Long-term tracking and automated analysis of tractor behaviour with Jdlink

Abstract

JDLINK is an application for farm management under a precision agriculture strategy (variable rate application of inputs with geolocated data and maintenance alerts among others. It allows accessing to isolated or groups of machines, with the possibility of downloading CAN BUS data in periods established by use (engine hours) or by date. JDLINK can be considered a SCADA (System Control - SC- and Data Acquisition -DA-) where the visualization of data is indispensable to the manager of the fleet of machines. Our approach in this research is clear: What practical recommendations in terms of energy efficiency, work performance and improvement of use (maintenance) can be made based on the automate analysis of the information contained in JDLINK? For this prospective study, two tractors were used: one 8345R of 380 HP (1000 hours) belonging to a cooperative in Navarra (Spain), and another one 7230R of 230HP (500 hours) belonging to a private farmer in Ciudad Real (Spain); we chose 42 variables: 17 motor, 10 on maintenance and 17 derived from the tasks performed. Then we programed and tested an automated multivariate analysis for the recognition of agricultural activity patterns which consists of a Principal Component Analysis (PCA), an unsupervised cluster (Cluster) and a multiple analysis of variance or MANOVA. The use of the automated procedure has led to the isolation of several patterns as identifiable tasks four in the case of 8345R and six different patterns for the 7230R tractor. As a consequence, a multivariate pattern became available and thus the identification of machine performance and fuel consumption pattern (l ha-1). Moreover, clear conclusions were derived upon the misuse of counterweights, low level and excessive variability of slippage (below 4% on average basis. At present this procedure is available for dealers to provide highly valuable agricultural consultancy for farmers.