Prediction Of Output Variable Research Articles

Development of an inverse simulation method for the analysis of train performance

Conventional methods of computer-based simulation allow prediction of output variables, often as a function of time, for a given model of a physical system for a given set of initial conditions and input variables. In the case of train performance simulation models, the possible output variables include train speed or distance travelled, both expressed as functions of time. The corresponding input variables, also expressed as functions of time, are the tractive force or power levels for given train characteristics and route information such as gradients, track curvature and speed restrictions. Inverse simulation methods, on the other hand, allow selected model variables (such as the tractive force at any time instant) to be found from other specified model variables applied as input (such as the train speed or distance travelled versus time) for a given set of route conditions and train characteristics. The specific inverse simulation method presented in the paper is based on feedback principles. Illustrative results are used to verify this inverse simulation approach for train performance applications, and further cases are used to show that the inverse formulation provides an insight that is different from that obtained using more conventional forward simulation techniques.

Aggregate models based on improved group methods for simulation and optimization of distillation systems

This note deals with aggregate models for complex distillation systems in large-scale flowsheets. Group methods were originally devised for simple absorber and stripper calculations with no major extensions for handling distillation. In this work, group methods are systematically analyzed and further improved by modifying some of the previously proposed approximations. As a result, the improved group method exhibits accurate predictions and this is demonstrated using simulation and optimization case studies for a variety of chemical systems and operating conditions. It is observed that the prediction of output variables is in close agreement with that of the rigorous equilibrium stage model. In case of optimization problems, the optimal number of trays and feed locations differ by only one or two trays. The aggregate model can be applied in a sequence of steps in order to improve the reliability and robustness of the solution procedure. A rounding heuristic is also proposed which can provide near-optimal solutions with a significant reduction in computational time.

Variable interaction network based variable selection for multivariate calibration

Multivariate calibration problems often involve the identification of a meaningful subset of variables, from a vast number of variables for better prediction of output variables. A new graph theoretic method based on partial correlations (variable interaction network—VIN) is proposed. Many well studied representative calibration datasets spanning different application domains are selected for investigating the performance. Partial least squares (PLS) regression models combined with variable selection techniques are employed for benchmarking the performance. Subsets of variables with different number of variables are retained for the final analysis after VIN selection and progressive prediction accuracies are used for comparison. VIN–PLS results show significant improvement in prediction efficiencies and variable subset optimization. Improvement of up to 45% over existing methods with significantly fewer variables is achieved using the new method. Advantages of VIN based variable selection are highlighted.