Icreasing The value of semiconductor plant assests with MES
01/01/1998
Increasing the value of semiconductor plant assets with MES
R. Sanborn Towle, Camstar Systems, Campbell, California
The value of a semiconductor manufacturing plant asset is directly related to its contribution to process repeatability and efficiency. A relatively small investment in manufacturing execution system (MES) software linked with business management systems such as enterprise resource planning (ERP) can dramatically increase this value. With the information obtained from these tools, manufacturers can increase profitability by optimizing both physical assets, such as plant and equipment, and intangible assets, such as vendor relationships.
Managing a semiconductor manufacturing process, some have said, requires a combination of art and magic. The art is the engineering management of hundreds of process steps involving millions of dollars worth of plant and equipment and thousands of process variables. The magic involves controlling a process that often appears to have a mind of its own, due to the large number of hard-to-control variables with the potential to affect process results. Everything from humidity, to the number of hours since each machine`s last preventive maintenance, to which operator works which shift can influence the semiconductor manufacturing process.
In this environment, achieving a level of process repeatability at which all plant assets operate at peak efficiency is the key manufacturing goal. While a process that operates within acceptable thresholds may be profitable, fully optimized plant assets provide the competitive advantage necessary for long-term success. MES makes a critical contribution to semiconductor manufacturing profitability by tracking previously hidden metrics - queue vs. in-process time, planned and unplanned maintenance, and operator identifier data, for example - that affect the ability of the resource to produce effectively. These metrics, captured on the shop floor as they occur, preserve a detailed history of actual events and provide the real-time data that process engineers need. Data from MES help engineers maintain the process within acceptable thresholds. MES software also passes process information upstream to the ERP systems used for enterprise-wide business management.
 |
Figure 1. Equipment utilization breakdown shows only 30% productive time.
How MES optimizes plant assets
Physical assets. The most cash-intensive asset in a semiconductor plant is the process equipment. Implanters, steppers, deposition equipment, and oxidation furnaces require initial investments ranging from $100,000 to several million dollars each. Rapid advances in semiconductor manufacturing technology require that these machines be periodically upgraded, retrofitted, and eventually replaced.
In addition to being complex and expensive, semiconductor manufacturing equipment is also just plain big. It requires lots of expensive, environmentally controlled floorspace. Floorspace costs combined with the cost of the initial purchase, ongoing maintenance, upgrades, and replacement of manufacturing equipment obviously make for an expensive physical asset.
Improved equipment utilization is the most immediate and cost-effective way to maximize fab output. According to Sematech, semiconductor equipment is actually productive only 30% of the time (Fig. 1). The remainder is spent in scheduled maintenance, unscheduled maintenance, engineering tweaking and testing, unscheduled shut-down, idle time, and set-up time.
Fabs today are approaching yield levels of 90%, so gaining another few percent is a complex and difficult undertaking with a relatively small return on investment. With a small investment in MES software, however, equipment utilization can be increased by approximately 20% with a significantly greater increase in overall output.
Equipment management by MES involves elements that on the surface might appear to be controlled by other systems. On closer inspection, the coarseness of those systems compared to MES becomes apparent. For example, the accuracy of semiconductor manufacturing equipment must be constantly maintained. Each machine has a recommended calendar-based preventive maintenance (PM) cycle. MES systems, however, can track the actual time in use of individual tools, thereby allowing more precise, usage-based maintenance cycles. Thus, MES can support PM cycles that only take equipment out of production when actual usage warrants.
Equipment utilization can also be tracked from a throughput perspective, measuring the queues for each piece of equipment to identify bottlenecks. Rerouting and reprioritizing work flow through a plant can increase equipment utilization at no additional cost, improving plant output and profitability. For example, MES data can indicate the queue time and processing time for each individual lot. Process steps that consistently report a higher queue time may suffer from a capacity shortage, and can be targeted for additional equipment purchases.
An MES system can also use utilization data to provide online planning and scheduling. Regardless of how good a material resource planning (MRP) system is and how carefully work flow is managed, unplanned events affect shop-floor operations: orders are canceled or postponed; material arrives late; or all the operators for a particular machine are unavailable. MES data, reflecting what has actually - as opposed to theoretically - happened, can be used to load models with actual average and standard deviation queue and processing times for particular product or process types. Because of the improved accuracy of the data used in its underlying models, online planning and scheduling software will more closely conform to the actual capabilities and characteristics of the manufacturing operation.
Record-keeping and reporting are critical to both equipment utilization and work flow. Tracking everything on the shop floor - work-in-process inventory, each lot through each operation, every machine`s maintenance history, every operator`s training and work history, plant environmental statistics, and more - creates a lot of data. To improve utilization, process engineers must know the percentage of time each piece of equipment spends actually processing material, which means tracking each lot through each substep in the process. By providing accurate data, MES applications can contribute to uptime and throughput, and make correlations such as the effect of individual operators on machine efficiency.
MESA, the MES developed and supported by Camstar, supports this level of tracking with its work-in-process, resource data collection, and quality data collection modules, and allows the interactive viewing and analysis of this data with a graphical reporting module. These modules combine lot, resource, and quality tracking to provide an accurate picture with the appropriate level of detail. The data shows when each lot completes each process step, how long it waits in each queue, and when it enters and leaves each substep in the process (such as the four substeps in the etch process). Process engineers can then determine what percentage of machine time was spent running, as opposed to waiting for an operator to load material, change sputter targets, or handle maintenance tasks (see Fig. 2).
 |
Figure 2. Graphical representations of accumulated MES data illustrate plant resource utilization.
MES tools can also manage other physical assets, such as interim inventories of substrate wafers and process consumables. These inventories are expensive to maintain, not only due to the cost of inventory, but also because products have a limited shelf life. Any product that expires becomes worthless because it does not meet the dates required for a vendor`s certificate of compliance.
While too much inventory adds needless cost, too little inventory may result in shortages. Sufficient inventory is critical in the semiconductor industry, because long manufacturing lead times prohibit a make-to-order strategy. Fab customers might even select another source rather than wait for the lead time required for make-to-order. While an ideal inventory includes the precise amount of product required exactly when it is needed, that level of control is difficult to attain.
MES helps fabs maintain the delicate balance by determining, on a case-by-case basis, whether it is better to maintain a sufficiently large inventory to avoid any shortage, or whether it is more cost-effective to risk occasional shortages that force an extended schedule for a small percentage of customers. Decisions are based on analyses of work-in-process (WIP), forecasting arrivals, and amounts of product available or currently in production. Correlations of actual on-time delivery results for different inventory level scenarios can be combined with MES WIP processing data to determine optimal inventory models. In developing these optimizations, MESA supplements manufacturing models and projections with real performance data from the shop floor. For example, while the manufacturing model may say that cycle time for a particular lot should be 30 days, real-world influences such as equipment problems and scheduling changes may give an actual cycle time of 35 days.
The software tracks factors such as actual equipment usage/process step to determine actual material and resource usage/step. This information is extremely useful for interim inventory decisions. If customer demand increases, for example, engineers can check WIP and know exactly how long material will take to reach stock, instead of merely checking a standard cycle time value. Then, they can use dispatching and lot-scheduling modules to match interim inventories to cycle times. Dispatching modules can alter lot priorities, while lot scheduling modules can start lots at particular times based on queue times and similar factors.
Intangible assets. In addition to easily measurable assets such as equipment utilization and inventory, MES helps optimize such related, but less tangible, assets as subcontractor relationships. Wafer growth, electrical test, final test, assembly, die preparation, packaging and other processes are commonly subcontracted, often to plants outside the US. Often, manufacturers will also subcontract the more complex and expensive processes, such as implant. Because subcontracted processes are often complex, and because quality means profits, good relationships with good subcontractors are a measurable benefit. Understanding manufacturing and business issues from the subcontractor`s perspective is a big part of maintaining a good relationship.
For example, on-time delivery of promised work allows the subcontractor to optimize resources. Erratic requirements add costs to the subcontractor`s operation; these are passed along to the manufacturer. Late delivery to the subcontractor might also cause product to miss a reserved slot, further delaying delivery of finished goods.
Good subcontractor relationships can generate volume discounts and other economic incentives for continued business. Subcontractors also control more difficult-to-measure elements such as acceptance of one manufacturer`s job instead of another. A subcontractor might even turn down a job from a manufacturer with a history of unreliability.
By identifying bottlenecks and shortages, MES can help ensure that a subcontractor has work as contracted, even if schedules or other factors must be rearranged. MES can also track history information, such as what material the manufacturer has shipped to each subcontractor, when material was shipped, and subcontractor performance data showing how often completions were late or failed to include required paperwork.
Personal productivity of operators and process engineers is an often-overlooked factor. Operator productivity can vary dramatically because incorrect or outdated procedures contribute to scrap and rework. MES overcomes many potential "operator-error" mistakes by providing real-time instructions, specifications, revisions, and procedures. Having this information immediately available not only eliminates potential errors, but saves time by providing immediate answers to procedural questions.
Similarly, process engineering productivity is an extremely costintensive asset because the decisions made by process engineers directly and dramatically affect process results. The better the process engineer`s decisions, the higher the profitability of the production run. One of the key elements of process engineering productivity is the availability of data. Three types of data reporting can help optimize personal productivity: standard, ad hoc, and analysis.
Standard graphical reports and real-time inquiries across shifts and across departments allow the entire fab to track the same measures. Engineers who have to spend too much time collecting data and determining whether the data is sufficiently accurate are not spending enough time actually improving the process. Ad hoc reports allow customization, which is a powerful tool for metrics specific to a type of product or piece of equipment. Engineers can also go into the MES database and analyze data to answer specific questions, such as whether a piece of equipment has low yields or the entire shift has a yield problem.
MES benefits flow to ERP for enterprise-wide asset optimization
MES clearly improves the efficiency of shop-floor operations, but these benefits can have an even greater impact on asset optimization when they are extrapolated to enterprise business planning systems. ERP is a demand-driven planning function that determines what must happen to pull product through the line. ERP systems commonly include general ledger and accounting, purchasing, inventory management, financial forecasting, MRP, master production scheduling, sales order management, and work order management.
To be truly effective, ERP systems need real-time information about shop-floor events. MES fills this critical information hole. The ERP system benefits from such MES information as current WIP locations, completion dates, and existing customer commitments.
On the financial side, MES provides accurate standard and activity-based costing. In addition to the value of WIP (a commonly used evaluation), MES can provide the actual cost of manufacturing. The results of this cost accounting function, such as WIP valuation and earnings, yield variance, and procedural variance, are then posted to the general ledger in the ERP system.
With the added data from MES, the ERP system`s accounting functions are far more accurate and comprehensive. ERP systems know what is budgeted for the month or year based on fixed costs such as leasing and maintenance; MES allows the comparison of actual costs with planned expenses to determine, for instance, whether a piece of equipment has adequate return on investment. By providing real-time data regarding actual shop floor events, MES also improves the ERP system`s forecasts, which increases the value of the ERP system and makes the business more profitable overall.
R. Sanborn Towle is director of marketing for Camstar Systems Inc., Campbell, CA. He is responsible for marketing Camstar`s MESA manufacturing execution system. Prior to joining Camstar he held marketing positions at NCA and SSA, both manufacturing software system developers, and with Saxpy Computer. Camstar Systems, Campbell, CA; ph 408/559-5700, fax 408/559-5719, e-mail [email protected].