IFORS 2011

July 10-15, 2011

Australian Society for Operations Research
Melbourne Chapter


Home | Contact | Program | Newsletter | Membership | OR in Melbourne | Links | Archive

2009 Program

The default venue for the monthly lectures is
AMSI Seminar Room
ICT Building
The University of Melbourne
111 Barry Street
Melbourne time

Scheduled Events
November 25
Full day, at RMIT
TBA Mini-Conference: Recent Advances in Operations Research
June 17, 6:00PM Ramamohanarao Kotagiri Contrast data mining: methods and applications
May 27, 6:00PM Anthony Bedford Simulation in Sport
April 15, 6:00PM David SierA model for stockpile blending with two components of uncertainty
March 18, 6:00PM Leonid ChurilovStroke Chain of Survival and Recovery — The scope for "The Science of Better" to do much better!
March 18, 5:30PMAGM


Consult the campus map for the exact location of the building. The ICT Building is Building 105 in block P14 on the map. It is shown as building A on this Google satellite image:

And here is a photo of the entrance to the building

View Larger Map

Date: Wed, November 25, 2009, Time: 9:00AM - 5:00PM
Location: RMIT, Building 8, Floor 9, Room 66

Recent Advances in Operations Research

Program: See details here (PDF file)

List of presentations: (Show/Hide all abstracts)

  • Rail Schedule Optimisation In The Hunter Valley Coal Chain
    Gaurav Singh, Andreas Ernst, and David Sier
    Show/Hide Abstract
    The Hunter Valley Coal Chain (HVCC) is the largest coal export operation in the world with total export of more than 90 million tonnes of coal in 2009. HVCC contains over 30 mines and the coal is exported by HVCC Logistics Team (HVCCLT) using over 20 coal load points spread across the region. Recently the CSIRO has developed a decision support tool to assist schedulers of HVCCLT in developing optimal schedules for rail operations. The tool reduces time taken to develop these schedules, and its fast execution time also allows schedulers to test the outcomes of various planning strategies before finalising the schedule. In this talk, we present the underlying mathematical model implemented in this decision support tool along with results from computational experiments. A demonstration of this decision support tool will also be presented.

  • Supply Chain management for Hunter Valley Coal Industry
    Riley Clement
    Show/Hide Abstract
    Australia is the world's largest exporter of coal and is also home to Port Waratah Coal Services (PWCS), who export more coal by volume than any other export coal terminal in the world. PWCS service the Hunter Valley Coal Chain, which consists of 35 mines spread over 350km in the Hunter Valley. Managing the supply chain requires the planning and coordination of train and ship movements, subject to a number of operational constraints. This task is critical as increasing coal export is limited by the capacity of the supply chain. Two mixed integer programming (MIP) models are developed for producing a detailed rail and port schedule --- a time indexed formulation, and a positional date and assignment formulation. The performance of these models is compared using historical data provided by PWCS, and Gurobi, a commercially available solver for MIP optimisation problems.

  • Why do I have to wait so long at the emergency department?
    A Ceglowski, R Sultana
    Show/Hide Abstract
    Despite years of research, applied projects and lots of hard thinking y bright people, the problems of sporadic long waits for attention persist at most hospital emergency departments (EDs) around the world. Long waits are symptomatic of deeper ills such as ED overcrowding, exit blockage (when the associated hospital is unable or slow to admit ED patients) and ambulance bypass (when ambulances have to be diverted to alternative EDs because the target Ed is unable to deal with new cases). Long waits might also affect the health and recovery time of the patient. We have learnt a fair bit about the dynamics of EDs over the years. For instance, long waits are often triggered when patients are slow to leave the ED, often because there is not a bed available in the hospital to admit them to. Many EDs have implemented "short stay" units where patients can be "admitted" in order to overcome the problem of matching patients to ward beds. It's readily apparent that this is akin to increasing ED resources, so the measure works to some extent, until demand once again overwhelms available resources. While the matter of hospital bed occupancy is very relevant to the ED state, it is "outside" the EDs control. Long waits for service are very obviously a resource issue (under the assumption that the clinical work is performed efficiently), so the question becomes one of balancing limited resources against the risk of triggering a queue for service. Measures to counter the problem of long wait times have concentrated on altering the queuing characteristics of the ED from a single preferential queue where patients are constantly rearranged according to triage class (most urgent patients "jump the queue" to be treated first) to multiple servers where non urgent patients also receive prioritised treatment ("fast track" initiatives), to other variations of multiple server systems based on their treatment (multitrack systems). As OR people we know that the most efficient use of resources is the single queue/multiple server option, so these options are just playing around the edges. Better forecasting of future demand is one of the ways of reducing uncertainty where scarce resources have to meet fluctuating demand. Many researchers have followed this route. This paper will demonstrate how attempts to forecast patient arrival rates are bound to failure because the ED environment is inherently chaotic. It describes the complexity of ED queue composition and how this impacts upon patient wait times.
    Presentation file (PDF, 750KB)

  • Heuristics for Project Scheduling with Stochastic Crashing Durations
    Heng-Soon Gan, Charles Hu and Tony Wirth
    Show/Hide Abstract
    The Minimax Theorem is the most recognized theorem for determining strategies in a two-person zero-sum game. Other common strategies exist such as the maximax principle and minimize the maximum regret principle. All these strategies follow the Von Neumann and Morgenstern linearity axiom which states that numbers in the game matrix must be cardinal utilities and can be transformed by any positive linear function f(x)=ax+b, a>0 without changing the information they convey. This talk describes risk-adverse strategies where the linearity axiom may not hold. Examples are given from zero-sum games, Prisoner's Dilemma from non-zero sum games and the Nash arbitration scheme.
    Presentation file (PDF, 356KB)

  • Applying Risk Theory to Game theory
    Tristan Barnett
    Show/Hide Abstract
    The Minimax Theorem is the most recognized theorem for determining strategies in a two-person zero-sum game. Other common strategies exist such as the maximax principle and minimize the maximum regret principle. All these strategies follow the Von Neumann and Morgenstern linearity axiom which states that numbers in the game matrix must be cardinal utilities and can be transformed by any positive linear function f(x)=ax+b, a>0 without changing the information they convey. This talk describes risk-adverse strategies where the linearity axiom may not hold. Examples are given from zero-sum games, Prisoner's Dilemma from non-zero sum games and the Nash arbitration scheme.

  • Solving stochastic scheduling problems by dealing with uncertainty one component at a time
    Geoff Robinson
    Show/Hide Abstract
    We are interested in obtaining good solutions to large scheduling problems such as stockyards for bulk materials that involve many sources of uncertainty and are too large to be solved exactly. One way to make progress is to approximate the losses associated with schedules by considering the effects of one component of uncertainty at a time. This is computationally cheaper than simulation. In our first example, we find an approximate solution to the scheduling of one job on one machine in the presence of three sources of uncertainty. The final example involves scheduling 50 jobs on 10 machines, including features such as: jobs which must be run on particular machines, precedences between jobs, uncertain release dates, and uncertain process times. Solutions based on deterministic evaluation, simulation and one component at a time are compared and discussed.

  • A dynamic programming approach to the solution of infinite-dimensional linear programming problems with Toeplitz structure
    Robin Hill
    Show/Hide Abstract
    I will show how to construct the optimal solution to an infinite-dimensional linear programming problem with constraints described by a linear time-invariant dynamic system. The non-linear dynamics driving the optimal solution are identified explicitly. I work backwards in time from a known point on the optimal trajectory, using standard dynamic programming ideas strengthened by the use of duality. The optimal solution can be pictured geometrically as evolving on the boundary of a convex, compact, centrally-symmetric body in Euclidean space.

  • Alternative Energy Planning using Mixed Integer Prgramming
    James Foster
    Show/Hide Abstract
    The expanding interest in renewable and alternative energy sources has suggested the need to find useful tools for planning new infrastructure. In the area of distributed generation, the traditional model of one power plant with many consumers has given way to a bilateral power flow model of interchanging demand and source nodes over the electricity distribution network. Our research has focused on using mixed-integer programming for the allocation of small-scale generators across a power system network with distributed generation in order to optimize energy efficiency. We compare solutions of a model involving a mixed-integer nonlinear program (MINLP) with nonconvex quadratic constraints to models with approximate linearised convex constraints - on both the speed of solution and quality of network designs obtained - for a set of networks well-known in the power engineering literature ranging from 6 to 57 nodes. The programs are implemented in the AMPL system with the MINLP solver Bonmin.

  • A Tale about Tails
    Alan Brown
    Show/Hide Abstract
    An investigation into a flawed model in sport exposed a fallacy caused by the use of an asymptotic formula in a situation where it produced unreliable results. At the same time an investigation into unsatisfactory investment results arising from the impact of the GFC was undertaken. By mixing the two lines of study, it was found that "Balanced Superannuation is not balanced". The wider implication, that other financial model results based Quadratic Programming may be unreliable, will be discussed.

  • A Critique of Info-Gap Decision Theory: From Voodoo Decision-Making to Voodoo Economics
    Moshe Sniedovich
    Show/Hide Abstract
    The title of this presentation is borrowed from a book that I am writing on my effort over the last six years to contain the spread of Info-Gap decision theory in Australia. However, the main question that I address in this presentation is not discussed in this book. Rather, it is one of the main questions addressed in my other book on this topic, which is tentatively entitled "The Rise and Rise of Voodoo Decision-Making". The basic question is this: given the very harsh and detailed criticism of this theory that is freely available and easily accessible to the public and which shows that this theory is a classic example of a voodoo decision theory, how is it that this fundamentally flawed theory is still promoted from the pages of respectable refereed journals? I address this fascinating question from an Operational Research perspective.
    Presentation file (PDF, 715KB)

  • On a Primal Proximal Point Heuristic and the Feasibility Pump in Discrete Optimization
    Andrew Eberhard
    Show/Hide Abstract
    We provide some mathematical analysis that models the feasibility pump heuristic as a discrete version of a primal-proximal point algorithm involving a polyhedral norm. This provides model algorithm closely related to the feasibility pump. We peruse this paradigm to develop a number of other model algorithms closely related to the feasibility pump strategy. This is work in progress and awaits numerical experimentation.

  • Student Projects: Reflections on 35 years Experience
    Dudley Foster
    Show/Hide Abstract
    Apart from 6 years as an academic, the author has spent most of the past 35 years as an OR practitioner, both within a corporate OR departments and as an external consultant. However, over the past 10 years, he has also developed a role as a specialist in the supervision student projects undertaken for organisations external to the university concerned. As might be expected, he has found that the issues pertaining to the management of client relations for such projects are very similar to those he encountered within a corporate OR department. The talk concentrates on experiences gained on the Masters in Logistics Management at RMIT (delivered in Hong Kong as well as Australia, the MBA at Victoria University (in both Kuala Lumpur and Melbourne) and, most recently, for the MSc in Operational Research at the University of Edinburgh. One of the keys to success with external projects, as with all projects is thorough problem definition and the author is a strong advocate for requiring students to produce a Scoping Document, which is not just something to be marked, but a commitment to provide constructive actionable written feedback from both the client and the supervisor.

Date: Wed, June 17, 2009, Time: 6.00pm.
Location: ICT Building, The University of Melbourne, 111 Barry Street, Carlton.

Topic: Contrast data mining: methods and applications

Speaker: Ramamohanarao Kotagiri, University of Melbourne.

Abstract: The ability to distinguish, differentiate and contrast between different data sets is a key objective in data mining. Such an ability can assist domain experts to understand their data, and can help in building classification models. This presentation will introduce the principal techniques for contrasting different types of data, covering the main data set varieties such as relational, sequence, and graph forms of data and clusters. It will also focus on some important real world application areas that illustrate how mining contrasts is advantageous.

Note the change in the venue of the lecture
Date: Wed, May 27, 2009, Time: 6.00pm.
Location: Room B113 at 221 Bouverie St, Carlton

Topic: Simulation in Sport

Speaker: Anthony Bedford, RMIT.

Abstract: Predictions in sport not only provide great challenges to the budding statistician, they provide great interest to the keen sports follower; a better chance to the punter; a different angle to the pushy publicist; a keener angle to the sports commentator; curiosity to the maths student and a novel story for the journalist. In this presentation, the multitude of questions from these sources are outlined. The fascinating problems arising from these questions are then defined. The success and failures through differing methods of data fitting and prediction are covered with a specific focus on AFL football, the Melbourne Cup and badminton. Treatment is also given on how simulation methods provide fascinating hypotheses to problems difficult to solve through conventional probabilistic methods, and how OR practitioners can utilise such problems to promote their field.

Date: Wed, April 15, 2009, Time: 6.00pm.
Location: ICT Building, The University of Melbourne, 111 Barry Street, Carlton.

Topic:A model for stockpile blending with two components of uncertainty

Speaker: David Sier, CSIRO.

Abstract: We look at the task of assembling a shipment using material stored in a set of separate stockpiles. The average grade of the parcel being assembled should be close to product specifications and is to be achieved by blending raw materials from the stockpiles that each have their own grade characteristics.
We formulate a version of this problem considering uncertainty about the average grade of material in each stockpile and allowing for the fact that portions taken from stockpiles will generally have grades different from the stockpile average grades. This formulation of the problem can be readily solved, despite non-convexity of the objective function.

Date: Wed, March 18, 2009, Time: 6.00pm.
Location: ICT Building, The University of Melbourne, 111 Barry Street, Carlton.

Topic: Stroke Chain of Survival and Recovery — The scope for "The Science of Better" to do much better!

Speaker: Leonid Churilov, National Stroke Research Institute and the Dept of Maths and Stats, University of Melbourne

Abstract: Stroke is responsible for 9%-14% of all deaths around the world and is one of the two most common causes of death. In Australia, approximately 50,000 people suffer a stroke each year and it is estimated that over 250,000 people live with the consequences of stroke, many surviving with permanent disability. The direct and indirect lifetime costs of first-ever strokes amount to approximately $2.1 billion per annum. In adults, with every decade of life, the risk of stroke doubles and due to the ageing of population, the burden of stroke is anticipated to increase significantly.
There have been many recent advances in treatmentsfor patients with acute stroke that have improved survival and reduced disability. Rapid diagnosis and management of stroke provides the best opportunity for clinicians to intervene early when combined with coordinated multidisciplinary stroke care and includes stroke rehabilitation. Stroke Chain of Survival (American Heart Association) captures the value-adding nature of stroke care and includes the following key value-adding steps: detection (recognizing the onset of signs and symptoms); dispatch (calling ambulance and having emergency medical services dispatched immediately); delivery (transporting patient to hospital with assessment and care); triage (immediate emergency department triage); data gathering (prompt laboratory and diagnostic imaging studies); decision (diagnosis and decision about appropriate therapy); treatment (administration of appropriate drugs and other acute care intervention); and rehabilitation (maximum feasible restoring of functional status).
Stroke Chain of Survival provides a virtually boundless scope for an integrated clinical, experimental, and decision making research effort in order to improve the outcomes for stroke victims. In this presentation we discuss how Decision Sciences can effectively contribute to this effort, using the activities of the Division of Statistics and Decision Support of the National Stroke Research Institute as an illustration. We argue that the major challenge facing Decision Sciences in this context is not necessarily in developing new or more sophisticated techniques, but in using currently available techniques in appropriate and persistent way to make much more of an impact on the delivery of stroke care.

The next big event

IFORS 2011 Conference
July 10-15, 2011, Melbourne, Australia