Development of a KPI-focused hybrid model for the Cummins QSB6.7 engine used in load haul dump vehicles

Abstract

Diesel engines are vital in mining operations, powering machinery in harsh environments where reliability is critical to avoid costly downtimes and economic losses. This study presents a hybrid model for the Cummins QSB6.7 turbocharged diesel engine to optimise predictive maintenance and operational efficiency in load-haul-dump vehicles. Controlled experiments generated key performance indicator (KPI) data under varied loads, capturing thermal, pressure, power, and efficiency metrics via QuantumX and CANedge2 systems. Physics-based models were calibrated using global and local optimisation, with complex metrics like turbocharger rotational speed as design variables. Neural networks mapped operational data, such as engine speed and load, to these variables, enabling accurate KPI predictions with minimal input. The model achieved MAPEs of 6.21% for thermal, 5.08% for pressure, and 4.12% for power, demonstrating strong predictive accuracy and practical applicability in mining contexts. These results underscore the model's potential to significantly reduce unplanned downtimes and associated economic losses.