Modelling the Impossible at Michelin

For a number of years Michelin has been using WITNESS simulation to continually improve production efficiency and effectively manage the process of engineering change. Benefits have been numerous including regular improvements in productivity as different elements of the production process have been refined. Here we examine how simulation has recently been used to lessen the effects of minor production stoppages and yet again, achieve a step change in performance.

Background

Michelin is one of the world's leading volume manufacturers of car tyres. Its plant in Dundee, UK, is one of the company's largest operations in Europe, producing more than 8.4 million tyres per year This plant is also the home of XM Services, the company's consulting division. One of XM Services' areas of responsibility is productivity improvement and WITNESS is a tool that is regularly used to meet this aim.

Tyre production at Michelin is undertaken using a proprietary machine, the "MAC" (Machine d'Automatique Confection), of which there are four at the Dundee plant. When the plant opened in 1972, the production process was simple; only four sizes of tyres were required and hence, each machine was dedicated to the production of one tyre size.

Today's production environment however, is very different with as many as 50 different tyres sizes available. On average, each machine now produces three or four tyre dimensions per day and the impact on production efficiency has been dramatic. Even though changeover times have already been reduced from four hours to only 15 minutes, size changes still reduce production availability by 15%. In addition, minor stoppages were reducing overall production performance by a staggering 10%!

With changeover operators already working extremely efficiently, XM Services realised that there was little chance of reducing changeover times significantly. The issue of minor stoppages however, offered more hope of being resolved. Minor stoppages at Michelin typically consist of processing trips, products creasing or products missing the line detectors. Although the actual number of stoppages was insignificant and indeed, acceptable within production standards, a stoppage at any one of a MAC's 20-25 working posts would cause the whole line to stop with the disastrous consequences for performance. XM Services realised the number of individual stoppages probably couldn't be reduced any further, but still asked how their effect on performance could be reduced.

According to XM's Simon Goring, this was the question that broke the rules. 

"Michelin operates a machine determinant cycle where each stoppage halts the whole line. We couldn't eliminate the stoppages so we considered masking them but, with a machine determinant cycle, how could stoppages possibly be masked?"

Eventually XM Services came up with the idea of de-coupling each post. There was a possibility that taking out the automatic links between each operation would increase production but still there were many questions; By how much would production be increased by de-coupling? Would de-coupling present a safety risk? And would quality production be affected? Unsure of the impact of the decision XM Services decided to simulate the change. Goring again 

“We had already used WITNESS countless times to quantify the effect of production changes so it made sense to use it again here. Using actual data a model of the current synchronised method was created, validated and then changed to reflect the proposed asynchronised system."

Results

The results confirmed XM Service's beliefs. Tyre carcasses are transported between posts by carriers called "tambours". Full tambours travel between posts along an overhead track and empty units return on a lower track, with lift cages at either end of the machine. Under this continuous method, full tambours must move through each of the posts (21 in this case) in order to finish the process and reach the lift at the end of the machine. As only one tambour can occupy one post at any one time, in order to be fully efficient, the approach depends on operations at all posts finishing simultaneously. Consequently, if a post finishes an operation before the others, the tambour cannot be moved on until all the posts ahead of it have finished. This also has the effect of blocking all posts behind it. Of course a stoppage at any of the posts has exactly the same blocking effect whilst leaving the posts ahead of it idle.

Under the proposed system however, once a post operation is complete, the tambour is shunted on to the next available post. There is no need to wait for all posts to finish and importantly, in the case of minor stoppages, the posts back along the line remain unconstrained. Somewhat surprisingly, the model showed that in order for this asynchronised system to work optimally, fewer tambours would be required and five were removed from the system.

The WITNESS results formed the basis for the justification of £140,000 to modify one of the MACS in order to trial the process. The first MAC has been modified and trials are on-going with a view to modifying the other three MAC’s over Christmas and early 1st quarter next year. Says Goring 

"WITNESS enabled us not only to test out an idea but also quantify its effectiveness. When all the machines have been modified payback will take about a year, but the huge benefit of such a large increase in productivity will no doubt to be felt for years to come. Without WITNESS such a critical improvement may never have been realised."

The simulation results were impressive with the model showing that without increasing the post cycle times, the asynchronised system would increase production by 70 carcasses per hour. This equates to an 8% increase in overall equipment efficiency and more than 550,000 additional tyres per year!