Premium
Scoring methods for verification and diagnostic performance in industrial fault‐finding problems
Author(s) -
DUNCAN K. D.,
GRAY M. J.
Publication year - 1975
Publication title -
journal of occupational psychology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.257
H-Index - 114
eISSN - 2044-8325
pISSN - 0305-8107
DOI - 10.1111/j.2044-8325.1975.tb00303.x
Subject(s) - context (archaeology) , set (abstract data type) , oil refinery , fault (geology) , computer science , simple (philosophy) , measure (data warehouse) , distillation , machine learning , artificial intelligence , psychology , reliability engineering , natural language processing , data mining , engineering , programming language , epistemology , seismology , geology , paleontology , philosophy , chemistry , organic chemistry , biology , waste management
The man trying to find a fault may make mistakes unless he asks effective diagnostic questions and verifies the answers. General principles for verifying indications in petroleum refineries are presented; some are easy to apply, others require instruction, practice or experience. Simple scores for verification or checking, C , and for effective diagnostic questioning, D , are proposed and their calculation is illustrated in the context of fault‐finding in a crude distillation unit. C and D probably measure different fault‐finding skills. The C score is improved by training, but comparisons are complicated by an apparent influence of the proportion of instrument faults in the problem set. Men may check instrument readings more often as their estimate of instrument unreliability increases. The D score is also sensitive to the effects of training and reflects differences in experience between novices and experienced men. D may over‐estimate effective diagnostic strategy, but only in a small percentage of cases for men with refinery operating experience or new entrants after training.