Composite tool cutting monitoring system and monitoring method thereof

Abstract

The present application discloses a composite cutter cutting monitoring system and a monitoring method thereof. The composite cutter cutting monitoring system is used for a machine tool. The cutting monitoring system includes a data acquisition module, a database and a cutting control module. The data acquisition module is used to acquire motor current data and a cutter wear data of the machine tool. The motor current data is used as training data of a cutter wear state prediction model for deep learning and prediction. The database is used to establish a cutter wear database for comparing a cutter wear state. The cutter wear state prediction model outputs a cutter wear state prediction data to the cutting control module. The cutting control module determines whether the cutter wear state is normal according to the cutter wear state prediction data.

Description

Technical Field

[0001] This invention relates to a composite cutting tool, and more particularly to a composite cutting tool cutting monitoring system and monitoring method thereof. Background Technology

[0002] When cutting with composite material tools, the tool temperature can rise rapidly, leading to overheating. Therefore, monitoring the tool's cutting condition to maintain stability is essential for achieving high-efficiency and automated cutting. However, high-efficiency cutting is not simply about increasing the spindle speed; it requires parameter adjustments based on the tool's condition (e.g., severe or normal wear) to prevent reduced tool life from impacting machine operation. Current research on tool life focuses primarily on composite tool material development, tool geometry improvement, and cutting force enhancement, neglecting improvements in composite tool machining optimization and cutting monitoring. Summary of the Invention

[0003] This invention provides a composite tool cutting monitoring system and method, which can use machine learning to predict the relationship between tool state and motor current. The trained prediction model can provide optimized parameters for the cutting control module to optimize the tool wear state, so as to achieve the purpose of intelligent cutting monitoring.

[0004] According to one aspect of the present invention, a composite tool cutting monitoring system is proposed for a machine tool. The cutting monitoring system includes a data acquisition module, a database, and a cutting control module. The data acquisition module acquires motor current data and tool wear data of the machine tool. The motor current data serves as training data for a tool wear state prediction model, enabling deep learning and prediction. The database is used to establish a tool wear database for comparing tool wear states. The tool wear state prediction model outputs tool wear state prediction data to the cutting control module. The cutting control module determines whether the tool wear state is normal based on the tool wear state prediction data.

[0005] According to one aspect of the present invention, a method for monitoring the cutting of composite cutting tools is proposed for a machine tool. The cutting monitoring method includes the following steps: Acquiring motor current data and measuring tool wear data of the machine tool; using the motor current data as training data for a tool wear state prediction model, performing deep learning and prediction; establishing a tool wear database for comparing tool wear states; outputting tool wear state prediction data generated by the tool wear state prediction model to a cutting control module; and determining whether the tool wear state is normal based on the tool wear state prediction data.

[0006] To provide a better understanding of the above and other aspects of the present invention, specific embodiments are described below in conjunction with the accompanying drawings: Attached Figure Description

[0007] Figure 1 This is a schematic diagram of an intelligent machining system for monitoring the cutting of composite tools according to an embodiment of the present invention;

[0008] Figure 2 This is a tool status data diagram;

[0009] Figure 3A , Figure 3B These are experimental data graphs from the tool wear database;

[0010] Figure 4 This is a schematic diagram of the tool wear condition;

[0011] Figure 5 A graph showing experimental data for different tool states used in variance analysis;

[0012] Figure 6 A graph showing the relationship between tool wear and cutting length;

[0013] Figure 7 A schematic diagram of a processing parameter optimization management platform;

[0014] Figure 8 This is a flowchart of a composite tool cutting monitoring method according to an embodiment of the present invention;

[0015] Figure 9A and Figure 9B Training graph for the prediction module;

[0016] Figure 9C and Figure 9D For the training graph of the prediction module; and

[0017] Figure 10 This is a schematic diagram of a machine deep learning model for tool wear.

[0018] Symbol Explanation

[0019] 100: Intelligent Manufacturing System

[0020] 102: Machine

[0021] 104: Controller

[0022] 105: Motor current data

[0023] 106: Sensors

[0024] 107: Tool Wear Data

[0025] 108: Processing parameters

[0026] 109: Tool Status Data

[0027] 110: Cutting Monitoring System

[0028] 112: Data Acquisition Module

[0029] 114: Database

[0030] 116: Tool Wear Condition Prediction Model

[0031] 118: Cutting Control Module Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description uses the same / similar symbols to denote the same / similar elements.

[0033] Please refer to Figure 1 , Figure 1 A schematic diagram of an intelligent machining system 100 for monitoring the cutting of composite tools according to an embodiment of the present invention is shown.

[0034] The intelligent machining system 100 includes a machine tool 102, a controller 104, multiple sensors 106, and a cutting monitoring system 110. The machine tool 102 is, for example, an industrial automation device such as a lathe, CNC machine tool, or robotic arm. It has a spindle motor and a composite cutting tool. The spindle motor drives the tool to rotate, feed, or cut. The current of the spindle motor can be provided by the controller 104 or an external power supply. The motor machining parameters, such as the current, speed, and feed rate of the spindle motor, can be detected by the sensors 106. Furthermore, the controller 104 can precisely control the speed and feed rate of the spindle motor via computer numerical control, and then precisely control the tool rotation, feed, or cutting via the spindle motor shaft to achieve precision control. Additionally, the controller 104 can set machining parameters 108, such as the speed and feed rate of the spindle motor / tool, as a reference for subsequent adjustments to the machining parameters 108.

[0035] In this embodiment, to achieve high-efficiency and automated cutting, the composite tool cutting monitoring system 110 can be used to monitor the tool cutting status and maintain a stable tool state. For example... Figure 1As shown, the composite tool cutting monitoring system 110 includes a data acquisition module 112, a database 114, a tool wear state prediction model 116, and a cutting control module 118. The data acquisition module 112 is used to acquire motor current data 105 and measure tool wear data 107 from the machine tool 102. For example, the data acquisition module 112 can exchange data within the machine tool 102 or between machines via the OPC UA (OPC Unified Architecture) communication protocol to collect information such as motor current data 105, tool wear data 107, and tool status data 109. OPC Unified Architecture is an open platform communication protocol that supports multi-threading and is not limited by operating systems or programming languages, making it suitable for information integration in industrial automation equipment.

[0036] In one embodiment, the tool wear data 107 is measured, for example, by observing the metallographic structure of the material and the tool wear state using a high-magnification optical microscope to calculate the tool surface roughness. Different tools have different processing conditions and wear. For example, tool types can be divided into straight cutters, spiral cutters, etc. Tool processing conditions can include parameters such as tool cutting speed, tool feed rate, and tool depth of cut. Tool wear states can be divided into initial wear, normal wear, and rapid wear. Generally speaking, the time from the start of cutting (initial wear) to the predetermined cutting time is called tool life. Tool life is generally quantified by the amount of tool wear within the normal wear range. If it exceeds the normal wear range, it is considered rapid wear. The occurrence of a certain phenomenon can also be used as a basis for judging rapid wear, such as vibration exacerbation, deterioration of the machined surface roughness, poor chip breaking, and chipping. When the tool reaches the end of its tool life, the cutting tip becomes blunt and is constantly subjected to large forces, high temperatures, and intense friction, resulting in wear. When the wear reaches a certain level, the tool can no longer perform its original cutting quality and can no longer be used. Therefore, tools should be used in a way that avoids rapid wear.

[0037] Generally speaking, when cutting with composite material tools, the temperature of the composite tool can easily increase rapidly, causing overheating. Therefore, it is necessary to strengthen the monitoring of the tool's cutting condition to maintain the tool's condition within the normal wear range. If the tool enters the rapid wear stage, it is necessary to change the tool's machining conditions (including parameters such as tool cutting speed, tool feed rate, and tool depth of cut) to reduce tool wear and extend tool life.

[0038] Furthermore, increasing the cutting speed can shorten the workpiece machining time and improve the machining surface accuracy, but it will also increase the tool wear rate, meaning the tool life will be shortened. The relationship between cutting speed and tool life is roughly constant. That is, the faster the cutting speed, the shorter the tool life; conversely, the slower the cutting speed, the longer the tool life. Therefore, the cutting control module 118 of this system 110 can increase tool life by reducing the cutting speed, but it can also increase tool life by changing parameters such as the tool feed rate or the depth of cut.

[0039] While increasing the cutting speed can optimize the surface roughness of composite cutting tools and improve workpiece machining quality, simply increasing the cutting speed can actually shorten tool life. Furthermore, poor tool condition can easily lead to overheating during cutting, and failure to promptly correct machining parameters 108 when the tool is in poor condition (such as entering a period of rapid wear) further reduces tool life. Therefore, in this embodiment, the cutting monitoring system 110 can perform machine learning and prediction on the relationship between tool wear state and motor current, and the trained prediction model 116 can provide optimized parameters for the cutting control module 118 to optimize the tool wear state, thereby improving tool life.

[0040] Please refer to Figures 2 to 4 ,in Figure 2 Tool status data 109, Figure 3A , Figure 3B The experimental data are from the tool wear data library 107. Figure 4 This refers to the tool wear condition. The parameter optimization process is as follows. First, step 1 is to obtain tool condition data 109 (e.g., ...). Figure 2 The data shown includes tool type, tool machining conditions, etc. Next, step 2 involves creating a tool wear data library (which can be generated from...). Figure 3A and Figure 3B The experimental data is compared to find the optimized parameters. Then, step 3 connects the cutting control module 118 to the machine tool 102 and transmits the optimized parameters to the controller 104 of the machine tool 102 for correction of the machining parameters 108.

[0041] In step 2, parameter optimization, for example, involves using variance analysis to identify a suitable machining parameter 108 (e.g., tool cutting speed, tool feed rate, and depth of cut) from database 114. Variance analysis uses a contribution index to represent the proportion of variation of a control factor to the total variation, where the contribution index represents the influence of variation on a control factor. For example, if the influence of tool cutting speed on tool wear is greater than the influence of other control factors, then the contribution index of tool cutting speed is greater than the contribution index of other control factors.

[0042] Contribution indicators, for example, rely on the F-distribution as the basis for probability distribution. The F-value is estimated using the sum of squares and degrees of freedom of the experimental data to calculate the between-group and within-group mean squares. The denominator of the F-value represents the variance estimated directly from the original experimental data, while the numerator represents the variance estimated from the original experimental data using the "sample mean." The denominator and numerator of the F-value represent two different methods for evaluating the variance in the same sample space. A large F-value indicates that the control factor is an influential parameter. Therefore, the F-value can be used as a contribution indicator for parameter optimization.

[0043] Please refer to Figure 5 and Figure 6 ,in Figure 5 Experimental data for different tool states 109 used for variance analysis. Figure 6 This is a graph showing the relationship between tool wear and cutting length. Figure 6 It can be seen that when the cutting speed decreases from 120m / min to 93.75m / min, under the same feed rate (e.g., 0.4 or 0.52mm / rev), the cutting length increases or the tool wear decreases. Under these machining conditions, the variation of the cutting speed has a greater impact on tool wear than the variation of other control factors. In other words, the cutting speed contributes more to tool wear. In addition, in Test 1 and Test 2, when the cutting speed was kept constant (e.g., 120 m / min), the feed rate was increased from 0.4 mm / rev to 0.52 mm / rev. Under these machining conditions, tool wear increased with the increase of cutting length. Alternatively, in Test 3 and Test 4, when the cutting speed was kept constant (e.g., 93.75 m / min), the feed rate was increased from 0.4 mm / rev to 0.52 mm / rev. Under these machining conditions, tool wear increased with the increase of cutting length, but compared to Test 1 and Test 2, the variation in feed rate had a limited (not significant) impact on tool wear, meaning the contribution of the feed rate to tool wear was small.

[0044] Please refer to Figure 7 The machining parameter 108 optimization management platform includes a cutting control module 118 connected to the machine tool 102, which transmits the optimized tool cutting parameters to the controller 104 for machining parameter 108 correction. After receiving the optimized tool cutting parameters, the controller 104 can adjust the current machining parameters 108 of the machine tool 102, including spindle speed, feed rate, tool type, and tool machining conditions. For details, please refer to [reference needed]. Figure 7These will not be elaborated upon here. The processing parameter 108 optimization management platform can be remotely controlled, and the optimized processing parameters 108 can be input into this management platform.

[0045] Please refer to Figure 8 , Figure 8 A flowchart illustrating a composite tool cutting monitoring method according to an embodiment of the present invention is shown. In this embodiment, using... Figure 1 The composite tool cutting monitoring system 110 performs the following cutting monitoring method, with the following steps: In step S110, motor current data 105 of the machine tool 102 and tool wear data 107 are acquired. In step S120, the motor current data 105 and tool wear data 107 are used as training data for a tool wear state prediction model 116, and deep learning and prediction are performed. The trained prediction model 116 can be stored in the application software for use by the machine tool 102. In step S130, tool wear state prediction data generated by the tool wear state prediction model 116 is output to a cutting control module 118. The tool wear state can be divided into initial wear, normal wear, and rapid wear. In step S140, the cutting control module 118 determines whether the tool wear state is normal (e.g., whether the machining parameters 108 need to be optimized). When the tool wear is within the normal wear range, the process proceeds to step S150, where the cutting control module 118 controls the machine tool 102 to perform machining with the current tool cutting parameters (no optimization required). When the tool wear is within the rapid wear range, the process proceeds to step S160, where the cutting control module 118 optimizes the current tool cutting parameters of the machine tool 102 and transmits the optimized tool cutting parameters to the controller 104 of the machine tool 102.

[0046] The tool wear state prediction model 116 is, for example, a Long Short Term Memory Network (LSTM) prediction model. In this system 110, the acquired motor current data 105 is used for machine learning and tool wear state prediction. Since the motor current data 105 has both temporal and continuous characteristics, and the Recurrent Neural Network (RNN) is an algorithm specifically designed to solve time series problems and has time memory capabilities, this system 110 can use the LSTM prediction model as a recurrent neural network to structure the connection between each input and output data. The input data includes, for example, motor current data 105 and tool wear data 107, and the output data includes, for example, the tool wear state. This system uses, for example, the root mean square (RMS) value of the motor current data 105 as training data input into the LSTM prediction model 116 to improve prediction accuracy.

[0047] Additionally, please refer to Figure 9A and Figure 9B The prediction module training graph, in which Figure 9A and Figure 9B Standardizing the acquired motor current data 105 significantly improves the training and prediction performance of the LSTM model, thereby increasing prediction accuracy. Conversely, without standardization, the LSTM model performs poorly in training and prediction, failing to improve prediction accuracy. In the figure, the LSTM train line is used to train accuracy, while the LSTM valid line is used to validate accuracy after training. A good training process requires validation accuracy to be close to the training accuracy.

[0048] Furthermore, to address the overfitting issue in LSTM models, this system incorporates a dropout parameter into the recurrent neural network. Dropout means that during each training iteration, a portion of neurons are randomly ignored. Specifically, the priming effect of neurons on downstream neurons is ignored during forward propagation, and their weights are not updated during backpropagation. Each neuron's neighbors rely on features derived from their neighbors' behaviors; excessive reliance on these features leads to overfitting. By randomly removing neurons each time, the remaining neurons need to compensate for the missing neurons' functionality, transforming the network into a collection of many independent networks (different solutions to the same problem). The effect of dropout is that the network becomes less sensitive to changes in the weights of individual neurons, increasing generalization ability and reducing overfitting.

[0049] Please refer to Figure 9C and Figure 9D The prediction module training graph, in which Figure 9C and Figure 9D Adding the Dropout parameter to the LSTM model prevents overfitting and can improve the training and prediction performance, thereby increasing prediction accuracy. In the diagram, the LSTM_Dropout train line represents the training accuracy, while the LSTM_Dropout valid line represents the validation accuracy, used with validation data after training. A good training process is considered successful if the validation accuracy is close to the training accuracy.

[0050] During the training of the prediction module, current data from different time points of the motor can be input into the prediction model 116. Each set of motor current data 105 is a time series data set. For example, there are 24 or more sets of motor current data 105 used as training data. After machine deep learning, the corresponding tool wear state (level) can be obtained. Tool wear states include, for example, initial wear (level 0), normal wear (level 1), and rapid wear (level 2). Please refer to... Figure 10The machine deep learning model takes motor current data 105 as input data, the first training layer of the model is, for example, LSTM, the second training layer is, for example, dropout model, the third training layer is, for example, LSTM, and the final output layer is, for example, tool wear status (level).

[0051] According to the composite tool cutting monitoring system and method described in the above embodiments, this system can use machine learning to predict the relationship between tool state and motor current. The trained prediction model can provide optimized parameters for the cutting control module to optimize the tool wear state, so as to achieve the purpose of intelligent cutting monitoring.

[0052] In summary, although the present invention has been disclosed in conjunction with the above embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A composite material cutting tool monitoring system for use on a machine tool, the cutting monitoring system comprising: The data acquisition module is used to acquire the motor current data and tool wear data of the machine tool. The motor current data is used as training data for the tool wear state prediction model for deep learning and prediction. A database is used to establish a tool wear database for comparison of tool wear status; as well as The cutting control module outputs tool wear state prediction data from the tool wear state prediction model to the cutting control module. The cutting control module then determines whether the tool wear state is normal based on the predicted tool wear state data. When the tool wear is within the range of rapid wear, the cutting control module optimizes the current tool cutting parameters and transmits the optimized tool cutting parameters to the machine tool. The parameter optimization includes: (1) obtaining tool status data, comparing the tool wear database to find the optimized parameters, and transmitting the optimized parameters to the machine tool for machining parameter correction; (2) Find suitable machining parameters for correcting tool wear from the tool wear database through variance analysis; and (3) Using the contribution index to represent the proportion of the variation of the control factor to the total variation, find the control factor with the largest contribution and use it as the machining parameter to correct the tool wear.

2. The composite tool cutting monitoring system as described in claim 1, wherein the tool wear state includes initial wear, normal wear, and rapid wear.

3. The composite tool cutting monitoring system as described in claim 1, wherein when the tool wear condition is within the normal wear range, the cutting control module controls the machine tool to perform machining with the current tool cutting parameters.

4. A method for monitoring the cutting of composite cutting tools, used in a machine tool, the method comprising: Collect the motor current data and measure the tool wear data of the machine tool; The motor current data was used as training data for the tool wear state prediction model to perform deep learning and prediction. Establish a tool wear database for comparison of tool wear status; Output the tool wear state prediction data generated by the tool wear state prediction model to the cutting control module; and The cutting control module determines whether the tool wear condition is normal based on the tool wear condition prediction data. When the tool wear is within the range of rapid wear, the cutting control module optimizes the current tool cutting parameters and transmits the optimized tool cutting parameters to the machine tool. The parameter optimization includes: (1) obtaining tool status data, comparing the tool wear database to find the optimized parameters, and transmitting the optimized parameters to the machine tool for machining parameter correction; (2) Find suitable machining parameters for correcting tool wear from the tool wear database through variance analysis; and (3) Using the contribution index to represent the proportion of the variation of the control factor to the total variation, find the control factor with the largest contribution and use it as the machining parameter to correct the tool wear.

5. The composite tool cutting monitoring method as described in claim 4, wherein the tool wear state includes initial wear, normal wear, and rapid wear.

6. The composite tool cutting monitoring method as described in claim 4, wherein when the tool wear condition is within the normal wear range, the cutting control module controls the machine tool to perform machining with the current tool cutting parameters.

Images