Judging the Abnormalities of Agricultural Machinery by Using Mahalanobis’ Distance

This study calculated the Mahalanobis’ distance, which is a multidimensional space distance with correlations, from the time–series data of 3–axis translational acceleration and 3–axis rotational angular speed, and examined whether this could be used in judging the abnormalities of the agricultural machineries. As the result, using the Mahalanobis’ distance to determine the abnormality of the data was not possible for the changes in not–so–big behaviors, such as turning and temporary stop, but determining the abnormality of the data using the Mahalanobis’ distance was clearly possible for sudden changes to the operating status of the equipment, such as a roll over and passing obstacles through an experiment using the model car. We hypothesized in the beginning that the distribution of the Mahalanobis’ distance at the signal space could be separated with the distribution of the Mahalanobis’ distance at the unit space. However, unless there is a large–scale change to the behavior, such as a roll over, etc., complete separation is difficult in reality, and we determined realistically that conducting the abnormality determination from the significant difference viewpoint by placing a threshold value to the normalized distribution of the Mahalanobis’ distance at the unit space is possible