Halina Pawlak
Ergonomic evaluation of workstations in the agricultural and food industry
The paper deals with the issues of the ergonomic evaluation of workstations in the agricultural and food industry. The analysis of the national and foreign literature which was carried out showed that the issues of the ergonomic evalu-
ation of workstations in the agricultural and food industry deals mainly with the occupational risk and the casualty rate while the methods used so far don’t take the ergonomic knowledge into consideration in a comprehensive way.
The methodological assumptions of the evaluation of the ergonomic workstation in the agricultural and food industry with the use of the formal system of knowledge representation have been shown in the paper. The notional model of ergonomic evaluation has been developed acknowledging that the proper notional pattern of the issue of the ergonomic evaluation is the classification problem. The evaluation system is the classifier whose input are the attributes of the evaluated workstation and output is its evaluation. The chosen attributes were presented in five thematic groups – work safety, work conditions, the influence of the physical work environment on an employee, psychophysical requirements and aesthetic requirements. Afterwards the transition from the informal conceptualization to the computer model i.e. representation in the formal and executable bayesian network language was made. Bayesian network in its visual aspect is an directed acyclic digraph in which nodes represent variables (both the attributes of evaluated workstation and the ergonomic evaluations) and arcs represent the types of connections between these variables. Taking into consideration the uncertainty of evaluation results which are due both to incomplete knowledge about the evaluated workstation and the vagueness of evaluation criteria it was probability which was accepted as the measure of uncertainty. The calculation schema implemented in the net and based on the theorem of conditional probability and Bayes formula enabled the introduction of qualitative and quantitative conclusions with the accuracy up to the probability distribution.
Designed demonstration modules were verified evaluating three chosen kinds of workstations in the agricultural and food industry. The developed method of ergonomic evaluation of workstations with the use of bayesian network as the system of knowledge representation has got the possibility of machine learning on the empirical examples as well as automatic reasoning in order to generate answers to the questions:
– What mark doest the workstation get?
– What attribute values should the workstation have to get the required mark?
– Which attributes have the greatest influence on the mark?
– What is the sensitivity of the mark on the attributes of evaluated 
The suggested method can be built into the process of a company management in order to automate ergonomic evaluation of workstations in agricultural and food industry and can be also used to evaluate workstations in the process of operating and to help to design them.