Thesis (MEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: Statistics from mining accidents expose that the potential for injury or
death to employees from occupational accidents is relatively high. This study
attempts to contribute to the on-going efforts to improve occupational safety
in the mining industry by creating a model capable of predicting the continuous
risk of occupational accidents occurring. Model inputs include the time
of day, time into shift, temperatures, humidity, rainfall and production rate.
The approach includes using an Artificial Neural Network (ANN) to identify
patterns between the input attributes and to predict the continuous risk of
accidents occurring. As a predecessor to the development of the model, a
comprehensive literature study was conducted. The objectives of the study
were to understand occupational safety, explore various forecasting techniques
and identify contributing factors that influence the occurrence of accidents and
in so doing recognise any gaps in the current knowledge. Another objective
was to quantify the contributing factors identified, as well as detect the sensitivity
amongst these factors and in so doing deliver a groundwork for the
present model.
After the literature was studied, the model design and construction was
performed as well as the model training and validation. The training and
validation took the form of a case study with data from a platinum mine
near Rustenburg in South Africa. The data was split into three sections,
namely, underground, engineering and other. Then the model was trained and
validated separately for the three sections on a yearly basis. This resulted
in meaningful correlation between the predicted continuous risk and actual
accidents as well as the majority of the actual accidents only occurring while
the continuous risk was estimated to be above 80%. However, the underground section has so many accidents, that the risk is permanently very high. Yet, the
engineering and other sections produced results useful for managerial decisions. / AFRIKAANSE OPSOMMING: Mynbou ongeluk statistieke dui aan dat die potensiaal vir besering of dood
as gevolg van beroepsongelukke relatief hoog is. Die studie poog om by te dra
tot die voortdurende verbetering van beroepsveiligheid in die mynbedryf deur
middel van ’n model wat die risiko van beroepsongelukke voorspel. Die model
vereis die tyd, tyd verstreke in die skof, temperatuur, humiditeit, reënval en
produksie tydens die ongeluk as inset. Die benadering tot hierdie model maak
gebruik van ’n Kunsmatige Neurale Netwerk (KNN) om patrone tussen die
insette te erken en om die risiko van ’n voorval te beraam. As ’n voorloper
tot die model ontwikkeling, is ’n omvattende literatuurstudie onderneem. Die
doelwitte van die literatuur studie was om beroepsveiligheid beter te verstaan,
verskeie voorspellings tegnieke te ondersoek en kennis van bydraende faktore
wat lei tot voorvalle te ondersoek. Nog ’n doelwit sluit die kwantifisering in van
geidentifiseerde bydraende faktore, asook die opsporing van die sensitiwiteit
tussen hierdie faktore en hierdeur ’n fondasie vir die voorgestelde model te
skep.
Na afloop van die literatuurstudie is die model ontwikkel, opgelei en gevalideer.
Die opleiding en validasie is deur middel van ’n gevallestudie in ’n
platinummyn naby Rustenburg in Suid Afrika gedoen. Die data is verdeel in
drie afdelings, d.i. ondergronds, ingenieurswese en ander. Die model is vir
elke afdeling apart opgelei en gevalideer op ’n jaarlikse basis. Hierdie het gelei
tot ’n betekenisvolle korrelasie tussen die voorspelde risiko en die werklike
ongelukke met die meerderheid van die werklike ongevalle wat voorgekom het
terwyl die risiko 80% oorskry het. In die ondergrondse afdeling is so baie voorvalle waarneem dat die risiko permanent hoog is. Die ander afdelings het wel
resultate verskaf wat sinvol gebruik kan word in bestuursbesluite.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:sun/oai:scholar.sun.ac.za:10019.1/96072 |
| Date | 12 1900 |
| Creators | Van den Honert, Andrew |
| Contributors | Vlok, P. J., Stellenbosch University. Faculty of Engineering. Department of Industrial Engineering. |
| Publisher | Stellenbosch : Stellenbosch University |
| Source Sets | South African National ETD Portal |
| Language | en_ZA |
| Detected Language | English |
| Type | Thesis |
| Format | xxi, 163 p. : ill. |
| Rights | Stellenbosch University |
Page generated in 0.0021 seconds