Monday, November 17, 2008

AMHS Issue and a Proactive OR Approach

Moving materials in a complex manufacturing environment is a difficult task. If you consider an average of 600-step process route requiring material movement between every step, managing those vehicles sometimes could be more difficult than managing the value added production itself. In contrast to machines, material handling vehicles are not fixed, they move around. Modeling them as another regular resource is not an option in this case.

Actually, it is not always the modeling or algorithm problem. Many times, it may be the infrastructure, which is beyond an Operations Research analyst's view. In many cases, it is hard to track vehicles or provide sufficient communication capabilities between machines, jobs and vehicles.

However, showing that an optimization based solution could make the overall system work better than the best solution that can be found within current infrastructure forces the hardware suppliers to design their equipments with better communication or tracking capabilities. This scenario is one of the many more examples existing in the real world, in which an OR expert can take proactive steps to change the environment to a better condition to work on. An optimization model is not always supposed to formed within the given constraints, it can also create its own constraint set.

Monday, November 3, 2008

Taking advantage of the market crash

A crash in the financial market is not essentially bad for everyone. There are a number of industries that thrive in such an environment. For example, the repossessed vehicle market is overflowing with inventory. While this may hurt banks and other financial institutions that are willing to auction off these vehicles to cut their losses, there are many people out there who wouldn't mind driving a nicer car without having to pay the price, well, at least not the sticker price...

According to npr, "vehicle repossessions are up 10% this year" and approximately 1.6 million vehicles are estimated to be repossessed by the end of the year. As a result, some lending companies are starting to 'adjust' the loans with the borrowers instead of repossessing their vehicles.

From the lenders point of view, an interesting optimization problem would be to identify the optimal number of vehicles to repossess versus adjust. A number of lending companies might be interested in such a problem due to the complexity of this decision. The amount of money saved due to repossessing the vehicle should offset the opportunity cost of receiving payments from the buyer as well as the logistics and inventory cost of transporting and holding repossessed vehicles. Such an optimization problem would also incorporate a 'risk' factor that includes the probability that borrowers may in fact pull themselves out of the economic trough sooner than later.

Food for thought...

Where does academy and industry differ?

Both academy and industry aim to produce work for the good of humanity. Although underlying personal aims could be different, such as greed to publish papers, higher profits, etc., they all satisfy in some demand due to human beings' needs or questions.

The methods that are used by both theoretical and practical sides support themselves, they all root back to scientific method. While academic production has very detailed reviewing steps, it is easier to assess the output of an industrial work. While many less credited academic works show great benefits in the industry, many academic milestone works cannot find a chance of direct implementation into practice.

Both industry and academy need themselves to survive. There would not be any problems to solve if there was not industry or industry would fail due to myopic solutions if there was not more global insight of the academy.

But where does an academic work differ from an industrial work? Even enabling a very highly complex system to work with little flaws needs a lot of intellectual work in itself. Everyday practice translates into improvement in many cases. However, this may not be considered as credible academically. Academy needs a new problem to be defined, or a new method to be developed to solve an existing problem or a better method than existing solution methods to solve an existing problem. Someone other than the researcher gives the final decision how academically credible the work the researcher has done. There are not always objective evaluation methods as opposed to evaluation methods that industry has. While an expert can consider a work substantially contributory, another cannot see any contribution. It is a hard job for both the researcher and the referee to decide on the quality of a paper.

Wednesday, October 29, 2008


Preventive maintenance is a type of maintenance that is done to prevent faults before they occur randomly. It is simply done to replace unscheduled maintenance due to unexpected breakdowns. Unexpected behaviors in a production environment creates variability, and variability is an unwanted effect since it indicates that some things are running out of control at the shop floor, which will hurt due dates, plans, etc.

Preventive maintenance can be simply checking if everything is fine with the equipment, cleaning, or replacing a part before it fails. It creates a down time in the machine as an unexpected breakdown does. It means both preventive maintenance or unscheduled maintenance uses some capacity of the equipment. In terms of using capacity, preventive maintenance and unscheduled maintenance are not very different than normal production process except that they do not produce any throughput.

So the difference between maintenance and production is throughput in terms of capacity usage. The difference between preventive maintenance and unscheduled maintenance is controlability. If we assume that we have an hour of preventive maintenance and an hour of unscheduled maintenance, actually they both take the same time, so they use the same capacity on an equipment. But while we can control where we can allocate that one hour capacity use when we prefer a preventive maintenance, if we don't do that then that one hour can steal capacity of the equipment randomly any time as an unscheduled maintenance.

Then, while we can plan preventive maintenance with other production jobs, unscheduled maintenance is out of control. Or, is it? Let's leave it to discuss some other time.

Monday, October 27, 2008

Why reducing variation?

Variance reduction is one of the core issues of simulation. For a reliable mean estimation, it is crucial to narrow down confidence interval as much as possible and the way is variance reduction, either by increasing run time or increasing number of replications. That, I understand for academic purposes. But why is that so crucial for the industry? Not for system-wide variation reduction my concerns are but for the simulation itself. Isn't the simulation a reflection of the real-life system? And aren't real-life systems under so many variability effects?

Take a semiconductor fab as an example. In such a complex environment with so much variability factors, if I see a smooth WIP profile, that concerns me more than seeing some fluctuating crazy WIP behavior with respect to validity of the simulation model. Who can argue that one can witness such a smooth WIP behavior in a real fab?

Having a smooth WIP profile by manipulating simulations (if you're not targeting a future ideal fab) hide problems while they are real and existing in the real-life system. Anyways, I never understand why industry is caring so much about the mean results of simulation performance. Why not asking for variation results? Yes, a high WIP or cycle time should tell something about high variation in the system but what is going to be the reference point for a simulation run that needs to be updated everyday for dayly schedule asessment? Evaluating the importance of simulation needs to be revisited by the industry and it should be more around the variation analysis rather than the interest in mean performance results.

Saturday, October 25, 2008

What does "optimization" mean?

The use of the word "optimize" dates back to 1857 in English and it means "to make as perfect, effective, or functional as possible". The word "optimization" was derived from "optimize" and means "an act, process, or methodology of making something (as a design, system, or decision) as fully perfect, functional, or effective as possible ; specifically : the mathematical procedures (as finding the maximum of a function) involved in this". The words "optimum", "optimize", "optimism", "optimist", "optimal", and "optimization" are all related were derived from each other at some period of time. I can refer more interested people to Online Etymology Dictionary.

That's enough with the etymology of the "optimization" related words. Let's come back to today and see how it is used. OR people know the meaning of the word "optimization" by their heart, "
the study of problems in which one seeks to minimize or maximize a real function by systematically choosing the values of real or integer variables from within an allowed set".

Apparently, there is a small difference in understanding this definition between OR people and people from other disciplines. Especially, the most difficult times to communicate between happens when an OR academician faces a non-OR industry person and tries to talk about about "optimizing" a system. The word "Optimal" means "the best with respect to a defined system" for an OR specialist, while it sounds more like "acceptably better" for an industry expert. It is an ironic representation of misevaluation of OR among the industry. It puts a responsibility to the shoulders of operations research experts and industrial engineers that they should devote some of their energy to tell what they are doing for science, for industry, and for life itself in general for an average person can understand well in a cople of minutes. Actually, there are very nice OR definitions that says what it means in a couple of very clear sentences but communicating the messag these definitions mean is a life-long struggle. It might be really good to take start with explaining what "optimal" means to a non-OR colleague of ourselves clearly and making sure he does not forget it by practicing it "optimally" at our work and research.

Friday, October 24, 2008

ISMI Symposium

This past week Austin has experienced another annual ISMI symposium at Hilton Airport hotel. Many members of semiconductor industry came to the city to share their work and create new networking opportunities. While there were short courses and pre-symposium events on Monday and Sunday, main presentations and panels were held on Wednesday and Thursday.

I had a chance to make an half-an-hour simulation demonstration in a simulation short course on Tuesday. It was just about showing how an AutoSched AP model looks like for people who are unfamiliar with it. AutoSched AP is probably one of the most widely used simulation packages in the semiconductor industry. It all started for automation industry but developed in semiconductor business by Brooks company and now is being supported by AMAT after Brook's acquisition to the company. It has a spreadsheet user interface without any graphical fancy stuff as opposed to other simulation packages such as Automod or Arena, but is the fastest and to run for semiconductor manufacturing. It has flexibility for customization by some knowledge of C++ coding for people who are not happy with its standard properties and extensions. It may not be the ideally fastest simulation package but is accepted as one of the best around for semiconductor manufacturing until a better ones come into competition.

On Thursday, I have been to a panel on the futur of semiconductor manufacturing in G7 countries, which was directed by Don Hutcheson of VLSI, by the participation of expert people of semiconductor industry. To summarise, panelists foresee that it is going to survive in G7 countries one way or another. While Asian countries are serious competitors for G7 countries, due to unique characteristics of semiconductor manufacturing (low labor, high investment requirements, etc.) they do not hold a greater threat in the future. One of the most important ways to hold semiconductor manufacturing profitable in G7 countries is government subsidies. Also, new technologies promise great opportunities ahead for semiconductor industry, such as solar and led.

I forgot to say that the food was good, too. I especially liked the vegetable lasagna served for lunch on Thursday.

Thursday, October 23, 2008

Science 2.0

Last Tuesday (October 23, 2008), I attended a talk by Don Hutcheson, CEO of VLSI Research Inc., which is hosted by Robert S. Strauss Center for International Security and Law in LBJ Library of University of Texas at Austin. Hutcheson, and his small but effective company VLSI serves as a big database for semiconductor industry. The title of the talk was "Science 2.0: Globalized Innovation in Electronics".

Science 2.0 is about applying principles of Web 2.0 to scientific research. Having presented transistors to public use and henceforth made the Internet available to world's rapidly globalizing communication and sharing needs, apparently semiconductor industry itself does not utilize it for its own benefit by the means of research and innovation as much as expected. Especially, research conducted in corporate labs have a very highly probable chance of staying in corporate's library shelves for years and nobody would hear of it for ages. As Web 2.0 applications are all about interaction, Science 2.0 model enables people share their research, innovations through web. Those days are not very far that people will work from their home using communication tools (video conferences, phone conferences, Internet, etc.) via sharing virtual offices.

The future of semiconductor industry lies in research and innovation. Sharing information, interaction of minds and collaboration will bring high quality and effective solutions. You can see Don Hutcheson's efforts in Science 2.0 at

It was not really clear to me how Science 2.0 reflects on manufacturing other than design but Hutcheson said he was going to give a talk on this subject in another conference soon. I am looking forward to learning about it afterwards.