Motion Control Resources
Optimized Platen Control Using DC Motors for Soil Testing
by Rory Cox & J. Antonio H. Carraro, Riordan Cox & Associates / The University of Western Australia
National Instruments Posted 03/17/2016
Developing accurate and simultaneous control of boundary loads and displacements applied to soil specimens tested in the UWA simple shear apparatus with both vertical and horizontal platen positions controlled/moved at a rate of several millimeters per second to an accuracy of a couple of micrometers.
Using the CompactRIO platform to control platen movement under either displacement or load feedback, with seamless switching between these two controls, to achieve accurate resolution by applying a pulsed signal to the motors using a dual-loop control system.
The Geotechnical Testing Laboratory of the Centre for Offshore Foundation Systems (COFS) at UWA started developing testing equipment in the 1990s. Since then, the laboratory has become a major provider of specialized geotechnical testing services for the offshore oil and gas industry. Over the last few years, the laboratory has upgraded its in-house equipment and now houses some of the most advanced geotechnical testing equipment in the world.
Rory Cox founded Riordan Cox and Associates in 1996 to deliver a service to the mining and petroleum industry in Australia for specialized soil and rock testing equipment. NI Australia contacted him about Riordan Cox becoming an NI Alliance Partner because he had used LabVIEW since 1991. Originally, Cox worked for the CSIRO Division of Geomechanics on a joint project with UWA for Woodside Offshore Petroleum on pile foundations in calcareous sands. The university has contacted him to assist with a number of projects since 1999. A new collaboration, which resulted in the development of the current control framework outlined in this paper, was initiated in 2012, after Professor Antonio Carraro joined COFS-UWA.
Simple Shear Equipment
COFS designed and built the simple shear device schematically shown in Figure 1. In this device, a cylindrical specimen is located between two rigid platens. The upper platen moves vertically, whereas the lower platen can move horizontally. A rubber membrane seals the specimen and isolates it from the cell pressure applied to the inside of the steel cylindrical chamber, using a cell pressure controller developed at COFS-UWA. A separate pump, manufactured by GDS Instruments, controls the pressure applied to the fluid filling the pore space within the specimen (back pressure).
Figure 1. Panel Showing Details of the COFS-UWA Simple Shear Apparatus
A separate motion actuator designed and constructed at COFS-UWA controls the linear motion of each platen. Each motion actuator includes a DC motor and encoder set attached to gear trains that convert motor rotation into platen movement. According to a 2003 article in Géotechnique by Mao and Fahey, this hardware is similar to that used in the previous generation of UWA simple shear apparatus.
In 2005, we started a project to convert the original control software for this equipment to LabVIEW. We originally selected a control board from NI to manage platen movement. However, selectable analogue feedback of both displacement and load on the platens precluded that device. In 2012, we replaced that motion control board with the cRIO-9076 controller to develop our current control system. The main advantage of the NI CompactRIO system was that we could conduct platen control with either displacement or load feedback, and that seamless transfer of control could occur at any stage during a test.
The CompactRIO was a natural choice because it could deliver flexibility and the program cycle speed necessary to achieve the motor control required. We used LabVIEW 2012 with real-time capability to program the FPGA chip, the target program on the CompactRIO, and the host program running on a Windows environment in a desktop computer.
Platen Control Program
The motors in this project operate with a control signal of +/-10 V. However, they require up to 2 V generating enough torque in overcoming static friction. As a result, we could not apply a conventional PID control algorithm to the motor to move the platens accurately, particularly when slow rates of movement are required. We designed an alternative control algorithm that split control into two parts. This provided smooth and accurate control regardless of the rate of movement required.
We achieved secondary control by incorporating a quadrature encoder onto the shaft of the motor, as shown in Figure 2, and applying a set of proportional and differential gains. In this case, we define the input signal as a rotational rate in encoder pulses per second. The quadrature encoder program measures two input signals (A and B). Both the rising and falling pulse of each input are recorded as counts. If the A pulse precedes B, then the count increases, otherwise they fall.
This control loop is unique to the specific motor, and is independent of what is happening at the platen being driven by the motors. We adjust the proportional gain to deliver the resolution required of the motor. We subtract the differential gain from the signal to start deceleration to overcome the inertia of the motor.
The operator can use the primary control loop (schematically shown in Figure 3 for the vertical platen) to select the type of control feedback signal desired. The operator can choose from three options: displacement of the platen, displacement of the motor (encoder), or load on the sample. We compare the feedback with the required signal. This is amplified by the primary gain to generate a rate at which the motor must run. We feed this signal to the secondary control loop to run the motor.
Once the gains for both the primary and secondary control loops have been optimized, system response improves within the capabilities of the hardware and electronics used. Figure 4 shows this schematically for the displacement control option. The platen moves over a range of 10 mm within 1 s, and then can maintain the displacement to within 0.001 mm.
Figure 4. Variation in Horizontal Platen Over Time (displacement control)
(a) Overall Response (b) Detailed Response Around Target Displacement of 5 mm
Encoder readings indicate that the motor starts turning before the platen actually moves. This is due to inherent backlash in the gear train for this system (horizontal actuator).
Figure 5. Variation in Horizontal Load Over Time (load control)
(a) Overall Response (b) Detailed Response Around Target Load of 100 N
The same applies for load control, as shown in Figure 5. In this example, we inserted a rubber specimen between the platens and applied a step horizontal (shear) load of 100 N. The specimen deflects by just over 0.5 mm in about 0.2 s. In this case, the rubber specimen continues to creep under a constant load as shown in Figure 5b and deforms by a further 0.07 mm in the next 7 s. The horizontal load is maintained within 0.2 N, which would be an acceptable resolution for the system having a horizontal load capacity of 4,000 N.
The CompactRIO controller together with the LabVIEW Real-Time Module provided the program loop time of 1 ms intervals on the FPGA chip necessary to control the two DC motors of the new generation of COFS-UWA simple shear devices. We measured the analogue displacement and load inputs using the NI 9215 input module, and controlled the analogue outputs to the motors using the NI 9263 output module. We used an NI 9401 digital I/O module to activate the motor controllers and monitor the vertical and horizontal limit switches. The second digital I/O module monitored the quadrature encoders attached to the motor shafts. The FPGA program that monitored the encoder pulses ran at a loop period of 5 ms intervals.
We transferred data from the FPGA program through the CompactRIO target program at 10 readings/s. The FPGA control program acquires this data at 1,000readings/s, so every 100 readings are averaged to improve the resolution of the readings in the host program.
With the optimal gains set, the motors could move both platens over 10 mm within 1 s and then maintain the resolution of the corresponding displacement and load readings within 0.001 mm and 0.2 N, respectively.
The system is flexible and we can adapt the solution to a large range of other applications using similar technology.
Riordan Cox and Associates have used NI hardware for more than 20 years, and the ability to upgrade earlier programs and hardware seamlessly was essential. The introduction of CompactRIO was critical in making this project successful. Using this technique, high-speed digital control was not only possible, but also relatively easy to implement and modify.