Nitin Padhiyar

Lexicographic optimization based MPC: Simulation and experimental study

Abstract Multi-variable prioritized control study is carried out using model predictive control (MPC) algorithms. The conventional MPC algorithm implements multi-variable control through one augmented objective function and requires weights adjustment for required performance. In order to implement explicit prioritization in multiple control objectives, we have used lexicographic MPC. To achieve better tracking performance, we have used a new MPC algorithm, by modifying the lexicographic constraint, referred to as MLMPC, where tuning of weights is not required. The effectiveness of MLMPC algorithm is demonstrated on a PMMA reactor for controlling the number average molecular weight and the reactor temperature. We have also verified the benefits of proposed algorithm on an experimental single board heater system (SBHS) for controlling temperature of a thin metal plate. These simulation and experimental studies demonstrate the superiority of the proposed method over conventional MPC and lexicographic MPC. Finally, we have presented generalized mathematical solutions to the optimization problem in MLMPC.

Multi-objective control of a fed-batch bioreactor using shrinking horizon MPC: A case study

This work focuses on offline and online multi-objective control of a fed-batch bioreactor for the induced foreign protein production by recombinant bacteria. Initially, open loop multi-objective control problem is formulated and solved using a single objective optimization after augmenting the individual objectives. The weighting parameters in the augmented objective function represent their priorities. The lexicographic optimization approach has been utilized for the multi-objective control in the MPC framework. Moreover, unlike the continuous processes, the batch processes operate for a definite batch time. Hence, the shrinking horizon approach along with the economic MPC framework is employed in the fed-batch bioreactor control.

Experimental study: Neural network based model predictive control of a distributed parameter system

An experimental set up has been designed for the purpose of spatial property control of a distributed parameter system. The setup consists of a thin metal chip with four temperature sensors along the length of the chip and four electric heaters to regulate the spatial temperature profile of the metal chip. A sequence of random pulse input signals with varying sampling times for the four manipulated inputs were used for live data generation of the transient temperature values at four locations along the metal chip length. Two data-driven models, namely linear state space model and neural network based model are identified and applied for online control. Model predictive control was used to control the spatial temperature profile for set-point tracking and disturbance rejection.

Lexicographic Optimization based Model Predictive Control: Application to Quadruple Tank System

Lexicographic optimization based model predictive control is a modified Model Predictive Control algorithm (MPC), which incorporates explicit controlled output priority by solving a series of optimization problems at every iteration in order of priority. We present an application of controlling liquid levels in quadruple cylindrical tanks using lexicographic optimization based MPC and compare the results with conventional MPC approach. The results show the superiority of lexicographic optimization based MPC over conventional MPC in terms of setpoint tracking. Extended Kalman filter is used as the state observer for the simulation results in this work for nonlinear quadruple cylindrical tank system.

Alien Genetic Algorithm for Exploration of Search Space

Genetic Algorithm (GA) is a widely accepted population based stochastic optimization technique used for single and multi objective optimization problems. Various versions of modifications in GA have been proposed in last three decades mainly addressing two issues, namely increasing convergence rate and increasing probability of global minima. While both these. While addressing the first issue, GA tends to converge to a local optima and addressing the second issue corresponds the large computational efforts. Thus, to reduce the contradictory effects of these two aspects, we propose a modification in GA by adding an alien member in the population at every generation. Addition of an Alien member in the current population at every generation increases the probability of obtaining global minima at the same time maintaining higher convergence rate. With two test cases, we have demonstrated the efficacy of the proposed GA by comparing with the conventional GA.