1. Machine Control Technologies
Because proportional-integral-derivative (PID) controllers are reliable and easy to implement, PID has become the most commonly implemented control algorithm for machine control. To implement PID, you can choose from several hardware platforms such as programmable logic controllers (PLCs), programmable automation controllers (PACs), digital signal processors (DSPs), and microprocessors. Before selecting one of these platforms for PID, you need to compare the reliability, hardware availability, and cost of the different options. You also need to determine your desired loop rate based on your machine. For example, high control-loop rates are critical in high-end machines because they directly affect machine performance and disturbance rejection. Figure 1 plots different step responses of the same PID brushless DC motor controller. Higher loop rates help the system move faster and with less overshoot, yielding better performance.
Figure 1. Effects of Different Step Responses of the Same PID Brushless DC Motor Controller on Loop Rates in Applied Step Response Test
PID has been around for decades, but another motion control technology has emerged in recent years: field-programmable gate arrays (FPGAs). This technology offers the same reliability and performance as custom-built circuits with the flexibility of a software development tool. Additional FPGA benefits include true parallel execution and fast execution cycles. This new technology allows for user-defined hardware execution behavior that you can change or tune as the design process evolves or you refine your specifications.
Whichever machine control technology you select, you likely will face resonance challenges. Typically, at a certain point, the machine starts to make a noise, thus signaling resonance, one of the most perplexing problems for machine builders. Figure 2 shows the step response of a simulated system displaying the oscillations caused by resonance.
Figure 2. Oscillations Caused by Resonance
Resonance occurs when the motor and load are not rigidly coupled. Shafts, belts, gearboxes, and so on introduce compliance, or “springiness,” and viscous damping between motor and load. Figure 3 shows a motor, load, and coupling schematic. Motor and load inertias combined with transmission compliance behave the same as a two mass-spring system where the oscillating frequency is the resonance frequency.
Figure 3. Motor, Load, and Coupling Schematic
Figure 4 clearly illustrates the resonance on the open-loop body plot of a simulated system. An ideal, rigidly coupled system exhibits this integrator behavior: magnitude decreasing at 20 dB/decade and phase fixed at -90 deg. Instead, the behavior is the same as the ideal system up to the antiresonance frequency (Far), but then there is an increase in both phase and magnitude with a maximum gain in resonance frequency (Fr).
Figure 4. Resonance on the Open-Loop Body Plot of a Simulated System
Depending on the resonance frequency placement, there are two types of resonance:
- Low frequency –The most common resonance type in industrial applications, low-frequency resonance occurs when resonance frequency is above (but not much) the servo controller bandwidth.
- High frequency –High-frequency resonance occurs when the resonance frequency is very high compared with the servo controller bandwidth. This is typical on high-end industrial machines, such as computer numerical control (CNC) or machine tools, where you increase stiffness to increase controller gains.
You can avoid resonance using different mechanisms. Among the first ones to consider are mechanical variations. By changing motor or load inertias or increasing transmission stiffness, you can move frequency resonance so that it does not appear in machine operation ranges. Unfortunately, these solutions require bigger and heavier transmissions, which increases commercial off-the-shelf (COTS) machine cost. Or your motor may already be specified based on price or performance, so you are not able to consider inertia.
Once you have exhausted the mechanical solutions, consider electrical solutions. These involve the machine motor velocity controller that affects the control commands sent to the servos. Figure 5 shows a typical motion control system.
Figure 5. Typical Motion Control System
You can use two technologies to address resonance by type. The first is lowpass filters, which reduce the gain at the resonance frequency. Lowpass filters are easy to implement because you do not have to know the exact resonance frequency. Unfortunately, lowpass filters are not very effective with low-frequency resonance. Figure 6 shows the effect on the frequency response of lowpass filters when applied with PID on the open-loop Bode plot.
Figure 6. FPGA PID and FPGA PID with Lowpass Filter Bode Plots
The second technology is notch filters. Using notch filters, you can target resonance frequency while maintaining low phase distortion to achieve higher gains in the control loop and improve performance. The drawback is that resonance frequencies are not fixed and change over time, so you may need to make a service trip to adjust the notch filter. Figure 7 illustrates the results of applying a notch filter.
Figure 7. Body Plot from an FPGA-Based PID Implementation with Lowpass and Notch Filters
You can implement your resonance-reducing technologies through off-the-shelf FPGA platforms and easy-to-use programming environments, which National Instruments offers to help machine builders apply advanced control technology while reducing design and development time. On the software side, the NI LabVIEW PID Control Toolkit seamlessly integrates with the NI LabVIEW FPGA Module to offer both one-channel and multichannel PID VIs. The new LabVIEW 8.5 introduces lowpass and notch filters. Figure 8 shows a simplified version of PID implementation for a velocity loop with both lowpass and notch filters. You can export the output to an analog output, as shown in the figure, or you can use it to drive an internal PWM current control loop. One benefit of this implementation is the online filter configuration. In addition, you can configure the system to be updated over the Internet to avoid expensive service trips.
Figure 8. Simplified PID Implementation for a Velocity Loop with Both Lowpass and Notch Filters
For FPGA hardware, consider high-end PXI systems from National Instruments with reconfigurable R Series devices featuring 1M or 3M gates or the industrial-rated reconfigurable I/O (RIO) devices with FPGAs embedded in the backplanes. Both options have the necessary I/O connectivity for motors, encoders, digital and analog signals, and so on. Another benefit of NI hardware is seamless code reuse. You can design and prototype machine controllers using the stronger processing power of a dual-core PXI system, and you can easily port code to NI CompactRIO hardware for product deployment.
Figure 9. CompactRIO System
This paper addressed how to increase machine performance while dealing with mechanical resonance by including lowpass and notch filters after the PID velocity loop on FPGA-based hardware targets. With new commercially available hardware technology and easy-to-use software tools, you can implement advanced and flexible control technology without needing low-level electronic knowledge. Breadth of analysis tools, code reuse, and hardware modularity contribute to faster design cycles and shorter development times.
For more information, visit:
NI LabVIEW FPGA