Analysis of forklift data – A process for decimating data and analyzing fork positioning functions

Abstract: Investigated in this thesis are the possibilities and effects of reducing CAN data collected from forklifts. The purpose of reducing the data was to create the possibility of exporting and managing data for multiple forklifts and a relatively long period of time. For doing that was an autoregressive filter implemented for filtering and decimating data. Connected to the decimation was also the aim of generating a data set that could be used for analyzing lift sequences and in particular the usage of fork adjustment functions during lift sequences. The findings in the report are that an AR (18) model works well for filtering and decimating the data. Information losses are unavoidable but kept at a relatively low level, and the size of data becomes manageable. Each row in the decimated data is labeled as belonging to a lift sequence or as not belonging to a lift sequence given a manually specified definition of the lift sequence event. From the lift sequences is information about the lift like number of usages of each fork adjustment function, load weight and fork height gathered. The analysis of the lift sequences gave that the lift/lower function on average is used 4.75 times per lift sequence and the reach function 3.23 times on average. For the side shift the mean is 0.35 per lift sequence and for the tilt the mean is 0.10. Moreover, it was also found that the struggling time on average is about 17 % of the total lift sequence time. The proportion of the lift that is struggling time was also shown to differ between drivers, with the lowest mean proportion being 7 % and the highest 30 %.