Tua Agustinus Tamba

A linear programming approach to network utility maximization

This paper describes a linear programming (LP) approach for solving the network utility maximization problem. The developed approach is inspired by a convex relaxation technique from non-convex polynomial optimization methods. In contrast to most of the existing results where concavity of the networks utility function is often assumed, the proposed LP approach may still be used to solve the NUM problem even in the absence of such a concavity assumption. Although the presented LP approach is originally formulated to compute upper bounds for the global optima of the NUM problem, we illustrate through simulation examples that the obtained bounds often correspond to the exact global optima.

Controller design for active suspension system based on skyhook reference model

In this paper, an alternative design technique for vehicles active suspension system using model reference adaptive control method is presented. To verify the performance of the design controller, a skyhook damping system was adopted as the reference model. An adaptive controller for a quarter-size vehicles active suspension system is designed. The controller parameters are determined using Lyapunov method and can be tuned to precisely achieve the type of desired response. The simulation results show that an active suspension system with the designed control strategy is able to improve the ride comfort and the road holding, significantly, compared with the conventional passive suspension systems as well with semi-active suspension.

Process and controller configuration of high purity binary distillation columns in the purification stage of petrochemical plant

This paper reports the systematic selection of process and controller configurations of four distillation columns operated at purification stage in an olefin petrochemical plant. These columns were demethanizer, deethanizer, depropanizer #1 and depropanizer #2 columns, which separated ethane (C2H6) and propane (C3H8) from methane (CH4) and other heavier components (i.e. C4+). The interconnection of these unit processes propagate the noise and cause severe instabilities towards the final unit process. Therefore, the decision about the choice of the process and control configuration is important in order to ensure the process stability. Such a decision includes the pairing of the manipulated and the controlled variables, the control rules and the inclusion of surge tanks to buffer the disturbances that caused the mentioned instabilities. In particular, the assessment selected three out of six control pairings available for distillation columns. A Proportional — Integral (PI) controller was utilized for all control pairs, and tuned by the Ziegler-Nichols method. In addition, two surge tanks were added as buffers between the demethanizer, the deethanizer and the depropanizer columns. This approach yields the process configuration that was capable of absorbing the disturbance, ensuring a 98–99% product purity and keeping a bias under 2%.

Optimal control of a nonlinear cryogenic separation process via SDRE method

The cryogenic separation process is an air separation technology which is used to produce specific components of the air in gases or liquid form with high purity. This technology is frequently used in liquified natural gas (LNG) industries, nuclear isotopes separation and cryogenic fuels production for space shuttle. This paper proposes an optimal control design method for a nonlinear model of a cryogenic separation process. The method combines the state-dependent Riccati equation (SDRE) method and the sum of squares (SOS) optimization technique. The paper first characterizes a stabilizing control law which minimizes a quadratic performance index of the process in an exact manner. Due to the difficulties in synthesizing such an exact optimal control (i.e. due to the need to solve the SDRE), the use SOS optimization techniques for computing sub-optimal control laws is then introduced. The proposed method is illustrated through the design of a controller which maximizes the fraction of a particular isotope production in a cryogenic separation process plant.

Design of Neuro-Fuzzy based inferential measurement of stripper unit in a fertilizer plant

This paper proposes an alternative solution to overcome the difficulties in measuring the primary variable of a stripper unit in a fertilizer plant using Adaptive Neuro-Fuzzy Inference System (ANFIS) technique. Inferential measurement is a method to predict the value of the primary variable of the model generated by the input-output relationships of the process affecting the primary variable the process. Using the real-time operational data collected from stripper unit of the fertilizer plant, the technique was able to estimate the value of the primary variable (benfield solution) with error criteria (RMSE value) of 0.467 in the learning stage, and 0.447 at the validation stage.

On the complexity of using Handelman relaxation for solving polynomial optimization problems

This paper investigates the computational complexity involved when using Handelman relaxation method for solving multivariable polynomial optimization problems. In particular, the relationship between the degree of the Handelman representation of non-negative polynomial function over convex polytopic set and the number of decision variables required to prove its existence through linear programming method is characterized. A simulation result to help illustrate the computational complexity and rate of convergence of Handelman relaxation method for solving multivariable polynomial optimization problem is also presented.

Algorithmic modeling and analysis of nonlinear biological systems

This paper proposes an algorithmic modeling and analysis method to study the dynamics of nonlinear biological systems. Motivated by the recent use of piecewise affine models in reachability analysis of continuous dynamical systems, we propose a multi-affine approximation method to study biological system models defined on hyperrectangle. We show that such approximation is useful for constructing systems abstraction in form of finite state transition system models whose properties can be verified algorithmically using model checking formalism. An example use of the proposed method to study the robustness of a biological system model to parameter variation is presented.