Search

Laboratory incubators are used to maintain and cultivate microbial and cell cultures. In order to ensure suitable growing conditions and to avoid cell injuries and fast rise and settling times, minimum overshoot and undershoot performance indexes should be considered in the controller design for incubators. Therefore, it is important to build proper models to evaluate the performance of the controllers before implementation. In this study, we propose an approach to build a model for a laboratory incubator. In this approach, the incubator is considered a linear time-invariant single-input, single-output system. Four different model structures, namely auto-regressive exogenous, auto-regressive moving average exogenous, output error and Box–Jenkins, are applied for modeling the system. The parameters of the model structures are estimated by using prediction error methods. The performances of the model structures are evaluated in terms of mean squared error, mean absolute error and goodness of fit. Additionally, residue analysis including auto-correlation and cross-correlation plots is provided. Experiments are carried out in two scenarios. In the first scenario, the identification dataset is collected from the unit-step response, while in the second scenario, it is collected from the pseudorandom binary sequence response. The experimental study shows that the Box–Jenkins model achieves an over 90% fit percentage for the first scenario and an over 95% fit percentage for the second scenario. Based on the experimental results, it is concluded that the Box–Jenkins model can be used as a successful model for laboratory incubators. [ABSTRACT FROM AUTHOR]

Laboratory incubators are used to maintain and cultivate microbial and cell cultures. In order to ensure suitable growing conditions and to avoid cell injuries and fast rise and settling times, minimum overshoot and undershoot performance indexes should be considered in the controller design for incubators. Therefore, it is important to build proper models to evaluate the performance of the controllers before implementation. In this study, we propose an approach to build a model for a laboratory incubator. In this approach, the incubator is considered a linear time-invariant single-input, single-output system. Four different model structures, namely auto-regressive exogenous, auto-regressive moving average exogenous, output error and Box–Jenkins, are applied for modeling the system. The parameters of the model structures are estimated by using prediction error methods. The performances of the model structures are evaluated in terms of mean squared error, mean absolute error and goodness of fit. Additionally, residue analysis including auto-correlation and cross-correlation plots is provided. Experiments are carried out in two scenarios. In the first scenario, the identification dataset is collected from the unit-step response, while in the second scenario, it is collected from the pseudorandom binary sequence response. The experimental study shows that the Box–Jenkins model achieves an over 90% fit percentage for the first scenario and an over 95% fit percentage for the second scenario. Based on the experimental results, it is concluded that the Box–Jenkins model can be used as a successful model for laboratory incubators.