Analysis and Application of Feedforward Control in Bus Servo Systems

With the introduction of slogans such as "Industry 4.0" and "Made in my country 2025," the industrial sector has placed higher demands on servo drive technology. Traditional pulse-type servo systems, due to their complex wiring, low reliability, and multi-axis motion systems, cannot meet market requirements, leading to the emergence of bus servo systems. However, CNC machine tools and industrial robots demand high speed, high precision, no overshoot, and rapid positioning from servo systems.

At present, the general servo drive technology adopts a three-loop feedback control of position loop, speed loop and current loop, plus speed feedforward and torque feedforward control [1][2][3]. The addition of feedforward control improves the response speed and tracking characteristics of the system. Reference [4] proposes to introduce position differential negative feedback on the basis of feedforward control to overcome the large overshoot of feedforward control for discontinuous position response. Reference [5] compares the dual feedforward control composed of speed feedforward and torque feedforward with the dual feedforward control technology composed of speed feedforward and acceleration feedforward, and shows that the latter dual feedforward control can better improve the ability of the servo system to track maneuvering targets. Reference [6] not only added speed and acceleration feedforward, but also position feedforward control. Experiments show that the position feedforward controller can meet the requirements of high-performance servo positioning. It can be used not only for point-to-point control, but also for contour control in interpolation control. However, regardless of whether it is velocity feedforward, torque feedforward, or acceleration feedforward, all of the above methods use differential operations on the position command to obtain the velocity feedforward quantity, differential operations on the velocity feedforward quantity to obtain the acceleration feedforward quantity, or direct differential operations on the position command to obtain the torque feedforward quantity. Without exception, they all use the differential extraction method.

This paper proposes a new extraction method for feedforward quantities based on a bus servo drive platform, which is proposed to address the traditional method for extracting feedforward quantities. This method is applied to a stamping robot and its performance is compared with that of the traditional differential feedforward control method.

1. AC servo motor control model

The mathematical model of an AC servo motor is a complex nonlinear system with multiple variables and strong coupling. In order to achieve high-performance control of this complex system, it is necessary to transform the mathematical model in the three-phase stationary coordinate system to the two-phase rotating coordinate system, and then control the excitation current component and the torque current component separately, that is, control the excitation and torque of the motor [7]. The mathematical model of the AC servo motor in the two-phase stationary coordinate system is given directly below:

The voltage equation is:

Figure 1. Block diagram of AC servo three-closed-loop control

2. Feedforward control

2.1 Traditional Feedforward Control

The AC servo three-loop control system shown in Figure 1 consists of a position loop, a speed loop, and a current loop. All three loops employ feedback control structures, resulting in large steady-state tracking errors and poor dynamic tracking performance, which fails to meet the requirements of current high-performance equipment. To address these shortcomings, the following control block diagram with feedforward control is provided:

Figure 2. Block diagram of traditional feedforward control

In Figure 2, point 1 represents the position command, point 2 represents the velocity feedforward, and point 3 represents the torque feedforward. Points 2 and 3 are obtained by differentiating the position command and then adding a certain filter. This is a general traditional feedforward control method, and most other improvements to feedforward control are based on this. The feedforward quantity extracted by differentiating the position command has at least one control cycle in timing. The following is a schematic diagram of the position command and the feedforward quantity obtained by differentiation in some trapezoidal curve planning:

Figure 3. Timing diagram of traditional position command and feedforward.

As can be seen from Figure 3, the speed feedforward and torque feedforward obtained by differentiation are both lagging behind the position command in terms of timing and are not synchronous output. With the addition of the smoothing filter, a certain delay is generated in the control, which inevitably fails to achieve the best effect of feedforward control.

2.2 Novel Feedforward Control

To address the timing lag issues inherent in traditional feedforward control, and considering that position trajectory planning in bus servo drives is internal to the drive itself, the following improvements are made using the CANopen servo drive as an example:

Figure 4. Block diagram of the novel feedforward control

In Figure 4, point 1 represents the commands sent by the CANopen master control device, including target position, target speed, target acceleration/deceleration, and control words. Since the trajectory planner is implemented within the driver, the driver can synchronously plan the position command at point 3, the speed feedforward at point 4, and the torque feedforward at point 5 according to the received commands. The timing diagram is as follows:

Figure 5. Timing diagram of the new position command and feedforward quantity

As can be seen from Figure 5, at the start of the position command planning at t1, the speed feedforward and torque feedforward planning start simultaneously, so that the acceleration segment, constant speed segment and deceleration segment are all synchronized with the position command planning, rather than the traditional method of taking the difference and differentiation of the control cycles before and after the position command.

3. Experiment on the application of stamping robot

A three-axis stamping robot was driven by a CANopen servo driver, and application tests were conducted on the stamping robot to compare the traditional feedforward control and the new feedforward control.

Figure 6 shows a real-world diagram of the stamping robot, which includes the swing arm axis, the upper and lower axes, and the telescopic axis, all driven by three CANopen servo drives:

The following are waveforms and locally magnified waveforms of the three axes under both traditional and novel feedforward control with a 300% speed feedforward. The horizontal axis represents time (ms):

3.1 Swing arm shaft

Figure 7. Waveform of traditional feedforward control

Figure 8. Novel feedforward control waveform

3.2 Upper and lower shafts

Figure 9. Waveform of traditional feedforward control

Figure 10 Novel feedforward control waveform

3.3 Telescopic Shaft

Figure 11 Waveform of traditional feedforward control

Figure 12 Novel feedforward control waveform

3.4 Data Comparison

From Figures 7 to 12 above, the positioning time when the position error is 100 pulses can be extracted:

From Figures 7 to 12 above, the maximum position error pulse count when each axis accelerates to its maximum speed can be extracted as follows:

The above data is based on a feedforward control ratio of 300%. When this ratio is set higher, the improvement effect is more significant.

4. Conclusion

As bus servo systems become increasingly popular in the market, higher demands are being placed on both high dynamism and high positioning accuracy. Feedforward control not only achieves accurate positioning without overshoot but also improves the dynamic performance of the servo system. This paper proposes a novel feedforward control method specifically for the curve planning characteristics of bus servo drivers. Compared with traditional feedforward control methods, experimental data shows that this method optimizes the extraction of speed and torque feedforward quantities and improves the tracking and dynamic performance of the servo system.

For more information, please visit the Server Systems channel.