Jun 9, 2010

Set-based design of capital-intensive manufacturing systems

Robert Simonis, director of global lean manufacturing for a division of Lear, makes a living spending money -- big money -- designing and installing manufacturing systems around the world. At the end of the day, however, his capital investment decisions have to pay off for the company. As he’s seen how lean is affecting production thinking, he’s also become more and more conscious of the constraints imposed by upstream decisions made in product and process design.

Addressing a Detroit SME meeting last month, he said we don’t yet have good names for lean in the design of production processes and accompanying systems design. Product design has the “design-for” tools -- design for manufacturability, design for reliability, design for maintainability, and so on. But what would “design for lean” or “design for continuous improvement” be all about?

Process design was once the bailiwick of manufacturing engineering alone, to be done after the product design was tossed over the proverbial wall by engineering. Many traditional manufacturing engineering and process design rules are based on mass production. They anticipate large returns on capital investment once volume achieves the breakeven point, the point where costs are recovered and the payback period starts.

That scenario is right for high-volume production, yet Simonis said that it can result in some bad bets being placed. For one thing, demand, when it must be predicted years before the fact, determines designed-for capacity, which in turn determines size, speed, complexity and cost of the equipment selected. And in Simonis’s experience, demand is sometimes estimated way too high.

Traditional methods may cause process designers to greatly underestimate variation in demand. If they design enough capacity to handle the expected peak in year three and it doesn’t pan out, the result will be overbuilding and overspending.

Their $1 million robotic system might reduce labor by four workers, but the cost of capital may not receive enough consideration. Process designers and manufacturing engineers need better information about interest and other costs of servicing the debt on the capital investment. They need methods of process design for capital investment, taking in how long equipment is going to be used, recovery time, and payback time.

Circumstances turn on the question, “How do you anticipate the future?” Besides what traditional methods give process designers, there are a number of other variables in the real world that impinge upon how well a process design proves out:

    Speed to market
    Niche markets
    Mass customization
    Variation in mix
    Variations in volume (short term, week-to-week)
    Variations in volume (long term, annual)
    Risk to capital
    Cost of capital
    Automation vs. labor

Rather than following formulas, Simonis sees process design for lean in terms of several goals:
    Minimize total cost and risk
    Produce profit
    Satisfy customers, both external (buyers of final product) and internal (employees)
    Safe process
    Good product
    Stockholder/stakeholder interests
    Support frequent change

To disrupt old manufacturing engineering routines, Simonis said we need to start process design and development with a concept earlier -- soon after the product concept is developed, not after the prototyping and piloting happens.

Product:    concept→    prototype→    pilot→   
Process:        concept→    prototype→    pilot→


Models that are often used to guide process design include replacing automation with low-cost labor, replacing people with machines, or outsourcing. In making decisions for where and how to produce a product, companies often believe that one of those models is the best practice. In Simonis’s experience, none of the three is necessarily the right strategy. Any one of them could be correct, depending upon circumstances.

To make better decisions, Simonis recommends turning to the concept of set-based design. His planning teams start decision making with a few scenarios. (Senseis say to start with seven sets, because the team usually has to really stretch to find that many, which stimulates creativity.) Which one will be the choice depends on the situation.

In set-based design, you start with multiple possibilities, develop them, and winnow them down:

----------\
----------  --------\
----------  --------  ---------\
----------  --------  ---------  --------X
----------  --------  ---------/
----------  --------/
----------/

Simonis describes three that could apply:

Set A
The first option might be an integrated system capable of immediately satisfying estimated peak demand. The rule of thumb is usually to design for 85% of forecast. Simonis said his own experience suggests that the customer actually orders at 70% of forecast, so with the figure that comes from a financial model, you may have overcapacity. The fully-integrated manufacturing system has another problem. The more complex it is, the less the inherent efficiency, because any failure will stop the whole system and there are no substitutions. Over time, components stop operating the way they do when new, for example, conveyer performance tends to degrade rapidly

Set B
A second concept might be a U-shaped cell with modular components and flexible staffing. You may start with the assumption that the customer will take 70-80% of forecast. You will then have the option to add capacity only at bottlenecks, and subtract people on the downside of fluctuating demand rate. Of course, the plan will have problems if it doesn’t properly allow for rapid expansion if necessary.

Set C
The third scenario might be to design for many small cells. In fact, that gives you more opportunity for continuous improvement, because each cell will develop ideas to apply to each new one as you set it up with growing capacity. Your initial proposal will show lower but more consistent returns rather than a big return after breakeven is reached. As with B, contingencies like rapid demand expansion need to be anticipated.

Simonis favors considering the third approach, if not as a solution, at least as an alternative. Working it out among several scenarios before fixing minds on Set A will bring out ideas. Even if Set A is the choice, the idea generation in the process will likely make for a more flexible integrated system.

People ask why they should belong to a professional engineering or business organization. This meeting is an example of when local organizations do their best. Robert Simonis is a member with special expertise, sharing it with his fellow SME members, without any financial motive. If a membership fee of $100/year or so seems like a lot of money, do the math: it works out to less than $10 a month. And a meeting like this is priceless.

Robert Simonis is Director of Global Lean Manufacturing, Electrical Power Management Systems Division at Lear. His comments represent his own views, not any official statements of Lear Corporation.

No comments:

Copyright @ 2005-2014 by Karen Wilhelm