An automatic system for categorizing and quantifying human factors

Abstract

Human factors (HF) are little understood, and particularly in non-destructive testing (NDT) experimental data is rare, samples are often small, and statistical methods are rarely used to evaluate results [1]. HF have been widely implicated in major occurrences of technical failure, for example at North Anna power plant and on United Airlines Flight 232. Understanding HF is vital for reliable detection and prevention of failures. Reliability assessments, though known to be affected by intrinsic capability, application factors, and HF, have thus far only concentrated to a sufficient degree on intrinsic capability. The addition of HF to that assessment has proven difficult due to the lack of a method of quantifying HF. This paper presents the first attempt from a psychological perspective to quantify HF from qualitative data. HF data was derived from qualitative human-oriented Failure Modes and Effects Analysis (Human-FMEA) workshops for visual inspection of tunnels by laser scanning and for ultrasonic testing of welds. Data was collected on human failure modes, causes, consequences and preventive measures, as well as eliciting a risk priority number (RPN). Using this data, a system of quantitative weightings was created to allocate errors to inductively derived HF categories for further allocation to existing HF categorisation models. This weighting model proved useful for creating quantitative summaries of HF, informing and validating qualitative FMEA results, and comparing existing HF categorisation models. Further potential lies within a planned interface to quantitative reliability assessment methods such as POD (Probability of Detection). While providing quantifications, the method retains a qualitative and holistic nature, can, thus, bridge the gap between psychological and engineering concerns of HF and reliability, contributing to future interdisciplinary work.