By putting into practice of optimization methods and implementing optimal solutions, many operational problems can be solved. However, real life problems are more complicated than the theoretically stated pure base mathematical models in some books or publications.
Do you have any success stories for solving small or large-scale real life problems?
Well, as your question is extremely wide in scope it is - i think - impossible to answer it in general. Instead, I can offer a discussion about a case study. Our optimization team recently had an industrial PhD student working on the scheduling of a multi-task cell at her work-place (GKN Aerospace in Trollhättan, Sweden). She devoted a very large portion of her time to modeling this multi-task cell, and the activities that could - and should - be optimized, as activities in a combinatorial optimization problem. A main problem is to learn what the industry wants - this is not always well understood by managers, in fact. Here, the goal was to find good enough schedules (not specifying the precise objective) that could be produced fast (within a quarter of an hour). Take a look at the PhD thesis here: http://publications.lib.chalmers.se/publication/181850 - the link also provides the individual papers. Let me know if the link does not work.
I believe that it is a success story, as it is established, I think, that provably good solutions can be produced in the time requested using mathematical optimization. Now, this does not solve the problem for the whole factory, as there are many machines, and what is produced also changes over time, meaning - perhaps - that at certain times some resources should be perhaps even be moved in order to facilitate more efficient flows through the factory. But it is a very nice first start ...
Hi Turkay,
I think your question is very broad, and any specific answer will only count towards certain problems. I have solved a 1 billion decision variable problem relating to the placement of wind turbines and solar panels across the contiguous USA using in depth weather data. The problem was solved using the GAMS/CPLEX software, however, it took a huge amount of tweaking and experimentation to get the code written in a fast solvable way.
The problem I solved is a convex optimization, which is important, and I solved it with LP via the barrier method. The simplex and dual were slower for my specific problem. In practice, if you can write the mathematical constraints and show that they are convex, you should be able to solve very large problems in polynomial time - however, this is not always guaranteed with the simplex algorithm. I think that CPLEX is the best option on the market, but it can be expensive. It has multiple options for different optimizations too. There are plenty of others out there, but I have not tried them all.
I hope that this helps and you find the best option for you!
In case of Vehicle Routing Problem (VRP) which is very popular in the filed of supply chain, i think Trans CAD is more appropriate
The model will always be a simplifcation of the real problem. How deep the simplification will be, depends on you.
For my calculations I use usually Excel. I know, it's slow, and the Excel's accuracy is not perfect. But in Excel you can easily simulate the explored problem. Look at my wrts, if you want to check out what you can do in Excel - it is about internal transport in the warhouse.
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2354827
I agree with Michael Patrikkson. The key is the mathematical structure of the optimization problem. MANY such applications occur in operations management and in the supply chain. Your question is so broad as to defy a succinct answer. Can you narrow down the type of application and/or optimization model that you have, sir? All the best,
Harold Benson
Emeritus Professor
Musician Now!
Dear Harold,
More specifically, but it is still broad, I would say the problem of berth allocation at container terminals. For instance, ship arrival times are uncertain. The crane operators’ performances vary. The objective is to minimize the time for loading and unloading of containers from/to vessels and thus to minimize waiting times for vessels at the berth in order to serve more vessels.
On the other hand, as far as optimized operations concerned, berth area operations cannot be considered apart from yard area operations. Yard area operations are also in connection with the other terminal processes. Some uncertainties in one process affect many other processes.
Narrowing down the optimization model? Is it possible to address real life problems by setting up few objective functions with some constaints?
Kind regards,
Let me give some references (those who like an apology if I partially refer to my own work; sorry for that):
doi:10.1016/j.ejor.2009.05.031 (specified, 2010)
doi: 10.1007/s00291-007-0100-9 (general, 2008)
doi: 10.1007/s00291-003-0157-z (general, 2004)
In the early 1990s we had done a berth planning prototype for a major terminal operator. Based on that experience as well as the experience gained from then on up to know I must confess that a lot of the academic literature calls itself "applied" but reveals that they probably never worked on a terminal (sorry for being rude; of course there are academics out there who know a lot about these problems in the real world). Nevertheless, it is possible to get appropriate optimization models done and also to incorporate practical constraints (like time windows, tidal conditions, maintenance of infrastructure, etc.) as well as integrate with other problems (e.g. crane split) and objectives (related to minimizing the time in port of a vessel, or reliability issues, etc.).
- Badges
- Science topic
- Similar topics
- Names