[pt] A maioria dos estudos existentes em controle de qualidade de processos focam no desempenho de máquinas e ferramentas. Assim, estes já contam com bons métodos para serem controlados. Contudo, erros humanos em potencial estão presentes em todos os processos industriais que contenham a relação homem-máquina, fazendo com que a necessidade de se avaliar a qualidade do desempenho humano seja de igual importância. A abordagem para se avaliar quão suscetível à falha humana estão tais processos baseiam-se em probabilidades de erro, supondo que o desempenho humano funciona da mesma maneira que o desempenho de máquinas, ou em PSFs (Performance Shaping Factors), variáveis representativas de características de desempenho humano. Embora esta última abordagem seja mais eficiente, ainda existem críticas a sua falta de contextualização: tais características são avaliadas separadamente uma das outras, e independentemente da tarefa que o operador esteja realizando. Sistemas de Inferência Fuzzy (SIFs) permitem que variáveis lingüísticas sejam avaliadas em conjunto, isto é, passa a ser possível criar um modelo que assimile as nuances da variação do comportamento de um PSF concomitantemente com a alteração de outro PSF. Dessa forma, a caracterização da confiabilidade humana, considerando que diversos PSFs afetam no desempenho dos demais, pode ser satisfeita ao se fazer uso de SIFs interligados seqüencialmente - SIFs hierárquicos. Para se contextualizar a caracterização da confiabilidade humana por tarefa realizada, necessita-se que os PSFs pertinentes a cada determinada tarefa sejam medidos novamente e realimentados ao sistema (desenvolvido nesta dissertação). O SIF geral (composto por nove camadas de SIFs hierárquicos) foi testado com dados hipotéticos e dados reais de operadores e tarefas de uma empresa do setor elétrico brasileiro. Os resultados encontrados foram satisfatórios e evidenciaram que a Lógica Fuzzy, na forma de SIFs hierárquicos, pode ser utilizada para caracterizar a confiabilidade humana, com a vantagem de fazê-lo enquanto seu contexto é considerado. / [en] Most of existing studies in quality control focus on machinery performance. There are effective and advanced control methods to deal with that. However, potential human errors are present in every industrial process operated by humans. Therefore, evaluating the quality of human performance becomes as important as evaluate machinery s. The approach to evaluate how much processes are susceptible to human error are based on error probabilities, by assuming that human performance is similar to machinery performance, or on PSFs (Performance Shaping Factors) – variables representing human features. Although this based approach is more efficient, there are still criticisms about its lack of context awareness: those features are evaluated separately from one another, and regardless of which task the employee is performing. Fuzzy Inference Systems (FISs) allow linguistic variables to be evaluated simultaneously, thus making it possible to develop a method that gathers the nuances of behavioral changes of a PSF whilst another PSF varies. With this method, and considering that different PSFs affect the performance of others, human reliability can be assessed through the use of sequentially interconnected FISs – Hierarchical Fuzzy Inference Systems. In order to contextualize this assessment by tasks, each of the PSFs that affects each task will have to be measured and fed into the system (as developed within this dissertation) once per task and per employee. The main FIS (which contains nine layers of hierarchical FISs) was tested by using both hypothetical and real data from operators and tasks of a Brazilian electricity company. Results were satisfactory and attested that Fuzzy Logic, in the form of hierarchical FISs, can be used to assess human reliability, with the advantage of also taking the context into account.

The presented doctoral thesis is focused on assessment of human factor impact on safety of industrial furnaces. Industrial furnaces are classified as machinery and belong to a group of industrial thermal equipment. The operation of industrial furnaces is burdened with the risks which the manufacturers and the furnace operators realize. The first part of the thesis presents an analysis of the current situation of legislation and scientific knowledges in the field of assessment of human factor impact on safety of industrial furnaces. In this part of the thesis the issue of human factor in other industrial branches: chemical industry, aviation and nuclear industry is also described. On the basis of conducted research the main aim of the thesis was established: preparation of a methodology for assessment of human factor impact on safety of industrial furnaces. Secondary targets were also listed. The second part deals with the assessment of risks connected with the operation of industrial furnaces and the assessment of human factor impact on safety of industrial furnaces. The furnace safety requirements, the process of furnaces risk assessment, the methods used for risk assessment and problems connected with the risk assessment are described. This part of the thesis explains the concept of human factor, presents a classification and description of the methods used for human reliability assessment, as well as the factors influencing the reliability of the operator. The third part of the thesis contains a proposed methodology for assessment of human factor impact on safety of industrial furnaces. Within the methodology performance shaping factors are stated, qualitative and quantitative assessment of human factor impact on safety of industrial furnaces is made and the system integration of the knowledges into the developed methodology is proposed – qualitative model for improvement of system state. The methodology is further verified on a real equipment – a hardening furnace.