| Humanistic problems such as thermal comfort, stress and fatigue are usually vague, complex and difficult to define or to express exactly in subjective terms. Fuzzy set theory was developed to handle this subjective vagueness and neural network is ideal for improving the representation to overcome, at least partially, this inexactness. In this dissertation, the recently developed fuzzy adaptive network (FAN), which combines the advantages of both fuzzy set theory and neural network, is used to define and to model this type of humanistic problems.; Thermal comfort forms the primary influence on the comfort and efficiency of a worker. However, thermal comfort cannot be defined exactly. In fact, even the various factors, which influence this comfort, is not exactly known. An initial model based on experimental design by considering the various relevant factors such as air velocity, temperature, humidity, and clothing is first formed. Then, the network of FAN was applied to analyze and to improve the relationships between thermal comfort and the various relevant parameters.; Standing fatigue is associated with time, allowed movement if any, and posture. This fatigue can be measured by the volume change of the lower leg. The degree of fatigue is also a function of time. Thus, as a first step in forming the model, a neural network approach with feedback from one time step before was proposed and implemented to model this problem.; Human stress plays an important role in occupational safety, worker's health, and efficiency. The factors influence this stress are associated with time of work/rest, workload, personal life, personality, work environment, and social relationship. Various stress indices were used to investigate and to compare the different factors. The results by applying FAN indicate that individual differences cannot be overlooked, and both subjective and objective measures of human stress are needed.; Another problem considered is the credit rating/loan approval system, to which credit line or the amount of loan to approve is based. Human judgement plays an important role in the credit rating practice. The existing very approximate and vague model is modeled and modified by the use of the network FAN. The results show clearly the effectiveness of this approach.; The results show clearly that the neuro-fuzzy network forms an effective approach to model and to improve this type of humanistic systems, which are very vague, difficult to define, complex, subjective, and greatly influenced by psychological factors. |
