Data mining and statistical analysis of construction equipment failure

Purpose Construction equipment is a key resource, and contractors that own a large equipment fleet take all necessary measures to maximize equipment utilization and minimize equipment failures. Although most contractors implement scheduled maintenance programs and carry out periodic inspections and repairs on their construction equipment, it is still difficult to predict the occurrence of a specific failure of a piece of equipment in the short or long term. According to a survey in the United States, approximately 46% of the major equipment repairs was undertaken as a result of an unexpected failure. Although it is not possible to predict all failure events, a slight improvement in their prediction represents a significant saving in time and cost for a large contractor. Statistical power law models and data-mining models were compared to investigate their pros and cons in predicting critical failure events of heavy construction equipment. Method With large amounts of equipment failure data accumulated in a surface mining project, two different types of failure models were created for comparative analysis from a practical point of view. For selected equipment units, failure data were collected along with the relevant factors which may cause variations of equipment failure rate (or mean time to failure). In a classical approach, Power law models of equipment failure rates are fitted using RGA 7.0; while in the data-mining approach, the mean time to failure is modeled using a data-mining algorithm-decision tree induction, establishing logical, mathematical, and statistical relations between MTTF (Mean Time Between Failures) and its various factor of impact (equipment conditions, failure history, environmental conditions, etc.). Both models are used for validation tests on randomly selected time periods and compared in terms of their performance. Results & Discussion The two types of models were compared.