An intelligent monitoring method for tool cutting force data in CNC boring and milling

Abstract

The present invention relates to the field of machine tool monitoring technology, specifically to an intelligent monitoring method for cutting force data of a CNC boring and milling tool, comprising the following steps: positioning a high-speed camera in the cutting area of ​​a CNC boring and milling machine, continuously capturing instantaneous images of particle splash and sparks, collecting and analyzing the distribution of particles and sparks in each frame in real time, and obtaining a cutting scene image data stream. In the present invention, by using a high-speed camera to capture images of particle splash and spark generation in the cutting area of ​​a boring and milling machine in real time, by analyzing the image data stream, the accuracy of prediction of tool wear and potential faults is improved, a deep learning network is used to optimize data processing, the particle and spark features extracted from the image provide quantitative indicators for tool wear status, the cutting force and vibration sensor data are integrated, the accuracy of abnormal point analysis is enhanced, and machine tool maintenance is more predictable. By adjusting cutting parameters in real time, the continuity and high efficiency of production are ensured, and the cost of downtime due to failures is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of machine tool monitoring, and in particular to an intelligent monitoring method for cutting force data of a CNC boring and milling tool. Background Art

[0002] The field of machine tool monitoring technology involves the use of various sensors and data processing systems to monitor the operating status of machine tools in real time, including tool wear, cutting force, machine temperature and vibration parameters, making the production process more automated and intelligent, helping to improve machining accuracy, optimize cutting parameters and extend tool life. Through real-time data analysis, machine tool failures can be prevented, downtime can be reduced, and high-efficiency and high-quality production can be achieved.

[0003] Among them, the intelligent monitoring method of tool cutting force data for CNC boring and milling focuses on using intelligent sensors on CNC boring and milling machines to collect data on tool cutting force in order to monitor and optimize the cutting process in real time. This method uses advanced data analysis technology to parse the collected cutting force data to determine the degree of tool wear, predict tool life, and optimize cutting parameters. This monitoring capability helps reduce machine tool downtime, improve production efficiency and processing quality. The method can also prevent potential production accidents by predicting tool failures, thereby ensuring the safe operation of the production line.

[0004] Although existing machine tool monitoring technologies provide basic cutting force and vibration monitoring, they are still limited in capturing small changes and early warning of faults. The sensors and data processing systems that traditional monitoring technologies rely on fail to reflect in real time when dealing with high-speed and small-change events due to data processing capabilities and resolution limitations. For example, early identification of small tool wear or microcracks is easily overlooked in traditional systems until it develops to the point where it affects production efficiency and product quality. Existing technologies have bottlenecks in data integration, and real-time analysis and application of multi-source data streams have largely failed to be achieved, which limits the scope and effect of monitoring systems in dynamically adjusting production processes, resulting in untimely machine tool maintenance and increased risk of production interruptions. At the same time, untimely identification also leads to greater maintenance costs. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an intelligent monitoring method for cutting force data of a CNC boring and milling tool.

[0006] In order to achieve the above object, the present invention adopts the following technical solution, a method for intelligently monitoring cutting force data of a CNC boring and milling tool, comprising the following steps:

[0007] S1: Position the high-speed camera in the cutting area of ​​the CNC boring and milling machine to continuously capture the instantaneous images of particle splash and sparks, collect and analyze the particle and spark distribution of each frame in real time, and obtain the cutting scene image data stream;

[0008] S2: Input the cutting scene image data stream into the deep learning network, extract the particle and spark features in the image, and obtain the preliminary cutting feature analysis results through image feature analysis;

[0009] S3: Based on the preliminary cutting feature analysis results, compare the historical data patterns to identify the changes in cutting status, evaluate the tool wear condition, identify the key changes in the cutting process, and construct a key morphological feature set;

[0010] S4: Based on the key morphological feature set, the cutting force sensor and vibration sensor data are integrated to form a time series data stream for multi-dimensional monitoring, perform time series analysis, and construct time series integrated analysis results;

[0011] S5: Perform outlier analysis on the results of time series integration analysis, use statistical models to identify and classify abnormal patterns in the data, and generate anomaly detection results by interpreting the data stream;

[0012] S6: Combine the anomaly detection results and key morphological feature sets to evaluate the real-time wear and fracture of the tool, adjust the cutting parameters and optimize the machine tool performance based on the evaluation results, and build an intelligent cutting force data control solution.

[0013] As a further solution of the present invention, the cutting site image data stream includes frame rate, image resolution and image contrast, the preliminary cutting feature analysis results include feature texture description, color change distinction and shape contrast, the key morphological feature set includes edge sharpness, morphological symmetry and color saturation, the time series integrated analysis results include frequency domain features, time window width and signal amplitude, the anomaly detection results include anomaly frequency, anomaly duration and anomaly impact range, and the intelligent cutting force data control scheme includes feed rate adjustment, cutting depth setting and tool angle modification.

[0014] As a further solution of the present invention, the steps of positioning a high-speed camera in the cutting area of ​​a CNC boring and milling machine, continuously capturing instantaneous images of particle splash and sparks, collecting and analyzing the particle and spark distribution of each frame in real time, and obtaining the cutting scene image data stream are specifically as follows:

[0015] S101: Position the high-speed camera through the cutting area of ​​the CNC boring and milling machine, aim at the cutting area, adjust the focus and exposure settings, match the shooting requirements under differentiated lighting and speed conditions, configure the image frame rate, capture particles and spark splashes, and obtain the initial image set;

[0016] S102: Based on the initial image set, identify the contours of particles and sparks in each frame, adjust the image resolution and contrast, encode and classify the collected data, form a structured data representation, and obtain particle spark feature data;

[0017] S103: According to the particle spark feature data, the information of each frame is integrated, the time series distribution is compared, a time-based dynamic image data model is constructed, data is archived and optimized for storage, and retrieval and analysis are performed to generate a cutting site image data stream.

[0018] As a further solution of the present invention, the steps of inputting the cutting scene image data stream into a deep learning network, extracting the particle and spark features in the image, and obtaining preliminary cutting feature analysis results through image feature analysis are specifically as follows:

[0019] S201: Initialize the network input interface according to the cutting scene image data stream, set the input frequency of the image data stream to 30 frames per second, adjust the input buffer capacity to accommodate continuous image data, configure the image resolution to 1,080p, adjust the clarity of the input image and the amount of data processed, and generate a stable image data stream;

[0020] S202: Based on the stable image data stream, adjust the image contrast and brightness, set a dynamic threshold, distinguish sparks from metal particles in the image, and extract key visual features in the image by adjusting the brightness distribution and color intensity to obtain image feature data;

[0021] S203: Using the image feature data, adjusting the parameter feedback mechanism, including adjusting the sensitivity and parameters of feature recognition, identifying and analyzing the frequency and position of particles and sparks in the image, and generating preliminary cutting feature analysis results.

[0022] As a further solution of the present invention, based on the preliminary cutting feature analysis results, the changes in cutting state are identified by comparing historical data patterns, tool wear conditions are evaluated, key changes in the cutting process are identified, and the steps of constructing a key morphological feature set are specifically as follows:

[0023] S301: based on the preliminary cutting feature analysis results, data points that are different from historical data are screened, the data points represent potential abnormal cutting state and tool wear, the data points are marked respectively, and after classification, data points with large data deviation are extracted to obtain a set of changed data points;

[0024] S302: According to the set of change data points, data associated with each tool use cycle is collected, the performance degradation rate of the tool is compared, time series analysis is performed on the data, the degree of tool wear is calibrated, and a wear degree index is obtained;

[0025] S303: Analyze the key changes in the cutting process using the wear index, combine the cutting parameters and wear data, construct the morphological features that affect the cutting quality, record and integrate the morphological features, and generate a key morphological feature set.

[0026] As a further solution of the present invention, based on the key morphological feature set, the data of the cutting force sensor and the vibration sensor are integrated to form a time series data stream for multi-dimensional monitoring, and time series analysis is performed to construct the time series integrated analysis results in the following steps:

[0027] S401: Based on the data of the key morphological feature set, the data collection frequency of the cutting force and vibration sensors is set, the data receiving speed and the operation of reducing data loss are optimized, the buffer configuration is adjusted, and the monitoring data stream is generated;

[0028] S402: Perform timestamp processing on the monitoring data stream, adjust the data stream structure, determine the input requirements for time series analysis, including time stamping each data packet, and obtain a time-synchronized data stream;

[0029] S403: Perform data segment analysis on the time-synchronized data stream, apply the weighted moving average method, set the time window and sliding window technology for data processing and maintain the continuity of the data, adjust the window coverage frequency, optimize the data coverage and analysis accuracy, and construct the time series integrated analysis results.

[0030] As a further solution of the present invention, the formula of the weighted moving average method is as follows:

[0031] ;

[0032] Where t represents the predicted value at time, Xt-i+1 is the real-time observation value at time t-i+1, Wi is the weight, f(i,λ,θ,δ) is the adjustment function, i is the time index, λ is the attenuation factor, θ is the trend adjustment parameter, and δ is the seasonal adjustment coefficient.

[0033] As a further solution of the present invention, the steps of performing anomaly point analysis on the time series integrated analysis results, using statistical models to identify and classify abnormal patterns in the data, and generating anomaly detection results by interpreting the data stream are specifically as follows:

[0034] S501: Based on the results of the time series integrated analysis, screening is performed to identify data points that deviate from the normal data pattern. The data points are recorded through time marking and abnormal intensity evaluation, including quantification of the degree of data deviation and preliminary classification of abnormal points, to obtain a preliminary abnormal point set;

[0035] S502: Based on the preliminary abnormal point set, historical data comparison is performed for each abnormal data point, by evaluating the difference between the historical behavior and the current performance of each point, recording the type and characteristics of each difference, and performing data comparison and type classification to obtain abnormal classification details;

[0036] S503: Using the anomaly classification details, interpret the anomaly patterns in the data stream, reconstruct the data relationship and archive the results according to the timestamps and change characteristics of the anomaly data points, and generate an anomaly detection result set.

[0037] As a further solution of the present invention, the steps of combining the abnormal detection results and the key morphological feature set to evaluate the real-time wear and fracture of the tool, adjusting the cutting parameters and optimizing the machine tool performance according to the evaluation results, and constructing an intelligent cutting force data control solution are as follows:

[0038] S601: Based on the abnormality detection result, configure the data acquisition unit, synchronize the abnormality detection result and morphological feature data of the tool monitoring sensor, adjust the acquisition frequency and synchronization mechanism and match the data stream speed output by the monitoring sensor, optimize the data transmission protocol, and generate a real-time tool status data stream;

[0039] S602: extracting wear and fracture indicators from the real-time tool status data stream, adjusting the cutting parameters of the machine tool based on the indicators, including cutting depth and feed speed, and dynamically adjusting the parameters by using the real-time feedback adjustment control unit to obtain adjusted and optimized cutting parameters;

[0040] S603: Apply the adjusted and optimized cutting parameters and use the support vector machine algorithm to reconfigure the machine tool's operating settings, including adjusting the spindle speed and tool load, monitor the adjusted machine tool operating data, determine the effectiveness of the parameter adjustment, and build an intelligent cutting force data control solution through machine tool performance debugging.

[0041] As a further solution of the present invention, the formula of the support vector machine algorithm is as follows:

[0042] y=w·x+b+α·p+β·q+γ·r

[0043] Among them, y is the predicted target value, w is the weight vector, x is the input feature vector, b is the bias term, α, β, γ are adjustment coefficients, p is the temperature change rate of the machine tool, q is the material hardness, and r is the tool life prediction value. Compared with the prior art, the advantages and positive effects of the present invention are:

[0044] In the present invention, a high-speed camera is used to capture images of particle splashing and spark generation in the cutting area of ​​the boring and milling machine in real time, thereby providing a fine-grained data monitoring method, which not only enhances the monitoring capability of minute changes in the cutting process, but also identifies the cutting state by analyzing the image data stream, thereby greatly improving the prediction accuracy of tool wear and potential failures. A deep learning network is used to further optimize the data processing flow, so that the particle and spark features extracted from the image can provide direct quantitative indicators for the tool wear state. The cutting force and vibration sensor data are integrated to construct a multi-dimensional monitoring information time series, which enhances the accuracy of abnormal point analysis, making the maintenance of the machine tool more predictable and less invasive. Through comprehensive data analysis, the cutting parameters of the machine tool can be adjusted in real time to ensure the continuity and high efficiency of the production process, while reducing the cost caused by downtime due to failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the main steps of the present invention;

[0046] Figure 2 It is a schematic diagram of the refinement of S1 of the present invention;

[0047] Figure 3 It is a schematic diagram of the refinement of S2 of the present invention;

[0048] Figure 4 It is a schematic diagram of the refinement of S3 of the present invention;

[0049] Figure 5 It is a schematic diagram of the refinement of S4 of the present invention;

[0050] Figure 6 It is a detailed schematic diagram of S5 of the present invention;

[0051] Figure 7 It is a detailed schematic diagram of S6 of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0053] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0054] See also Figure 1 The present invention provides a technical solution, a method for intelligently monitoring cutting force data of a CNC boring and milling tool, comprising the following steps:

[0055] S1: Position the high-speed camera in the cutting area of ​​the CNC boring and milling machine to continuously capture the instantaneous images of particle splash and sparks, collect and analyze the particle and spark distribution of each frame in real time, and obtain the cutting scene image data stream;

[0056] S2: Input the cutting scene image data stream into the deep learning network, extract the particle and spark features in the image, and obtain the preliminary cutting feature analysis results through image feature analysis;

[0057] S3: Based on the preliminary cutting feature analysis results, compare the historical data patterns to identify the changes in cutting status, evaluate the tool wear condition, identify the key changes in the cutting process, and construct a key morphological feature set;

[0058] S4: Based on the key morphological feature set, the cutting force sensor and vibration sensor data are integrated to form a time series data stream for multi-dimensional monitoring, perform time series analysis, and construct time series integrated analysis results;

[0059] S5: Perform outlier analysis on the results of time series integration analysis, use statistical models to identify and classify abnormal patterns in the data, and generate anomaly detection results by interpreting the data stream;

[0060] S6: Combine the anomaly detection results and key morphological feature sets to evaluate the real-time wear and fracture of the tool, adjust the cutting parameters and optimize the machine tool performance based on the evaluation results, and build an intelligent cutting force data control solution.

[0061] The cutting site image data stream includes frame rate, image resolution and image contrast. The preliminary cutting feature analysis results include feature texture description, color change distinction and shape contrast. The key morphological feature set includes edge sharpness, morphological symmetry and color saturation. The time series integrated analysis results include frequency domain characteristics, time window width and signal amplitude. The anomaly detection results include anomaly frequency, anomaly duration and anomaly impact range. The intelligent cutting force data control solution includes feed rate adjustment, cutting depth setting and tool angle modification.

[0062] Specifically, if Figure 2 As shown in the figure, by positioning the high-speed camera in the cutting area of ​​the CNC boring and milling machine, the instantaneous images of particle splash and sparks are continuously captured, and the particle and spark distribution of each frame is collected and analyzed in real time. The specific steps for obtaining the cutting scene image data stream are as follows:

[0063] S101: Position the high-speed camera through the cutting area of ​​the CNC boring and milling machine, aim at the cutting area, adjust the focus and exposure settings, match the shooting requirements under the conditions of differentiated lighting and speed, configure the image frame rate, capture particles and spark splashes, and obtain the execution process of the initial image set as follows;

[0064] After the CNC boring and milling machine locates the cutting area through the high-speed camera, the operator needs to ensure that the focus of the camera is adjusted to the clearest position of the cutting area, and fine-tune the exposure settings to ensure that the image can be effectively captured under different lighting and speed conditions. The configuration of the image frame rate is the key. The optimal frame rate must be determined according to the cutting speed and the luminous properties of the material to achieve effective capture of spatter particles and sparks, and then generate an initial image set. The exposure time is adjusted using the following formula: ;

[0065] Among them, T represents the exposure time (seconds), v represents the cutting speed (m / s), L represents the light intensity (lumens), and K is an adjustment coefficient determined by experiments to balance the brightness and clarity of the image. The application of the formula ensures that high-quality images can be obtained under different working conditions.

[0066] S102: Based on the initial image set, identify the outlines of particles and sparks in each frame, adjust the image resolution and contrast, encode and classify the collected data, form a structured data representation, and obtain the particle spark feature data. The execution process is as follows;

[0067] Each frame in the initial image set is analyzed, and the resolution and contrast are adjusted using image processing software to highlight the outlines of particles and sparks. The image data is structured using coding technology to classify various types of sparks and particles and convert the information into usable feature data. The feature data is presented in the form of a structured data table, which is easy to process and analyze later. The following formula is used to process this data:

[0068] F = a·Pavg+b·σ

[0069] Among them, F represents the feature intensity, Pavg represents the average pixel intensity, σ represents the standard deviation of the pixel intensity, and a and b are weight factors obtained through data fitting. The formula helps optimize the feature discrimination and improve the accuracy of data analysis.

[0070] S103: According to the particle spark feature data, the information of each frame is integrated, the time series distribution is compared, a dynamic image data model of the time reference is constructed, data archiving and optimized storage are performed, and retrieval and analysis are performed to generate the cutting scene image data stream. The execution process is as follows;

[0071] By integrating the particle spark feature data in each frame and combining it with the time series data, a dynamic image data model is constructed. This model supports continuous monitoring of the cutting site, realizes real-time archiving and optimized storage of data, and uses time series analysis to model the time-dependent dynamics using the following formula:

[0072] ;

[0073] Among them, M(t) represents the model value at time, Fi(t) represents the feature data at time, and wi is the weight of the feature, which determines the influence of different features on the overall model.

[0074] Specifically, if Figure 3 As shown in the figure, the cutting scene image data stream is input into the deep learning network, the particle and spark features in the image are extracted, and the preliminary cutting feature analysis results are obtained through image feature analysis. The specific steps are as follows:

[0075] S201: Initialize the network input interface according to the cutting scene image data stream, set the input frequency of the image data stream to 30 frames per second, adjust the input buffer capacity to accommodate continuous image data, configure the image resolution to 1,080p, adjust the clarity of the input image and the amount of data processed, and generate a stable image data stream. The execution process is as follows;

[0076] In order to process the continuous image data acquired by the high-speed camera, the network input interface is first initialized, and the input frequency of the image data stream is set to thirty frames per second to adapt to the rapid changes in high-speed cutting operations. The capacity of the input buffer is adjusted to ensure that the image data can be stored continuously and uninterruptedly to avoid data loss in high-frequency operations. The image resolution is configured to 1080p to ensure that the details of the image can be clearly captured to meet subsequent processing requirements. The clarity of the input image is adjusted to the optimal level, and the amount of data processed is adjusted accordingly to generate a stable image data stream.

[0077] S202: Based on the stable image data stream, adjust the image contrast and brightness, set the dynamic threshold, distinguish the sparks and metal particles in the image, and extract the key visual features in the image by adjusting the brightness distribution and color intensity to obtain the image feature data. The execution process is as follows;

[0078] For the image data stream stably received through the network input interface, the contrast and brightness of the image are first adjusted to enhance the recognizability of tiny sparks and metal particles in the image. A dynamic threshold is set to automatically adjust the key elements in the image according to the changes in the image content. The adjustment of brightness distribution and color intensity is achieved through the following formula:

[0079] D(v)=a·vk+b

[0080] Among them, D(v) represents the adjusted pixel value, v represents the original pixel value, a and b are coefficients for adjusting brightness and contrast, and k is an index for adjusting image contrast. By fine-tuning the values ​​of a, b, and k, the key visual features in the image can be effectively extracted to obtain more accurate image feature data.

[0081] S203: Using the image feature data, adjusting the parameter feedback mechanism, including adjusting the sensitivity and parameters of feature recognition, identifying and analyzing the frequency and position of particles and sparks in the image, and generating preliminary cutting feature analysis results. The execution process is as follows;

[0082] Using the image feature data extracted from the stable image data stream, the parameter feedback mechanism is further adjusted to enhance the accuracy of feature recognition, including adjusting the sensitivity and parameters of feature recognition to adapt to different image features. The recognition parameters are adjusted by the following formula:

[0083] F(x,y)=α·f(x,y)+β·g(x,y)

[0084] Among them, F(x,y) represents the adjusted eigenvalue, f(x,y) represents the original eigenvalue extracted from the image, g(x,y) represents the influence of background noise, α and β are weight factors used to balance the influence of image features and background noise. By adjusting α and β, the frequency of particles and sparks and their positions in the image can be more accurately identified and analyzed, thereby generating more accurate preliminary cutting feature analysis results.

[0085] Specifically, if Figure 4 As shown in the figure, based on the preliminary cutting feature analysis results, the changes in cutting state are identified by comparing historical data patterns, tool wear conditions are evaluated, key changes in the cutting process are identified, and the steps for constructing a key morphological feature set are as follows:

[0086] S301: Based on the preliminary cutting feature analysis results, data points that are different from historical data are screened. The data points represent potential abnormal cutting conditions and tool wear. The data points are marked respectively, and data points with large data deviations are extracted after classification. The execution process of obtaining the changed data point set is as follows;

[0087] By conducting an in-depth analysis of the initial cutting feature analysis results, we first screen out data points that are significantly different from historical data. These data points indicate abnormal cutting conditions or tool wear. In order to effectively perform such screening, we use the following formula to calculate data deviation:

[0088] ΔP=│Pcurrent-Phistorical│

[0089] Among them, ΔP represents the deviation between the current data point and the historical data, Pcurrent is the value of the current data point, and Phistorical is the value of the corresponding historical data point. By setting a deviation threshold, only data points exceeding this threshold are regarded as significant change points and are marked and classified. The marked data points are then grouped into a set of change data points to provide a basis for the next step of analysis.

[0090] S302: According to the set of change data points, data associated with each tool use cycle is collected, the tool performance degradation rate is compared, time series analysis is performed on the data, and the degree of tool wear is calibrated. The execution process of obtaining the wear degree index is as follows;

[0091] According to the set of changing data points, detailed data associated with each tool use cycle is collected. This data includes tool use time, cutting material type, cutting parameters, etc. By comparing the performance degradation rate of the tool in different cycles, the data is analyzed in time series, and the tool wear degree is calibrated using the following formula:

[0092] W(t)=k·exp(-rt)

[0093] Where W(t) represents the degree of wear at time t, k and r are the initial value and attenuation rate of wear, respectively. This formula helps evaluate the degree of wear of the tool at different stages of use, thereby obtaining an accurate wear degree index, which will be used for further cutting quality analysis.

[0094] S303: Analyze the key changes in the cutting process using the wear degree index, combine the cutting parameters and wear data, construct the morphological features that affect the cutting quality, record and integrate the morphological features, and generate the key morphological feature set. The execution process is as follows;

[0095] Using the wear index obtained, we deeply analyze the key changes in the cutting process, and combine the real-time monitored cutting parameters and the previously obtained wear data to construct the morphological features that affect the cutting quality. The recorded and integrated morphological features will be quantified by the following formula:

[0096] ;

[0097] Among them, Q represents the comprehensive quantitative index of morphological characteristics, Ti is the morphological characteristic value corresponding to the i-th cutting parameter, and ai is the corresponding weight coefficient, which reflects the contribution of different cutting parameters to the morphological characteristics. This formula generates a set of key morphological characteristics by systematically integrating multiple morphological characteristics, which provides data support for optimizing the cutting process.

[0098] Specifically, if Figure 5 As shown in the figure, based on the key morphological feature set, the cutting force sensor and vibration sensor data are integrated to form a time series data stream for multi-dimensional monitoring, and the time series analysis is performed to construct the time series integrated analysis results. The specific steps are as follows:

[0099] S401: Based on the data of the key morphological feature set, the data collection frequency of the cutting force and vibration sensors is set, the data receiving speed and the operation of reducing data loss are optimized, the buffer configuration is adjusted, and the execution process of generating the monitoring data stream is as follows;

[0100] Based on the data of the key morphological feature set, the data collection frequency of the cutting force and vibration sensors is first set. The setting takes into account the technical characteristics of the sensors and the dynamic changes of the cutting process to ensure the real-time and accuracy of data collection. In order to optimize the data reception speed and reduce the probability of data loss, a systematic communication and storage configuration adjustment is performed. The specific operation includes adjusting the network bandwidth and buffer capacity to adapt to high-frequency data transmission. The buffer configuration is determined by the following formula:

[0101] ;

[0102] Among them, C represents the buffer size (bit), F represents the data collection frequency (Hz), D represents the size of a single data point (bit), T represents the target data storage time (seconds), and S represents the maximum supported speed of the buffer (bits / second). By adjusting the parameters, we can ensure that the buffer can handle the input data stream without overflowing, thereby generating a stable and continuous monitoring data stream, providing a high-quality data source for subsequent analysis.

[0103] S402: Perform timestamp processing on the monitoring data stream, adjust the data stream structure, determine the input requirements for time series analysis, including time stamping each data packet, and obtain the execution process of the time synchronization data stream as follows;

[0104] Perform timestamp processing on the monitoring data stream to ensure the timing accuracy of the data, including adding a time stamp to each data packet to ensure that subsequent data analysis can be performed based on accurate time information. Data stream structure adjustment involves programming and configuration to adapt to different analysis tools and storage systems. The input requirements of time series analysis accurately define the data structure, including the format and precision of the data type timestamp. The following formula is used to optimize the processing and storage of timestamps:

[0105] Ts=To+n·Δt

[0106] Among them, Ts represents the timestamp of each data point, To is the initial timestamp, n is the serial number of the data point, and Δt is the data collection interval, which helps the system record the time of each data point at a predetermined frequency, ensure the consistency and accuracy of the time-synchronized data stream, and provide a solid time foundation for in-depth analysis and decision support.

[0107] S403: Perform data segment analysis on the time-synchronized data stream, apply the weighted moving average method, set the time window and sliding window technology for data processing and maintain the continuity of the data, adjust the window coverage frequency, optimize the data coverage and analysis accuracy, and construct the execution process of the time series integrated analysis results as follows;

[0108] Perform data segment analysis on time-synchronized data streams, including applying weighted moving average to smooth data and reduce the impact of random fluctuations, setting time windows and sliding window technology for data processing, which not only maintains data continuity but also optimizes data coverage and analysis accuracy. By adjusting the coverage frequency of the sliding window, the data refresh speed and analysis response time can be flexibly controlled, improving the flexibility and efficiency of overall data processing, and constructing time series integrated analysis results.

[0109] The formula for the weighted moving average method is as follows:

[0110] ;

[0111] Where yt represents the predicted value at time t, xt-i+1 is the real-time observation value at time t-i+1, Wi is the weight, f(i,λ,θ,δ) is the adjustment function, i is the time index, λ is the attenuation factor, θ is the trend adjustment parameter, and δ is the seasonal adjustment coefficient.

[0112] The execution process is as follows:

[0113] Determine the time window size n, set the weight wi for each time point, collect the actual observations at the corresponding time point t-i+1, and design an innovative adjustment function f(i,λ,θ,δ), which adjusts the contribution of each observation by referring to the time index i, the attenuation factor λ, the trend adjustment parameter θ, and the seasonal adjustment coefficient δ. Then, the weighted average is calculated by combining the weight wi and the adjusted observation xt-i+1·f(i,λ,θ,δ) to generate the predicted or calculated value yt at time t. In the step of confirming the weight coefficient, historical data analysis is applied to determine the optimal value of each parameter to ensure the accuracy of the calculation.

[0114] Specifically, if Figure 6 As shown in the figure, the steps of performing outlier analysis on the time series integrated analysis results, using statistical models to identify and classify abnormal patterns in the data, and generating anomaly detection results by interpreting the data stream are as follows:

[0115] S501: Based on the results of the time series integrated analysis, screening is performed to identify data points that deviate from the normal data pattern. The data points are recorded through time marking and abnormal intensity evaluation, including quantification of the degree of data deviation and preliminary classification of abnormal points. The execution process of obtaining a preliminary abnormal point set is as follows;

[0116] Based on the results of time series integrated analysis, data point screening is performed, including points that deviate significantly from the normal data pattern. Each data point is tracked by its time tag, and then its abnormal intensity is evaluated. The evaluation includes quantifying the degree of data deviation and preliminarily classifying the abnormal points. The quantification of the degree of data deviation is performed using the following formula:

[0117] ;

[0118] Where D represents the degree of data deviation, x is the value of the current data point, µ is the average value of the normal data pattern, and σ is the standard deviation. The anomaly level is divided according to the size of its value. The data point is then classified into a preliminary anomaly point set, laying the foundation for further analysis and processing.

[0119] S502: Based on the preliminary abnormal point set, historical data comparison is performed for each abnormal data point, by evaluating the difference between the historical behavior and the current performance of each point, recording the type and characteristics of each difference, and performing data comparison and type classification, the execution process of obtaining abnormal classification details is as follows;

[0120] For each data point in the preliminary anomaly point set, historical data comparison is performed to evaluate the difference between the historical behavior and the current performance of each point. The steps further refine the classification of anomalies by recording the type and characteristics of each difference in detail, and performing data comparison and type classification. The comparative analysis relies on the following formula:

[0121] ;

[0122] Among them, ΔH represents the degree of difference between the current data point and its historical data, xcurrent,i and xhistorical,i represent the data values ​​at the current and historical moments respectively, providing a method to quantify the difference between historical and current behaviors, helping to determine the specific type and intensity of the anomaly, thereby obtaining more accurate anomaly classification details.

[0123] S503: Using the anomaly classification details, interpret the anomaly pattern in the data stream, reconstruct the data relationship and archive the results according to the timestamp and change characteristics of the abnormal data points, and generate the execution process of the anomaly detection result set as follows;

[0124] Using the obtained anomaly classification details, the abnormal patterns in the data stream are interpreted, including reconstruction of data relationships and archiving of results based on the timestamps and change characteristics of the abnormal data points, detailed analysis of the abnormal patterns, and integration and archiving of data through the following formula:

[0125] ;

[0126] Among them, R represents the reconstructed anomaly detection result, fj is the processing function for the jth type of anomaly, ΔHj is the degree of difference of this type of anomaly, and wj is the weight factor, which reflects the importance of different types of anomalies in the overall anomaly assessment, helps to systematically interpret and process abnormal patterns, and generates the final anomaly detection result set by reconstructing data relationships, providing a scientific basis for subsequent decision-making.

[0127] Specifically, if Figure 7 As shown in the figure, the real-time wear and fracture of the tool are evaluated by combining the abnormal detection results and the key morphological feature set. The cutting parameters are adjusted and the machine tool performance is optimized according to the evaluation results. The specific steps for building an intelligent cutting force data control solution are as follows:

[0128] S601: Based on the abnormality detection result, configure the data acquisition unit, synchronize the abnormality detection result and morphological feature data of the tool monitoring sensor, adjust the acquisition frequency and synchronization mechanism and match the data stream speed output by the monitoring sensor, optimize the data transmission protocol, and generate the real-time tool status data stream. The execution process is as follows;

[0129] Based on the anomaly detection results, configure the data acquisition unit, synchronize the anomaly detection results and morphological feature data of the tool monitoring sensor, determine the operation interface of the data acquisition unit, customize specific input ports for data acquisition, ensure that the anomaly detection results can be smoothly input from the sensor, set the data synchronization threshold through the program interface, ensure that the morphological feature data and the anomaly detection results can be synchronized and updated in the same time frame, design the data flow management logic, match the data flow speed output by the monitoring sensor, and optimize the data transmission protocol. The formula is:

[0130] ;

[0131] Among them, Fs represents the final synchronization frequency, xi represents the acquisition frequency of the i-th data point, n is the total number of data points, σ ​​is the standard deviation of the data flow speed, and kp is the weight coefficient, which is used to adjust the deviation between the synchronization frequency and the data flow speed. By adjusting the value, the synchronization efficiency can be optimized, data delay and loss can be reduced, and a real-time tool status data stream can be generated.

[0132] S602: extracting wear and fracture indicators from the real-time tool status data stream, adjusting the cutting parameters of the machine tool based on the indicators, including cutting depth and feed speed, and dynamically adjusting the parameters using the real-time feedback adjustment control unit to obtain the adjusted and optimized cutting parameters. The execution process is as follows;

[0133] Wear and fracture indicators are extracted from the real-time tool status data stream, and the cutting parameters of the machine tool are adjusted based on the indicators. The tool wear index is analyzed by statistical analysis methods, and the standard deviation and mean are calculated to provide real-time data support for the machine tool control unit. Based on the wear indicator, a logical judgment threshold is set. If the indicator exceeds the preset range, the parameter adjustment is triggered. The real-time feedback is used to adjust the control unit to dynamically adjust the cutting depth and feed speed. The calculation formula is as follows:

[0134] Popt=α×Pprev+β×Dnew+γ×Snew;

[0135] Among them, Popt is the optimized cutting parameter, Pprev represents the previous parameter setting, Dnew and Snew are the new cutting depth and feed speed respectively, α, β, γ are adjustment weights to balance the influence of the old parameters and the new feedback indicators to obtain the adjusted and optimized cutting parameters.

[0136] S603: Apply the adjusted and optimized cutting parameters, use the support vector machine algorithm, reconfigure the machine tool's operating settings, including adjusting the spindle speed and tool load, monitor the adjusted machine tool operation data, determine the effectiveness of the parameter adjustment, and build an intelligent cutting force data control solution through machine tool performance debugging. The execution process is as follows;

[0137] Apply the adjusted and optimized cutting parameters, use numerical simulation methods to reconfigure the machine tool's operating settings, adjust the spindle speed and tool load force in the process, and determine the best parameter configuration through sensitivity analysis to ensure the best performance of the adjusted machine tool. Through real-time monitoring and adjustment, ensure that the spindle speed and load force match, optimize tool performance and durability, monitor the adjusted machine tool operation data, determine the effectiveness of parameter adjustment, and build an intelligent cutting force data control solution through machine tool performance debugging.

[0138] The formula of the support vector machine algorithm is as follows:

[0139] y = w·x+b+α·p+β·q+γ·r;

[0140] Among them, y is the predicted target value, w is the weight vector, x is the input feature vector, b is the bias term, α, β, γ are adjustment coefficients, p is the machine temperature change rate, q is the material hardness, and r is the tool life prediction value. The execution process is as follows:

[0141] It is necessary to determine the weight vector w and the bias term b. By using new features including machine tool temperature, material hardness and predicted tool life, the new features p, q, r are integrated with historical cutting parameters and performance indicators as input vectors. By adjusting the coefficients α, β, γ, the coefficients are optimized on the training set through the cross-validation method to improve the prediction accuracy. Finally, the optimal spindle speed and tool load force are calculated. The machine tool temperature change rate p, material hardness q, and tool life prediction value r are newly introduced parameters. They are combined with the original cutting force, speed and other parameters to improve the adaptability and accuracy of the model in actual operation. By including actual influencing factors, the model can more comprehensively predict the optimal settings of the machine tool.

[0142] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for intelligently monitoring cutting force data of a CNC boring and milling tool, characterized in that: The following steps are involved: By positioning the high-speed camera in the cutting area of ​​the CNC boring and milling machine, the instantaneous images of particle splash and sparks are continuously captured, the particle and spark distribution of each frame is collected and analyzed in real time, and the image data stream of the cutting site is obtained; Input the cutting scene image data stream into the deep learning network, extract the particle and spark features in the image, and obtain preliminary cutting feature analysis results through image feature analysis; Based on the preliminary cutting feature analysis results, compare historical data patterns to identify changes in cutting conditions, evaluate tool wear conditions, identify key changes in the cutting process, and construct a key morphological feature set; Based on the key morphological feature set, the cutting force sensor and vibration sensor data are integrated to form a time series data stream for multi-dimensional monitoring, and time series analysis is performed to construct a time series integrated analysis result; Performing outlier analysis on the time series integrated analysis results, using statistical models to identify and classify abnormal patterns in the data, and generating anomaly detection results by interpreting the data stream; Combine the abnormal detection results and the key morphological feature set to evaluate the real-time wear and fracture of the tool, adjust the cutting parameters and optimize the machine tool performance according to the evaluation results, and build an intelligent cutting force data control solution; The cutting scene image data stream includes frame rate, image resolution and image contrast, the preliminary cutting feature analysis results include feature texture description, color change distinction and shape contrast, the key morphological feature set includes edge sharpness, morphological symmetry and color saturation, the time series integrated analysis results include frequency domain features, time window width and signal amplitude, the anomaly detection results include anomaly frequency, anomaly duration and anomaly impact range, and the intelligent cutting force data control scheme includes feed rate adjustment, cutting depth setting and tool angle modification; The steps of positioning the high-speed camera in the cutting area of ​​the CNC boring and milling machine, continuously capturing the instantaneous images of particle splash and sparks, collecting and analyzing the particle and spark distribution of each frame in real time, and obtaining the cutting scene image data stream are as follows: Position the high-speed camera through the cutting area of ​​the CNC boring and milling machine, aim at the cutting area, adjust the focus and exposure settings, match the shooting requirements under differentiated lighting and speed conditions, configure the image frame rate, capture particles and spark splashes, and obtain the initial image set; Based on the initial image set, the outlines of particles and sparks in each frame are identified, the image resolution and contrast are adjusted, the collected data are encoded and classified to form a structured data representation, and particle spark feature data is obtained; According to the particle spark characteristic data, the information of each frame is integrated, the time series distribution is compared, a dynamic image data model of the time reference is constructed, data archiving and optimized storage are performed, and retrieval and analysis are performed to generate a cutting site image data stream; The steps of inputting the cutting scene image data stream into the deep learning network, extracting the particle and spark features in the image, and obtaining the preliminary cutting feature analysis results through image feature analysis are as follows: Initialize the network input interface according to the cutting scene image data stream, set the input frequency of the image data stream to 30 frames per second, adjust the input buffer capacity to accommodate continuous image data, configure the image resolution to 1,080 p, adjust the clarity of the input image and the amount of data processed, and generate a stable image data stream; Based on the stable image data stream, adjusting image contrast and brightness, setting a dynamic threshold, distinguishing sparks from metal particles in the image, and extracting key visual features in the image by adjusting brightness distribution and color intensity to obtain image feature data; Using the image feature data, adjusting the parameter feedback mechanism, including adjusting the sensitivity and parameters of feature recognition, identifying and analyzing the frequency and position of particles and sparks in the image, and generating preliminary cutting feature analysis results; Based on the preliminary cutting feature analysis results, the steps of comparing historical data patterns to identify changes in cutting conditions, assess tool wear conditions, identify key changes in the cutting process, and construct a key morphological feature set are as follows: Based on the preliminary cutting feature analysis results, data points that are different from historical data are screened, the data points represent potential abnormal cutting state and tool wear, the data points are marked respectively, and after classification, data points with large data deviations are extracted to obtain a set of changed data points; According to the set of change data points, data associated with each tool use cycle is collected, the performance degradation rate of the tool is compared, time series analysis is performed on the data, the degree of tool wear is calibrated, and a wear degree index is obtained; Utilizing the wear degree index, analyzing key changes in the cutting process, combining cutting parameters and wear data, constructing morphological features that affect cutting quality, recording and integrating the morphological features, and generating a key morphological feature set; Based on the key morphological feature set, the data of cutting force sensors and vibration sensors are integrated to form a time series data stream for multi-dimensional monitoring, and time series analysis is performed to construct the time series integrated analysis results. Specifically, the steps are as follows: Based on the data of the key morphological feature set, the data collection frequency of the cutting force and vibration sensors is set, the data receiving speed and the operation of reducing data loss are optimized, the buffer configuration is adjusted, and the monitoring data stream is generated; Performing timestamp processing on the monitoring data stream, adjusting the data stream structure, and determining the input requirements for time series analysis, including time stamping each data packet, to obtain a time-synchronized data stream; Perform data segment analysis on the time-synchronized data stream, apply weighted moving average method, set time window and sliding window technology for data processing and maintain data continuity, adjust window coverage frequency, optimize data coverage and analysis accuracy, and construct time series integrated analysis results; The steps of performing anomaly analysis on the time series integrated analysis results, using statistical models to identify and classify abnormal patterns in the data, and generating anomaly detection results by interpreting the data stream are as follows: Based on the time series integrated analysis results, screening is performed to identify data points that deviate from normal data patterns, and the data points are recorded through time marking and abnormal intensity assessment, including quantification of the degree of data deviation and preliminary classification of abnormal points, to obtain a preliminary abnormal point set; Based on the preliminary abnormal point set, historical data comparison is performed for each abnormal data point, by evaluating the difference between the historical behavior and the current performance of each point, recording the type and characteristics of each difference, and performing data comparison and type classification to obtain abnormal classification details; Using the anomaly classification details, interpret the anomaly patterns in the data stream, reconstruct the data relationships and archive the results according to the timestamps and change characteristics of the anomaly data points, and generate an anomaly detection result set; Combining the anomaly detection results and the key morphological feature set, evaluating the real-time wear and fracture of the tool, adjusting the cutting parameters and optimizing the machine tool performance according to the evaluation results, and constructing an intelligent cutting force data control solution are as follows: Based on the abnormality detection result, a data acquisition unit is configured to synchronize the abnormality detection result and morphological feature data of the tool monitoring sensor, adjust the acquisition frequency and synchronization mechanism and match the data stream speed output by the monitoring sensor, optimize the data transmission protocol, and generate a real-time tool status data stream; Extracting wear and fracture indicators from the real-time tool status data stream, adjusting cutting parameters of the machine tool based on the indicators, including cutting depth and feed speed, and dynamically adjusting the parameters using a real-time feedback control unit to obtain adjusted and optimized cutting parameters; Applying the adjusted and optimized cutting parameters, the support vector machine algorithm is used to reconfigure the operating settings of the machine tool, including adjusting the spindle speed and tool load force, monitoring the adjusted machine tool operation data, determining the effectiveness of the parameter adjustment, and building an intelligent cutting force data control solution through machine tool performance debugging.

2. The intelligent monitoring method for cutting force data of CNC boring and milling tools according to claim 1 is characterized in that: The formula of the weighted moving average method is as follows: ； Among them, y t represents the predicted value at time t, X t-i+1 is the real-time observation value at time t-i+1, W i is the weight, f(i,λ,θ,δ) is the adjustment function, i is the time index, λ is the attenuation factor, θ is the trend adjustment parameter, and δ is the seasonal adjustment coefficient.

3. The intelligent monitoring method for cutting force data of CNC boring and milling tools according to claim 2 is characterized in that: The formula of the support vector machine algorithm is as follows: y=w·x+b+α·p+β·q+γ·r Among them, y is the predicted target value, w is the weight vector, x is the input feature vector, b is the bias term, α, β, γ are adjustment coefficients, p is the temperature change rate of the machine tool, q is the material hardness, and r is the predicted value of tool life.

Images