A data-driven based machine tool feed axis cutting force monitoring method
By constructing a dataset through monitoring the servo signals of the machine tool feed axes and employing a time series prediction model, the problems of high cost and large error in feed axis cutting force monitoring are solved, achieving low-cost and accurate online monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
- Filing Date
- 2023-12-26
- Publication Date
- 2026-06-02
AI Technical Summary
Existing methods for monitoring feed axis cutting force are costly, have large errors, and are not suitable for online applications. Traditional benchtop force gauges affect workpiece clamping and machine tool workspace.
By monitoring servo signals such as the output torque or three-phase current and motion status of the machine tool feed axis motor, a servo monitoring signal-cutting force dataset is constructed, and online monitoring is performed using a data-driven time series single-step or sequence prediction model.
It enables low-cost, non-disruptive, and space-saving online monitoring of feed axis cutting forces, improving the accuracy and applicability of monitoring.
Smart Images
Figure CN117773652B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of CNC machine tool technology, and in particular to a CNC machine tool machining status monitoring technology that uses servo signals of the machine tool feed axes to monitor the cutting force of the feed axes. Specifically, it is a data-driven method for monitoring the cutting force of the machine tool feed axes. Background Technology
[0002] Monitoring the machining status during machine tool processing is crucial for ensuring machining quality, improving machining efficiency, and identifying equipment malfunctions. Cutting forces arise from the interaction between the tool and the workpiece, while feed axis cutting forces reflect the component of the cutting force along the feed axis. Monitoring feed axis cutting forces provides direct and effective information for applications such as identifying tool breakage, predicting tool wear, and optimizing feed rates online. Traditional methods for monitoring machine tool feed axis cutting forces based on benchtop force gauges are costly, restrict workpiece size, increase clamping difficulty, and introduce cables into the machine tool workspace, making them unsuitable for actual production applications. Machine tool servo monitoring signals contain cutting force information as an external disturbance, allowing for online prediction of cutting forces. This solution is low-cost, does not affect the machine tool workspace, and has significant practical value.
[0003] Chinese Patent 2022105841099, "A Dynamic Cutting Force Monitoring Method and System in the Cutting Process," discloses a method for online prediction of feed axis cutting force based on the three-phase current and root mean square (RMS) of the spindle motor. However, the current signal of the spindle motor is difficult to accurately reflect the load condition of the feed axis. Chinese Patent 2021107781399, "A Data-Driven Self-Monitoring Method for Cutting Force of CNC Machine Tools," authorizes a method for online prediction of feed axis cutting force based on feed axis motor current and motion state. This method defines the servo monitoring signal-cutting force prediction problem as a nonlinear autoregressive modeling problem with a specified time delay order. On the one hand, the time delay order needs to be manually specified through trial and error; on the other hand, the single-step training iterative prediction mode is prone to accumulating errors over long periods, leading to prediction failure. Therefore, there is an urgent need for a method that can accurately monitor feed axis cutting force over a long period at low cost. Summary of the Invention
[0004] The purpose of this invention is to address the problems of long cycle time, high cost, and large error in existing feed axis cutting force monitoring methods by proposing a data-driven machine tool feed axis cutting force monitoring method. It monitors servo signals such as the output torque (or three-phase current) of the machine tool feed axis motor and the feed axis motion state, simultaneously monitoring the cutting force in the corresponding feed axis direction. A feed axis servo monitoring signal-cutting force dataset is constructed through cutting experiments, and then a data-driven time series single-step prediction model is trained using a sequence-sequence supervised training method. In practical applications, the feed axis servo monitoring signals are input into the prediction model for single-step real-time prediction or sequence prediction, thereby achieving online monitoring of the feed axis cutting force.
[0005] The technical solution of this invention is:
[0006] A data-driven method for monitoring the cutting force of a machine tool feed axis, characterized by comprising the following steps:
[0007] 1) Set process parameters such as rotational speed, depth of cut, and width of cut, and continuously change the actual speed of the specified feed axis within a certain motion trajectory to carry out cutting experiments. During the cutting process, synchronously collect servo signals such as the output torque (or three-phase current) of the feed axis motor, the motion state of the feed axis, and the cutting force in the direction of the feed axis to obtain a cutting experiment sample in the form of a time series. The feed axis servo signal is the characteristic signal, and the cutting force in the direction of the feed axis is the label.
[0008] 2) Select different combinations of process parameters such as rotational speed, depth of cut and width of cut, and use the method in step 1) to obtain a set of cutting test samples, and construct a servo monitoring signal-cutting force dataset containing different combinations of process parameters and different changes in the actual speed of the selected feed axis.
[0009] 3) The model adopts a data-driven time series prediction model, which is trained to a single-step prediction model of feed axis servo monitoring signal and cutting force in a sequence-sequence supervised manner;
[0010] 4) When applying, the feed axis servo monitoring signal is input into the prediction model. The model starts from the zero initial state and performs single-step real-time prediction or sequence-sequence prediction of arbitrary length for the cutting force in the selected feed axis direction. The internal state of the model after the previous prediction should be used as the initial state for the next prediction.
[0011] It should be noted that the feed axis cutting force mentioned here refers to the component of the cutting force in the direction of the machine tool feed axis movement, which can be measured using, but is not limited to, benchtop or rotary cutting force measuring instruments.
[0012] Step 1) describes continuously changing the actual speed of a specified feed axis within a certain trajectory. This is achieved by adjusting, but not limited to, the feed per tooth or the tool trajectory direction, so that the actual speed of the feed axis changes continuously after interpolation. Preferably, this includes continuous changes in speed reversal. Furthermore, the trajectory progresses from the point where the tool is not in contact with the material at all, to the point where it enters the material and the commanded speed starts from zero, and to the point where the tool completely exits the material and the commanded speed decreases to zero.
[0013] Step 1) involves monitoring the output torque (or three-phase current) of the machine tool feed axis motor. If monitoring conditions permit, the output torque should be used preferentially. Alternatively, the torque current can be monitored instead of the output torque (or three-phase current). The root mean square (RMS) of the three-phase current can also be used instead of the torque current, but the RMS current needs to be assigned a corresponding positive or negative sign based on the direction of the feed axis speed.
[0014] The feed axis motion state described in step 1) should at least include the speed at the motor end or the table end, or both can be monitored simultaneously. Furthermore, the acceleration at the motor end and the table end can be monitored, and as many motion state monitoring quantities as possible should be used as a priority.
[0015] The servo signal mentioned in step 1) should preferably be obtained directly from the CNC system. If it cannot be read from the CNC system, it can be obtained by using external Hall sensors to monitor (three-phase current of the feed axis motor), external encoder signals (motor end motion status), and external grating ruler signals (worktable end motion status).
[0016] Step 2) The selection of different combinations of process parameters such as rotation speed, cutting depth and cutting width should be sampled according to their actual application range and the degree of dispersion of values.
[0017] The data-driven time series prediction model described in step 3) is recommended, but not limited to, neural network-based time series prediction models, such as Long Short-Term Memory (LSTM) networks and Gated Recurrent Unit (GRU) networks.
[0018] Step 3) involves inputting the feed axis servo monitoring signal into the prediction module. Monitoring should begin when no cutting has occurred and the feed rate is zero.
[0019] In addition, the sampling frequency of the feed axis servo monitoring signal and the cutting force should be consistent, and should be at least twice the maximum cutting frequency in the predicted requirements, with a recommended range of 5 to 10 times.
[0020] In summary, the key feature of this invention is the monitoring of servo signals such as the output torque (or three-phase current) of the machine tool feed axis motor and the motion state of the feed axis, synchronously monitoring the cutting force in the corresponding feed axis direction, constructing a feed axis servo monitoring signal-cutting force dataset through cutting experiments, and then training a time series single-step prediction model using a sequence-sequence supervised method. In practical applications, the corresponding feed axis servo monitoring signal is input into the prediction model for single-step real-time prediction or sequence prediction, thereby achieving online monitoring of the feed axis cutting force.
[0021] The beneficial effects of this invention are:
[0022] The method disclosed in this invention is lower in cost, does not affect workpiece clamping, and does not interfere with the machine tool workspace compared to traditional monitoring methods based on benchtop force gauges, making it more suitable for application in actual production processes. Attached Figure Description
[0023] Figure 1 This is a structural diagram of the GRU model of the present invention.
[0024] Figure 2 This is a schematic diagram of the circular hole milling experiment of the present invention.
[0025] Figure 3 The Y-axis cutting force prediction result is shown in the embodiment of the present invention when the sampling frequency of the cutting edge is 57.5Hz and the sampling frequency of the cutting edge is 500Hz.
[0026] Figure 4 The Y-axis cutting force prediction result is shown in the embodiment of the present invention when the sampling frequency is 72.5Hz and the cutting edge frequency is 500Hz. Detailed Implementation
[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0028] like Figures 1-4 As shown.
[0029] A data-driven method for monitoring cutting force on machine tool feed axes includes the following steps:
[0030] 1) Set process parameters such as rotational speed, depth of cut, and width of cut, and continuously change the actual speed of the specified feed axis within a certain motion trajectory to carry out cutting experiments. During the cutting process, synchronously collect servo signals such as the output torque (or three-phase current) of the feed axis motor, the motion state of the feed axis, and the cutting force in the direction of the feed axis to obtain a cutting experiment sample in the form of a time series. The feed axis servo signal is the characteristic signal, and the cutting force in the direction of the feed axis is the label.
[0031] 2) Select different combinations of process parameters such as rotational speed, depth of cut and width of cut, and use the method in step 1) to obtain a set of cutting test samples, and construct a servo monitoring signal-cutting force dataset containing different combinations of process parameters and different changes in the actual speed of the selected feed axis.
[0032] 3) The model adopts a data-driven time series prediction model, which is trained to a single-step prediction model of feed axis servo monitoring signal and cutting force in a sequence-sequence supervised manner;
[0033] 4) When applying, the feed axis servo monitoring signal is input into the prediction model. The model starts from the zero initial state and performs single-step real-time prediction or sequence-sequence prediction of arbitrary length for the cutting force in the selected feed axis direction. The internal state of the model after the previous prediction should be used as the initial state for the next prediction.
[0034] Details are as follows:
[0035] This implementation uses a classic neural network model for time series forecasting—the Gate Recurrent Unit (GRU) network—to construct a dynamic prediction model for feed axis servo monitoring signals and cutting forces, and validates it using experimental data from circular trajectory milling. The GRU network is suitable for adaptive learning with long short-term memory and is a powerful time series forecasting model.
[0036] Figure 1 The diagram shows the structure of a GRU network. The GRU network has two gate structures, with the reset gate and update gate from left to right. The reset gate formula is as follows:
[0037] r t =σ(W ir x t +b ir +W hr h t-1 +b hr )
[0038] Where, x t Given the input at time t, h t-1 Let W be the hidden state at time t-1. ir and W hr For the corresponding weight matrix, b ir and b hr Let σ be the corresponding bias vector, and r be the sigmoid function. t To reset the output of the gate.
[0039] The updated gate formula is as follows:
[0040] z t =σ(W iz x t +biz +W hz h t-1 +b hz )
[0041] Among them, W iz and W hz For the corresponding weight matrix, b iz and b hz For the corresponding bias vector, z t To update the output of the gate.
[0042] To generate the hidden state at time t, the candidate hidden state at time t is first calculated.
[0043]
[0044] Among them, W ih and W hh For the corresponding weight matrix, b ih and b hh Let be the corresponding bias vector, and ⊙ denote element-wise multiplication, i.e., the Hadamard product. Further, after obtaining the update gate output z... t and candidate hidden state Then, the hidden state at time t can be calculated using the following formula:
[0045]
[0046] Ultimately, the GRU network outputs y t By h t After transformation by a linear fully connected layer, the following is obtained:
[0047] y t =W l h t +b l
[0048] Among them, W l For the corresponding weight matrix, b l This is the bias vector.
[0049] In this invention, there are 8 main modes for inputting x, as follows:
[0050] x = [T] e ω] T
[0051]
[0052] x = [T] e v] T
[0053] x = [T] e va]T
[0054] x = [T] e ω v] T
[0055]
[0056] x = [T] e ω va] T
[0057]
[0058] Among them, T e ω represents the output torque of the feed axis servo motor, and ω represents the motor speed. Let v be the angular acceleration of the motor, a be the velocity at the end of the worktable, and α be the acceleration at the end of the worktable. The output quantity is the spindle cutting torque T. m T e It can be replaced by the three-phase current of the feed axis servo motor, or by the torque current. Alternatively, the root mean square of the three-phase current can be used to further replace the torque current, but the root mean square current needs to be assigned the corresponding positive and negative values according to the forward and reverse rotation of the motor.
[0059] In this implementation case, the second input mode is used, where T e Torque current was used as a substitute. A variable-speed hole milling experiment was conducted on a DMG DMU80P machining center to acquire data, monitoring the Y-axis servo signal and cutting force at a sampling frequency of 500Hz. The experimental scenario is as follows. Figure 2 As shown. Since circular toolpaths can provide typical time-varying complex cutting conditions, a hole with a diameter of 82 mm was designed for milling. The cutter diameter is 20 mm, much smaller than the hole diameter, so there are multiple circular toolpaths running from the inside out at each layer during milling. For all circular toolpaths, the spindle speed was selected from [500, 600, ..., 1800] rpm. The cutter used was a three-flute end mill, so the corresponding cutting edge frequency was [25, 30, ..., 90] Hz. Training was performed using the above data, and testing was conducted using cutting edge frequencies of 57.5 Hz and 72.5 Hz.
[0060] The GRU unit has a 3D input, a 1D output, and a 256-dimensional hidden layer; it is followed by a 256-dimensional to 1-dimensional linear fully connected layer. The initial state is set to a zero vector. Training is sequence-to-sequence, with a set of feed axis servo monitoring signals and motion state signals under different parameters input at a time. The label data is the corresponding feed axis cutting force sequence. Tests were conducted using data at cutting edge frequencies of 57.5Hz and 72.5Hz, and the results are as follows... Figure 3 and Figure 4 As shown in the figure, the method disclosed in this invention achieves good prediction results.
[0061] The parts not covered in this invention are the same as those in the prior art and are implemented using existing technologies.
Claims
1. A data-driven method for monitoring the cutting force of a machine tool feed axis, characterized in that, Includes the following steps: 1) Set process parameters: including rotational speed, depth of cut, and width of cut. Continuously change the actual speed of the specified feed axis within a certain motion trajectory to conduct a cutting experiment. During the cutting process, synchronously monitor the feed axis servo monitoring signal and the cutting force in the feed axis direction. The feed axis servo monitoring signal includes the output torque or three-phase current signal of the feed axis motor and the feed axis motion state signal. Obtain a cutting experiment sample in the form of a time series, where the feed axis servo monitoring signal is the characteristic signal and the cutting force in the feed axis direction is the label. 2) Select different combinations of process parameters such as rotation speed, depth of cut and width of cut, and use the method in step 1) to obtain a set of cutting test samples, and construct a feed axis servo monitoring signal-cutting force dataset containing different combinations of process parameters and different changes in the actual speed of the selected feed axis; 3) The model adopts a data-driven time series prediction model, which is trained to a single-step prediction model of feed axis servo monitoring signal and cutting force in a sequence-sequence supervised manner; 4) Input the feed axis servo monitoring signal into the prediction model. The model starts from zero initial state and performs single-step real-time prediction or arbitrary length sequence-sequence prediction of the cutting force in the selected feed axis direction. The internal state of the model after the previous prediction should be used as the initial state of the next prediction.
2. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: The cutting force in the feed axis direction refers to the component of the cutting force in the direction of the machine tool's feed axis movement, and should be measured using a benchtop or rotary cutting force measuring instrument.
3. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 1) describes continuously changing the actual speed of a specified feed axis within a certain motion trajectory. This is achieved by adjusting, but not limited to, the feed per tooth or the tool trajectory direction, so that the actual speed of the feed axis changes continuously after interpolation. Furthermore, the trajectory changes from when the tool is not in contact with the material at all to when it enters the material and the commanded speed starts from zero, and ends when the tool completely cuts out of the material and the commanded speed decreases to zero.
4. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 1) Monitor the output torque or three-phase current of the feed axis motor. If the monitoring conditions permit, use the output torque or monitor its torque current to replace the output torque or three-phase current. The root mean square of the three-phase current is used to replace the torque current, but the root mean square of the current needs to be assigned the corresponding positive and negative values according to the direction of the feed axis speed.
5. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: The feed axis motion state described in step 1) should at least include the speed at the motor end or the table end, or monitor both simultaneously, or further increase the monitoring of the acceleration at the motor end and the table end, and adopt multiple motion state monitoring quantities.
6. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 1) The feed axis servo monitoring signal is obtained directly from the CNC system; for cases where it cannot be read from the CNC system, external Hall sensor monitoring, external encoder signal connection, and external grating ruler signal connection are used to obtain the signal.
7. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 2) The selection of different combinations of process parameters such as rotation speed, cutting depth and cutting width should be sampled according to their actual application range and the degree of dispersion of the values.
8. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 3) The data-driven time series prediction model adopts a neural network-based time series prediction model, which includes a gated recurrent unit (GRU) network.
9. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: Step 4) involves inputting the feed axis servo monitoring signal into the prediction model. Monitoring should begin when no cutting occurs and the feed rate is zero.
10. The data-driven machine tool feed axis cutting force monitoring method according to claim 1, characterized in that: The sampling frequency of the feed axis servo monitoring signal and the cutting force should be consistent, and should be at least twice the maximum cutting frequency in the predicted requirements; if monitoring conditions permit, it should be more than five times.