| Overview | Travelers | WIP | SQC/SPC | MFG.Specs | Quality | Features | System | ||||||||||||
| Consider | Bar Codes | Users | Equipment | Pricing | |||||||||||||||
MES Considerations | ||
|
Considerations For Manufacturing Execution System Implementation .: This article first appeared in FabTech It is perhaps unfortunate and likewise quite telling that quantification of Manufacturing Execution Systems (MES), is really just, so much black art. It is not that the information available is fraught with large holes in the knowledge base that must be filled in with speculation and experience, but rather that the experience, expectations and true needs of a foundry implementing an MES system are very difficult to pin down. They are in a constant state of flux and unless some overriding themes can be culled from the moving targets, significant dilution of the Implementation Teams efforts will occur. This is not to say there are no examples of well implemented systems, running well within the constraints put upon them; there are. However, for every success story, you can find multiples of others where the Grand Scheme just isn't playing out as expected. And it can be hard to ferret out this kind of information. Consider the personal consequences of a $2million dollar project with a $1million dollar maintenance budget being declared by the implementation staff as: not really what was expected; and that the gains hardly seem worth the cost and effort to keep it alive. This kind of statement does not do well for career paths of the upwardly mobile MIS staff. And yet in spite of this, in fact, flying in the face of this, MES systems are becoming more critical to a foundries success every day. The amount of information available and the amount of information necessary to run a plant can at times be overwhelming. MES systems definitely have their place in this type of environment. Yet in order to not get burned with a mediocre implementation you must begin by examining both your needs and your expectations. MES systems are arguably one of the more complex Computer Information Systems in existence. This is not to say that you can't think of several other systems that you personally consider more complex, perhaps you can. And it's not that the code is necessarily complex either, but rather that the whole system itself is. The interaction of all the components is critical to the net result. And what is that Net Result? It is Information presented on-line to the Production Operator and to the desk of Manufacturing Management. So what becomes complex is the necessary checks and balances required of such a system to assure the users that the information they are viewing is accurate and that the database does not have slowly creeping and growing data integrity issues. Some of today's systems have multiples of millions of lines of Cobol and Fortran Code being executed and you end up having to worry about the stability of such a system. These are not unjustified worries. In general, an MES systems costs are not strongly linked to the size of a plant or the volume of a foundries production. Yes, there are sometimes a per user or per workstation cost that are a piece of the financial analysis but they are usually a small piece of the total picture when all is said and done. MES systems that run on large main-frame / mini-frame types of hardware have significant costs and support issues that have a life all their own, independent of the MES software running on them. These costs are very similar whether helping to run the new billion dollar fab or the 50 million dollar fab producing 5 micron CMOS in volume. When a plant purchases a piece of production equipment, that equipment has an inherent ability to produce product. The costs are leveraged by the capacity to produce. Buy more equipment and produce more product. A typical Dec/VAX/VMS implementation of an MES system on the other hand has large up front and continuing support costs that are for the most part in no way related to plant capacity. And here in lies the rub for most of the mid size and smaller facilities. How do they justify the very large installation / implementation / support costs when they essentially are paying a price that is similar to that paid by a larger foundry producing 5 times more product than they are. And on the other hand how do they dare not to implement. That is in fact what has been going on for the longest time in the medium and smaller fabs. They run with home brewed systems or no systems at all rather than incur the costs of a typical MES implementation. O f the smaller and mid scale foundries that do implement, how successful are they? If we look at our own market research it becomes clear that only the fundamental pieces are ever really brought on-line in most cases. The more esoteric features of say, automatic wafer processing equipment program downloading that is linked to a bar code on a cassette of wafers almost never gets implemented, except in the largest of fabs. And even then there is the perception that the effort to build, maintain and the constant need to verify the link is more costly than the gain it is supposed to produce. This is an example of the complexity of MES systems. You must be able to trust that the program being downloaded is the correct one else your "lights out" facility always has people running around checking, defeating the whole purpose of automation. The biggest gains to be made by an MES system comes in the form of the fundamental pieces of information that previously were collected and assimilated by hand. This is especially true in the smaller fabs. It is clearly an out of kilter situation when 80% of a systems implementation costs are related to the top 20% of it's functionality, functionality that is infrequently used and perhaps never trusted. So what then are the key concerns when implementing an MES system? Lets start with the term itself, Manufacturing Execution Systems (MES). The name itself tries to be all encompassing, all knowing, all powerful, all capable. It sounds unlimitless and as you might predict those are the kinds of resources it takes to implement it. If a system is to be truly MES in nature that implies an openness of sorts. How does one integrate so many pieces under one roof? Most likely it is not done in a proprietary framework but rather in one that is expandable, one that is by nature a piece of a larger whole. This is the very nature of client/server architectures today. The data being generated by factory floor operators and collected by WIP tracking routines does not belong to the software. In the final analysis if you are truly going to have a MES system then you must begin by realizing that first the data belongs to the organization, not just to the software that is generating it and second, that it takes an open system and architecture to accomplish this. So once again lets ask the question, what then are the key concerns when implementing an MES system in a second tier fab facility? Again my answer is similar. What was wrong with the term Shop Floor Control Systems? Why is there this constant creep of acronyms? Why are we trying to integrate JIT automatic photoresist reordering from our suppliers into our MES systems, while the fundamentals of product movement from workstation to workstation still challenge our operators abilities to easily interface with it and quickly obtain meaningful information from it? The fundamentals must come first and that is what the key concerns must be.  Real-time WIP On-line bar code driven travelers On-line wafer manufacturing specifications and drawings linked to the parts being produced Integrated SPC data collection, monitoring and warning.
For the small to medium fab to get seduced by all the other possibilities when the fundamentals are not rigorously covered is to spend time on outcomes that will ultimately be swamped by the core issues. Sure it' great to have a system re-order your photoresist for you. Sure you have lost money in the past when you ran out unexpectedly. But those costs are second and third order effects compared to the losses that occur when 300 wafers are exposed and etched with the wrong mask set because real-time information necessary for the operator to do their job was not available and the operator, doing what operators do, made an educated best guess and was wrong. Besides overspending for secondary functionality, there is a second major roadblock to successful MES implementation. And that is not knowing your factory. This weakness usually surfaces most frequently on a foundries first MES implementation. There is significant information that is carried around inside of the heads of critical and usually unrecognized people in a fab. It is not necessarily the Mfg. Mgr. who has all the system configuration answers. It is frequently the expediter, or production control clerk or wafer fab lead operators or street smart engineering technicians who really know how a facility is run. More frequently it is a synthesis of information from these employees that truly give an overall complete picture of a facility. These types of employees carry around large amounts of undocumented information in order to perform their tasks and keep their immediate supervisors happy. It is the unspoken collective consciousness that frequently has many of the answers to the MES configuration and implementation questions. To not identify and include these people in the planning and implementation stages is to unnecessarily handicap your efforts. In conclusion then: Stay focused. Don't let yourself get seduced by the possibility of also including the kitchen sink in your implementation. Utilize your resources, namely your undocumented collective knowledge about how you plant runs and is organized. Choose an open kind of systems architecture like client/server networks to implement your system on and in. Choose an open systems software that allows for the data to be used across your entire organization. As Published in SemiConductor FabTech May 9th About the author
|
Copyright © 1989 - 2008 Santa Barbara Analysis.
All rights reserved.
Home About Us wiptrac Test Drive Support Contact Us
Terms of Use - Privacy Policy
48@!%7470th%2-63
