Customer SolutionsTesting Control Strategies on a Combine Harvester with NI LabVIEW Real-Time, PXI
Author(s):Tom Coen, Katholieke Universiteit Leuven
Industry:Machines/Mechanics, Research, University/Education
Product:CAN, LabVIEW, LabVIEW Real-Time, PXI/CompactPCI
The Challenge:Developing an environment that can rapidly implement different strategies to control the speed of a combine harvester and optimize efficiency.
The Solution:Increasing harvesting efficiency and maximizing machine throughput using National Instruments LabVIEW Real-Time and PXI for optimized, rugged control of the machine’s velocity.
The Need for a More Efficient Control System Harvesting machines, such as combines, are costly and only used during peak periods. To be most efficient, these machines must achieve a certain speed to optimize crop throughput. Several variables in the harvesting process factor into this, including crop density per square meter and the amount of grains in the crop. Our original combine control system ran on a Microsoft Windows environment and was written in C++. It used a PCMCIA-CAN board to interface with the machine’s hardware. This setup made it difficult to keep the system deterministic. For example, mouse movements could cause the system to miss certain messages; extending to a higher control loop frequency was not possible. Adding other types of measurements was also not possible. Implementation of the New Control System When defining the new system, we looked at different model-based design tools, including National Instruments LabVIEW Real-Time. One of the reasons we chose the setup with NI LabVIEW Real-Time and PXI is that integrating LabVIEW, data acquisition, and PID control algorithms is very easy, and we can acquire analog data at both low sample rates (Hz) and very high sample rates (hundreds of kHz).The PXI solution also provides the ruggedness we need in the harsh environment of a combine harvester. The program we wrote achieves two principle goals – to apply excitations to the machine for modeling purposes (up to 20 Hz sampling frequency), and to test and implement new control laws (5 Hz sampling frequency). We can perform machine excitations at higher frequencies to capture all machine dynamics. Most of the machine measurements are transmitted to the NI PXI-8461 via a Controller Area Network (CAN). Some other experimental sensors are directly connected to the PXI system through a data acquisition card. The sampling frequency for these sensors may be as high as 100 kHz. The PXI system has to function even under very harsh conditions, so it is placed in the cab of the combine harvester where it is protected from extreme temperatures and rain. The system is subjected to vibrations caused by the engine and the threshing process, as well as the rough terrain, so the PXI system is equipped with a solid state hard disk. The experiments require the integrated use of CAN bus and analog data acquisition. The machine measures the input density of the crop as well as the output flow rate of corn grains. Sensors are installed at the back of the machine to measure grain losses. Counter boards are used to count the impacts on these loss sensors, and these inputs are used together with several others to maintain the harvester’s most efficient velocity (km/h). Control Laws Many different control laws are implemented, including fuzzy and PID control. For the PID controller, the process amplification is estimated online using a recursive least squares estimator. The estimate is used to keep the total amplification of the process and the controller constant. These controllers are used to control the machine speed. The driver often accelerates with a step signal (from 0 km/h to 7 km/h, for instance). As such, the controller has to be robust to process disturbances around the set point, and has to remain stable when large set point errors occur. This is achieved through gain scheduling. In this way, a fast but stable transient response can be combined with a good disturbance rejection. In rough terrain, the controller needs to take the engine load into account. A heuristic rule is added to the PID controller to lower the speed when the constraint on engine load is broken. The fuzzy controllers implemented are actually fuzzy versions of a PID controller. The inputs are, for instance, set point error, change of error, and integrated error. Self-written code was used to implement fuzzy control and to adapt control parameters easily on the real-time system. For this application, the fuzzy controllers delivered similar performance to that of the PID controllers; however, gain scheduling can be implemented more naturally in a fuzzy controller. Conclusion and Future Work Without previous LabVIEW programming experience, we were able to develop a robust, fuzzy control system for controlling the harvester’s speed in a short period of time. In the future, we plan to add image processing to the system, for which the modularity of PXI will be a big advantage. We also plan to implement a model-based process control in future versions. For more information, contact: Tom Coen
|
