all are welcome series/past...these boundary layer fluid flows considered maxwell, williamson,...

INSTITUTE OF MATHEMATICAL SCIENCES

UNIVERSITY OF MALAYA

COLLOQUIUM SERIES

Title : The Development of MatHeuristics for Production-Inventory-

Distribution Routing Problem

Speaker : Dicky Lim Teik Kyee, ISM, UM

Date : 23- 02- 2018 (Friday)

Time : 10 - 11 AM

Venue : MM3, Institute of Mathematical Sciences

ABSTRACT

Production, inventory and distribution are amongst the most important components of a supply chain network.

Integrating production, inventory and distribution decisions is a challenging problem for manufacturers trying

to optimize their supply chain. In general, the problem of optimally coordinating production, inventory and

transportation is called the production-inventory-distribution routing problem (PIDRP). The PIDRP model

considered in this study includes a finite planning horizon, a single production facility, a set of customers with

deterministic and time-varying demand, and a fleet of homogeneous capacitated vehicles. The aim of solving

the model is to construct a production plan and delivery schedule which minimizes the overall costs while

fulfilling customers’ demand over the planning horizon. We propose an optimization algorithm designed by the interpolation of metaheuristics and mathematical programming techniques, known as MatHeuristics

algorithm, to solve the model. In this study, we develop two different two-phase approaches within a

MatHeuristic framework, namely MatHeuristic1 and MatHeuristic2. In MatHeuristic1, Phase 1 solves a mixed

integer programming (MIP) model which includes all the constraints in the original model except the routing

aspects. The routing constraints are replaced by an approximated routing cost in the objective function. A

variable neighborhood search (VNS) is proposed in Phase 2. The VNS is employed to improve the solution and

it incorporates some aspects of Tabu search to escape from local optima. Whilst in MatHeuristic2, we adopt a

slightly different approach. The algorithm starts with the pre-processing stage where the routes are determined

to act as input to the MIP in Phase 1. Phase 2 uses VNS and we utilize both phases in an interactive manner.

The models in both approaches are solved by using Concert Technology of CPLEX 12.5 Optimizers with

Microsoft Visual C++ 2010. Both algorithms are tested on a set of 90 benchmark instances with 50, 100 and

200 customers and 20 periods (Boudia et al., 2007) and the results are competitive when compared to the Memetic Algorithm with Population Management (MA|PM) (Boudia & Prins, 2009), Reactive Tabu Search

(RTS) (Bard & Nananukul, 2009) and Scatter Search (SS) (Moin & Yuliana, 2015). MatHeuristic1 performs

well in large instances when compared to other metaheuristics but MatHeuristic2 performs well in smaller

cases.

All are Welcome



COLLOQUIUM SERIES Title : DETERMINISTIC AND STOCHASTIC INVENTORY ROUTING

PROBLEMS WITH BACKORDERS USING ARTIFICIAL BEE COLONY

Speaker : HUDA ZUHRAH AB HALIM, ISM, UM

Date : 23 - 02 - 2018 ( Friday)

Time : 11 - 12 AM


ABSTRACT

This thesis is devoted to solving the inventory routing problem (IRP) and its variants. The inventory

routing problem comprises the coordination of two components: inventory management and vehicle routing

problem. The details of the component define the variation of IRP. Three different variants of IRP were studied

in this thesis. First variant is a many-to-one distribution network, consisting of a single depot, an assembly

plant, and geographically dispersed suppliers where a capacitated homogeneous vehicle delivers a distinct

product from the suppliers to fulfil the deterministic demand specified by the assembly plant over the planning

horizon. The inventory holding cost is assumed to be product specific and only incurred at the assembly plant.

A metaheuristic, Artificial Bee Colony (ABC) algorithm is proposed to solve the problem. The ABC were

tested on existing dataset and compared with two other metaheuristics: Scatter Search (SS) and Genetic

Algorithm (GA). A statistical analysis was carried to shows the difference between algorithms is at 95% level

of significance. An enhanced ABC were also developed, which performs better in terms of quality as compared

to the previous ABC, SS and GA. IRP with backordering (IRPB) represents the second variant of IRP explored in this thesis. The IRPB define the conditions where unsatisfied demand will be delayed and fulfilled in future

period. The distribution network of IRPB consists of a single supplier and geographically scattered customers,

where a set of vehicle performs the delivery to fulfil customer’s demand. The backorder decisions considered

here depends on two situations, first, if vehicle capacity is insufficient to satisfy a customer demand, and

secondly if the saving in the transportation cost is larger than the backorder penalty imposed by a customer. An

ABC algorithm with two different inventory updating mechanisms is proposed to solve IRPB. The two

mechanisms are random exchange and guided exchange. Results from both algorithms are compared with lower

bound found by CPLEX and from the literature. The final variant investigated is IRP with stochastic demand.

Two main characteristics of the demand are, first, the demand is known in a probabilistic sense, which in our

study the demand follows the binomial and uniform probability distribution and secondly, the demand is

dynamic since it is gradually revealed at the end of each period. The problem is known as dynamic and stochastic inventory routing problem (DSIRP). In DSIRP, we consider a distribution network consists of a

supplier and a set of retailers. An order-up-to level (OU) inventory policy is applied, and each unit of positive

inventory will incur a holding cost whilst a penalty is incurred for each negative inventory level. We assumed

that the transportation of the product from supplier to the retailers is handled by a third party. A matheuristic,

combination of Mixed Integer Linear Programming (MILP) and hybrid rollout algorithm is proposed. The

control in each state is obtained either from MILP or ABC algorithm. The DSIRP is then modified to be able

to handle backorder decision (DSIRPB). A new formulation of MILP model is presented. Both DSIRP and

DSIRPB were tested on existing instances.

All are Welcome



COLLOQUIUM SERIES

Title : Lazy cop number and other related graph parameters

Speaker : Sim Kai An, ISM, UM

Date : 21- 03- 2018

Time : 3.00-4.00 PM


ABSTRACT

The game of cops and robbers is a two-player game played on a finite connected

undirected graph G. The first player occupies some vertices with a set of cops, and the second

player occupies a vertex with a single robber. The cops move first, followed by the robber.

After that, the players move alternately. On the cops’ turn, each of the cops may remain

stationary or move to an adjacent vertex. On the robber’s turn, he may remain stationary or

move to an adjacent vertex. A round of the game is a cop move together with the subsequent

robber move. The cops win if after a finite number of rounds, one of them can move to catch

the robber, that is, the cop and the robber occupy the same vertex. The robber wins if he can

evade the cops indefinitely. The cop number of G, denoted as c(G), is the main graph

parameter in the game of cops and robbers. Here, we investigate the cop number and lazy

cop number of a graph G, the minimum order of graphs for small value of cop number and

the capture time. Our results focused on a variant of the game, the lazy cops and robbers,

where at most one cop moves in any round. Burning a graph is a process defined on the vertex set of a simple finite graph. Initially,

at time step t = 0, all vertices are unburned. At the beginning of every time step t ≥ 1, an

unburned vertex is chosen to burn (if such a vertex is available). Thereafter, if a vertex is

burned in time step t - 1, then in time step t, each of its unburned neighbours becomes burned.

A burned vertex will remain burned throughout the process. The process ends when all

vertices are burned. The burning number of a graph G, denoted by b(G), is the minimum

number of time steps required to burn a graph. We give a survey on some known results of

burning number of certain graphs and present the bounds on the burning number of the

generalized Petersen graphs.

All are Welcome



COLLOQUIUM SERIES

Title : Effective Portfolio Investment Strategy

Speaker : Prof. Dr. Tan Ken Seng (University of Waterloo), Canada.

Date : 29- 08- 2018 (Wednesday)

Time : 1400 -1500


ABSTRACT

The optimal construction of portfolio investment strategy has remained a fascinating topic

since the ground breaking work Markowitz in 1952 who pioneered the mean-variance efficient

portfolio. More recently, the 1/N strategy, commonly known as the naïve investment strategy, has

been gaining popularity due to its simplicity and its effectiveness. Numerous studies have shown

that none of the commonly known investment strategies could consistently outperform the naïve

strategy. Motivated by such studies, we propose a new investment strategy that is based on the

notion of effective portfolio dimension. Extensive empirical studies are provided to demonstrate

that the resulting effective portfolio strategy has the potential of outperforming the naïve strategy.

All are Welcome



COLLOQUIUM SERIES

Title : STAGNATION-POINT FLOW OF NON-NEWTONIAN NANOFLUIDS WITH ACTIVE AND

PASSIVE CONTROLS OF NANOPARTICLES

Speaker : NADHIRAH BT ABDUL HALIM, ISM, UM

Date : 05- 09- 2018 (Wednesday)

Time : 3.00-4.00 PM


ABSTRACT

In this seminar, newly upgraded non-Newtonian nanofluids models near a stagnation point are proposed under

the influence of active and passive controls of the nanoparticles. These boundary layer fluid flows considered

Maxwell, Williamson, second-grade, Carreau and Powell-Eyring non-Newtonian fluids. The flows are represented

by the conventional partial differential equations in fluid dynamics added with unique expression of stress tensor

in the momentum equation which satisfy the continuity equation for conservation of mass. The Buongiorno’s

model is used as a base model in this analysis as it takes into consideration the effect of Brownian motion and

thermophoresis of the nanoparticles in the energy and mass transport equations of the flows. All these equations

are reduced into a set of simpler partial differential equations via boundary layer approximation. The governing

equations are later converted to a set of nonlinear ordinary differential equations by using similarity

transformation. Shooting technique is employed to reduce these resulting equations into a set of boundary value

problem in the form of nonlinear first order ordinary differential equations subject to the specific initial and

boundary conditions which reflect the effect of active and passive controls of the nanoparticles in two different

occasions. The bvp4c function, developed based on finite difference method by MATLAB is utilized to further

solve the newly upgraded Maxwell, Williamson, Carreau and Powell-Eyring models while the BVPh 2.0 package

in Mathematica is employed to solve the newly upgraded second grade nanofluids flow model. The effects of

active and passive controls of the nanoparticles are compared graphically and tabularly. The influences of other

considered parameters towards the flow profiles are also presented while the numerical values of skin friction

coefficient, Nusselt number and Sherwood number are listed. The stagnation parameter increases the heat

transfer of all the non-Newtonian nanofluids flows studied. Furthermore, the heat transfer rate of the boundary

layer flows under passive control of nanoparticles is consistently higher in magnitude as compared to the ones

under active control of nanoparticles.

All are Welcome

all are welcome series/past...these boundary layer fluid flows considered maxwell, williamson,...

Documents