Automation of workflow control of construction and road machines based on intelligent data processing algorithms

Abstract

Onboard measuring systems of construction and road machines generate about 0.35 GB of data per machine over a ten-hour shift, yet for most work processes the control decision is still made by the operator with a lag comparable to the correlation time of the external disturbance. Greater algorithmic complexity raises the computational delay and cancels the gain in prediction accuracy, and this mismatch remains unresolved. The aim of the work is to quantify the conditions under which an intelligent algorithm processing onboard data keeps a technological parameter within the regulatory tolerance. An author's computational model of the closed loop «measurement – processing – decision – actuation» was built for soil compaction by a vibratory roller, layer grading by a motor grader and excavation of a cut by a hydraulic excavator; the residual deviation was estimated through the minimum attainable variance under dead time and verified by simulation over 4 million samples. Replacing threshold logic with gradient boosting cuts the share of degree-of-compaction readings outside the 0,04 tolerance from 16,8 to 2,6%, while a recurrent network with a computational delay of 0,214 s yields 2,6%, so the accuracy gain is absorbed by the lag. The critical computational delay for the compaction loop equals 0,206 s. The limit of work process automation is set by the ratio of the total loop delay to the disturbance correlation time.