What is the point of SMED?

Publié le 03 décembre 2013

By Xavier Perrin (xperrin@xp-consulting.fr)

When they are talking about their customers’ expectations, all the managers of manufacturing companies say without hesitation products and services quality, short lead times (speed), high customer service level (dependability), and the ability to adjust to the demands (flexibility), in addition to low costs. When they are asked about the evolution of their setup times during the past 5 years, they’re way fewer to give spontaneously precise and detailed answers. In the same way, it is interesting to look at their Key Performance Indicators to see if they include indicators related to setups. In my experience those are often forgotten.

It is generally when a TPM (Total Productive Maintenance) program is launched, with the setting up of the OEE (Overall Equipment Effectiveness) indicator that measures of the loss of capacity due to setup times appear. The goal is then to improve the equipment utilization in order to have more capacity, while lessening the costs.

The SMED (Single Minute Exchange of Die) method to reduce the setup times is then deployed as a tool to reduce the loss of capacity, and to limit the wastes (the setup times do not bring any adding value). The process seems perfectly logical:

  • Measuring OEE of each equipment,
  • Identifying equipment with the lowest OEE,
  • Making a Pareto analysis of capacity losses for each of these equipment,
  • Initiating SMED workshops when setup times appear at the top of the Pareto analysis.

Thus every reduction in setup times translates into a reduction of the loss of capacity, which immediately improves the OEE.

It is interesting to compare such approaches with the customers’ expectations stated before. How would reductions of setup times increase customer service level, reduce lead times, and increase flexibility?

Reducing setup times could be viewed as an opportunity to increase safety capacity which allows absorbing all or part of demand variations, especially when bottleneck resources are affected. Unfortunately, I have rarely seen this argument proven true: all reduction of capacity loss is primarily seen as an opportunity to reduce production costs.

Actually, increasing capacity per se doesn’t contribute directly to the reduction of lead times and to the improvement of service level and flexibility. Except perhaps for some companies with overloaded capacities. In fact, these ones have, first and foremost, a serious planning issue to address!

Reducing lead times and improving service level and flexibility first necessitates appropriate decisions regarding inventories and capacities as part of the S&OP (Sales & Operations Planning) and MPS (Master Production Scheduling) processes. A pull system should be preferred for scheduling operations.

When these basic principles are set down, the overriding factor for flexibility and responsiveness is the lot size.

Determining the lot sizes is probably one of the oldest issues of production and inventory control. Everybody is familiar with the Economic Order Quantity model and the famous Wilson formula, which has been one of the essentials of all good production and inventory control classes for decades. The Wilson EOQ Model (1934) has actually been introduced by M. Harris since 1913! Its goal is to determine an Economic Order Quantity which is a compromise between the ordering or setup costs and the inventory carrying cost caused by the lot size. According to this paradigm, the lot size is the unknown, the variables being the annual usage, the setup cost (function of the setup time), the item cost and the carrying cost. Yet, even if the EOQ Model has been criticized and reconsidered for a while now and the lean approach considers overproduction as the worst of wastes, it still prevails in the reasoning of many managers who are convinced that long runs must be favored. For them, beyond the shadow of a doubt, SMED should be used for increasing equipment productivity.

SMED has been created during the sixties. It has been suggested by Mr Shigeo Shingo, a Japanese consultant working at the time for Toyota in order to systematize the reduction of changeover times. The vision of Toyota’s leaders at that time was to produce “just-in-time”, which means only the right quantity only when it is needed. Thus overproduction was (and still is) considered the worst of the 7 wastes (Muda). Thus, the goal of the production system is the One-Piece-Flow principle, where only one part goes from a work station to another without any stop. Of course, this is an ideal situation which is unattainable in many situations. However, expressing such a vision is useful in the sense that it helps directing and coordinating the efforts in the same direction.

The goal of SMED is therefore to reduce lot sizes. The goal is not to reduce the portion of capacity which is dedicated to setups. Deciding which part of the capacity has to be “invested” for setups is a strategic decision. To put it differently, managers have to provide the means for flexibility and short lead times. Many companies in the automotive industry choose to spend about 10% of the capacities for setups.

Thus, the SMED method, while helping to reduce setup times, allows first and foremost to reduce lot sizes, and eventually to reduce “intervals”.

The “interval” or EPEI (Every Part Every Interval) principle is illustrated in the following diagram. In this model, we consider a weekly loading time of 2,400 minutes (5 days x 8 hours x 60 minutes). 6 references (A, B, C, D, E and F) are assigned to the production process. The initial setup time is 40 minutes. If it is decided to dedicate 10% of the loading time to setups, thus 240 minutes are available. The first chart shows that one week is needed to produce all 6 references. The corresponding lots stand for one week of customer demand for each reference. If applying the SMED method allows dividing the time for one setup by 2, which becomes 20 minutes, then 12 changeovers can be made in a week. The interval is then cut down to 2.5 days. It can be cut down to 1.25 days if the duration of one setup is reduced to 10 minutes.

As shown on the diagram, the lessening of the setup time doesn’t lead to an improvement of the OEE: this is not the goal of SMED.

Thereby, what is to gain in reducing the interval? The interval determines the level of work-in-process inventories, so it determines throughput times, so it determines lead times. It also determines the frozen horizon of production schedules. Every time it is lessened, WIP and lead time’s decrease and production processes become more responsive thanks to shorten frozen horizon. Short intervals also help to improve the usefulness of inventories: a reduced lot size means a reduced cycle stock (the part of the stock cycling according to the lot size). It allows dedicating more stock as safety stock to compensate demand variations. Stocks become then more “useful” since they allow a better customer service, and thus a better customer satisfaction.

2 réactions sur What is the point of SMED?

  • Mike Clayton dit :

    HOWEVER…we did focus on SETUP TIMES when designing automated process tools with automated handling systems, and eventually operator jobs were eliminated for many steps. After the batch operations for thousands of IC »s per wafer, the wafer IC’s are « singulated » by sawing or laser cutting and then individual parts are processed in packaging operations and tested individually, so that « back end of line » is now the highest cost set of steps, and SMED is very important as is predictive maintenance and of course yield detractors are traced back to root causes with support of « id » tags and manufacturing execution systems with relational databases tracking every one of the 1000’s of steps across many countries in some cases. Lean and Six Sigma methods are merged in appropriate situations.

  • Mike Clayton dit :

    In making IC’s we found that overall cost was reduced by making lot sizes BIGGER for some processes. In early transistor days, each chip was processed separately but when the « step and repeat » photo-lithography masks and wafers with many IC’s or transistors on each wafer we found « Moore’s Law » drove us to larger and larger wafers, and smaller and smaller devices to cut cost 2x every 18 months roughly from 1960 to recent times. HOWEVER, for some process steps we load one wafer and if process times are in hours we load many wafers to minimize capital equipment costs. Sooooo..it depends. The military funded early work and called it « Batch Electronics » before the name « Integrated Circuit » was coined.

  • Derniers articles

    Nuage de tags

    Amélioration Continue APS Continous Improvement Coûts DBR ERP flux Flux tiré Genshi Genbutsu GPAO Heijunka Hoshin Informatique Intervalle IT Jidoka Judoka Just-in-Time Juste-à-Temps Kaizen kanban Lean Lecture Management MES MPS MRP2 Muda PDP PIC Productivité Progrès S&OP Setup SMED Stocks Stratégie Supply Chain Taille de lot Throughput TOC Toyota TPM TPS VSM


    XP Consulting – Le BLOG est fièrement propulsé par WordPress et le thème SubtleFlux traduit par WordPress tuto.

    Copyright © XP Consulting – Le BLOG.