Mohd Azri Abdul Aziz

An improved event selection technique in a modified PSO algorithm to solve class scheduling problems

In solving class scheduling problems, among the initial steps taken by optimization algorithms is to generate initial solutions. These initial solutions which are random or feasible will be improved iteratively to obtain better solutions. In this iterative process, the re-assignment of events to a new time slot and the selection of event to be reassigned plays an important role in assuring the reassignment process is able to improve the quality of the timetable. This paper proposes an event selection technique namely Event Selection based on Selection Limit (ESSL) applied in a modified PSO algorithm to solve class scheduling problems. This technique was based on Tabu List used in Tabu Search Algorithm. Using this technique, the selection of event in each iteration step will be limited based on percentage selection limit. Five percentage selection limits plus adaptive limit were tested in order to see their effect towards the final penalty of the solution. The performance of the proposed technique was measured based on percentage penalty reduction (%PR) and execution time. Five sets of data from International Timetabling Competition were used in the experiment. The experimental results shows that ESSL with adaptive limit managed to produce the highest percentage of penalty reduction. Furthermore, this technique also managed to reduce the effect of early convergence which is known to be one of the weaknesses of the original PSO.

Assignments acceptance strategy in a Modified PSO Algorithm to elevate local optima in solving class scheduling problems

Local optima in optimization problems describes a state where no small modification of the current best solution will produce a solution that is better. This situation will make the optimization algorithm unable to find a way to global optimum and finally the quality of the generated solution is not as expected. This paper proposes an assignment acceptance strategy in a Modified PSO Algorithm to elevate local optima in solving class scheduling problems. The assignments which reduce the value of objective function will be totally accepted and the assignment which increases or maintains the value of objective function will be accepted based on acceptance probability. Five combinations of acceptance probabilities for both types of assignments were tested in order to see their effect in helping particles moving out from local optima and also their effect towards the final penalty of the solution. The performance of the proposed technique was measured based on percentage penalty reduction (%PR). Five sets of data from International Timetabling Competition were used in the experiment. The experimental results shows that the acceptance probability of 1 for neutral assignment and 0 for negative assignments managed to produce the highest percentage of penalty reduction. This combination of acceptance probability was able to elevate the particle stuck at the local optima which is one of the unwanted situations in solving optimization problems.

The effect of varying inertia weight on Particle Swarm Optimization (PSO) algorithm in optimizing PID controller of temperature control system

The implementation of PSO as an optimization algorithm in finding the optimal values of PIDs parameters has attracted a significant interest among researchers. One of the important parameters in PSO algorithm is inertia weight (ω) which directly affects the percentage of previous velocity used in determining the next position of particles. A high value of ω will force the particles to fly with a significant influence of previous velocity while with a low value of ω, current velocity will contribute more to the particles trajectory. This paper studies various types of inertia weight to be implemented with PSO in optimizing PID controllers parameters for a temperature control system. Among popular types of inertia weight are fixed, linearly decreasing, exponentially decreasing and stochastic. The experimental results shows that the Linearly Decreasing Inertia Weight (LDIW) approach managed to improve the performance of PID controller as compared to several other types.

The effects of event selection based on soft constraint violation (ESSCV) in a modified PSO algorithm to solve class scheduling problems

Generating an initial solution and then iteratively improve the solution are common initial steps in solving class scheduling problems. In this improvement process, the re-assignment of events to a new time slot and the selection of event to be reassigned play a significant role in reducing the soft constraint violation and finally improving the quality of the timetable. This paper proposes a local search heuristic which handle an event selection namely Event Selection based on Soft Constraint Violation (ESSCV) applied in a modified PSO algorithm to solve class scheduling problems. This technique was based on the idea of re-assignment of event which contributes to soft constraints violation, hoping the new assignment will alleviate the violation. The performance of the proposed technique was measured based on percentage penalty reduction (%PR) and execution time. Five sets of data from International Timetabling Competition were used in the experiment. The experimental results shows that ESSCV managed to reduce the event contributed to soft constraint violation, thus increases the %PR even though also demonstrated a small increase in execution time.

Moment of inertia of a DC motor as significant factor on the performance of PSO algorithm utilizing WTRI based fitness function

The implementation of PSO algorithm in optimizing PID controllers parameters has been in active research area recently. One of the popular plant to be controlled using this control strategy is a DC motor and various types of DC motors with variations in parameters value have been used in this research area. Moment of inertia is one of the important parameters of a DC motor which can affect the output transient response including rise time, settling time, overshoot and steady state error. This paper investigate the moment of inertia as a significant factor on the performance of PSO algorithm utilizing the weighted transient response index (WTRI) based fitness function. Some popular weight combinations of WTRI were used in the experiment with 3 DC motors with the same specifications except for the moment of inertia. The experimental result concluded the existence of some kind of threshold moment of inertia where the performance of PSO algorithm was not affected by any variation of weight combinations in WTRI based fitness function if the moment of inertia of the DC motor is less than the threshold value. With moment of inertia above this threshold value, the significant variation in the performance of PSO algorithm with different weight combinations in WTRI based fitness function can be seen.

The significance of prioritized weight in WTRI based fitness function on PSO algorithm under different range of moment of inertia of a DC motor

In optimizing PID controllers parameters using any optimization algorithm, fitness function is one of the important area to be explored. Among frequently used fitness function is a weighted transient response index (WTRI) where the weight is given to each of the transient response component including rise time, settling time, overshoot and steady state error based on their priority set in specification of the control system. This paper analyzed the significance of 3 weight categories (2 prioritized and 1 non-prioritized) on each of the transient response components under different moment of inertia, represented by 3 DC motors. The result show a significant effect of both prioritized weight on the transient response component for a moment of inertia of 0.038 Kgm2 and above. Below this value of moment of inertia, no significant difference can be seen either on prioritized or non-prioritized weight categories.

The functionality validation of PSO algorithm in optimizing PID controller's parameters using multi-dimensional test functions

In recent years, PSO algorithm has been one of the popular soft computing technique and increasingly used in optimizing the parameters of PID controller. However, before PSO or any optimization algorithm in general can be successfully used to solve optimization problems, the validation of their functionality must be done earlier. This paper proposed the functionality validation of PSO algorithm which will be focusing on the fitness functions and convergence of the particles. Due to the existence of P (1 dimensional), PI or PD (2 dimensional) and PID (3 dimensional) controllers, the functionality validation has been done with 1, 2 and 3 dimensional problems. The test functions used are Rastrigin and Rosenbrock. Experimental results shows the proposed PSO algorithm coded in MATLAB has been successfully proven fit to be employed in optimizing 1, 2 and 3 dimensional problems which later can be used to optimize PID controllers and its variants with the same dimensions of problems.