps to reduce the erroneous deletion of normal signal components while maintaining the denoising effect, optimizes the data denoising process, and effectively improves the accuracy and efficiency of drilling tool monitoring.

[0088] S3: Monitor the drilling tool according to the denoised feed speed data sequence.

[0089] The specific monitoring method includes: setting the fluctuation range of the feed speed according to the historical feed speed data, and issuing an early warning when it is identified that the value of a data point in the denoised feed speed data sequence exceeds the fluctuation range, indicating possible tool wear, unstable workpiece clamping or other processing abnormalities, otherwise, no early warning is issued.

[0090] The present invention also provides an artificial intelligence-based drilling tool precision monitoring system, the monitoring system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the monitoring method.

[0091] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0092] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A drilling tool accuracy monitoring method based on artificial intelligence, characterized in that: include: Collecting a feed speed data sequence of the drilling tool during the drilling process, and calculating the possibility that any data point in the feed speed data sequence is a feature point according to the numerical difference between the data points on both sides thereof, wherein the possibility is positively correlated with the numerical difference; According to the possibility, all characteristic points in the feed speed data sequence are screened out, and every two adjacent characteristic points are regarded as a drilling cycle; Performing data fitting on the feed speed data sequence, and taking the ratio of the fitting loss of the data point to the fitting loss of the data points on both sides thereof as the independence of the data point from the data points on both sides thereof, and taking the larger independence as the noise performance degree of the data point; According to the deviation value between the value of the data point and the average value of the corresponding data points in all other drilling cycles, the noise performance degree of the data point is corrected, and the corrected noise performance degree is negatively correlated with the deviation value; The corrected noise performance satisfies the following relationship: ； In the formula, is the noise performance degree of the data point after correction, is the noise performance of the data point, is the value of the data point, is the average value of the corresponding data points of the data point in other drilling cycles, is the natural exponential function, is the linear normalization function, To find the absolute value sign; The noise performance degree of the data point after correction is used to estimate the data point The desired signals in the algorithm include: ； In the formula, For the Data points in The expected signal in the algorithm, For the The value of the data point, For the The noise performance after the correction of the data point is is the natural exponential function, For the The value of the data point, For the The feed speed data sequence is denoised based on the expected signal, and the drilling tool is monitored according to the denoised feed speed data sequence.

2. The drilling tool accuracy monitoring method based on artificial intelligence according to claim 1 is characterized in that: The independence of the data point and the data points on both sides thereof can be obtained by: The ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its left is taken as the independence of the data point from the data points on the left; the ratio of the fitting loss of the data point to the cumulative value of the fitting losses of all the data points on its right is taken as the independence of the data point from the data points on the right.

3. The drilling tool accuracy monitoring method based on artificial intelligence according to claim 1 is characterized in that: All feature points in the feed speed data sequence are screened out according to the possibility, including: Data fitting obtains a fitting curve of the possibility that all data points in the feed speed data sequence are feature points, wherein the abscissa of the fitting curve is the acquisition time of the data point, and the ordinate is the possibility that the data point at the acquisition time is the feature point; Extract all peaks in the fitting curve; Based on the peak values ​​of all the peaks, all the peaks are divided into two categories with large inter-class differences and small intra-class differences. The classification method uses clustering or maximum inter-class variance method; The data points corresponding to the peaks with larger peak mean are taken as feature points.

4. The drilling tool accuracy monitoring method based on artificial intelligence according to claim 2 is characterized in that: The data point and the data points on both sides of it are centered on the data point, and the left and right sides of the data point are respectively taken data points as the data points on both sides of the data point; if there are insufficient data points on one side , then the number of data points selected on the other side is the same as that on this side.

5. The drilling tool precision monitoring method based on artificial intelligence according to claim 3 is characterized in that: The possibility that the data point is a feature point satisfies the following relationship: ,in, is the possibility that the data point is a feature point, is the average value of all data points to the left of the data point, is the average value of all data points to the right of the data point, To find the absolute value sign.

6. The drilling tool precision monitoring method based on artificial intelligence according to claim 3 is characterized in that: Extract all peaks in the fitting curve, including: In the fitting curve, if the first-order derivative of the possibility that a data point is a feature point is zero and the second-order derivative is less than zero, the coordinate corresponding to the data point is a peak.

7. The drilling tool accuracy monitoring method based on artificial intelligence according to claim 1 is characterized in that: The drilling tool is monitored based on the denoised feed rate data sequence, including: The fluctuation range of the feed speed is set according to the historical feed speed data. When it is identified that the value of a data point in the denoised feed speed data sequence exceeds the fluctuation range, an early warning is issued; otherwise, no early warning is issued.

8. A drilling tool precision monitoring system based on artificial intelligence, characterized in that: The monitoring system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the monitoring method according to any one of claims 1 to 7.

Images