Jahmy Hindman

Monitoring the Condition of a Valve and Linear Actuator in Hydraulic Systems

Abstract The topic of condition monitoring has been a growing area of research in both academia and industry for much of the last two decades. Condition monitoring of fluid power equipment has been no exception to this trend. Much of the research work associated with monitoring the condition of fluid power equipment has centered on pump and motor components due to their relatively high cost and complexity. The work in this paper focuses on the lesser expensive, but more common components of valves and linear actuators. The primary focus of the work presented here pertains to assessing the independent component condition of a valve-controlled linear actuator circuit. The paper first presents simulation studies to establish techniques for proper data collection, neural network training and output interpretation. The neural network approach is then applied to a valve and linear actuator of a John Deere 410E Backhoe Loader. The results indicate that the concept can be applied to a commercial system and is feasible for implementation.

An Artificial Neural Network Approach to Payload Estimation in Four Wheel Drive Loaders

Estimation of the manipulated payload mass in off-highway machines is made non-trivial by the nonlinearities associated with the hydraulic systems used to actuate the linkage of the machine in addition to the nonlinearity of the kinematics of the linkage itself. Hydraulic cylinder friction, hydraulic conduit compressibility, linkage machining variation and linkage joint friction all make this a complex task under even ideal (machine static) conditions. This problem is made even more difficult when the linkage is mobile as is often the case with off-highway equipment such as four-wheel-drive loaders, cranes, and excavators. The rigid body motion of this type of equipment affects the gravitational loads seen in the linkage and impacts the payload estimate. The commercially available state-of-the-art load estimation solutions rely on the mobile machine becoming pseudo-static in order to maintain accuracy. This requirement increases the time required to move the material and decreases the productivity of the machine. An artificial neural network solution to this problem that enables the machine to remain dynamic and still accurately estimate the payload is discussed in this paper. Development and implementation on an actual four-wheel-drive loader is shown.Copyright