Scheduling tirage champagne production

Penfolds Wines

Penfolds has used time staged linear programming models to analyse their production of tirage champagne, and to develop a technique for smoothing production and reducing stock levels.

Penfolds has used a linear programming model and spreadsheet optimisation software to analyse their production of champagne.

Two principal methods for champagne production are used in Australia:

  • Methode Champagnoise, in which maturation takes place entirely in one bottle, and
  • Transfer method champagne, where two bottles are often used.

For the transfer method, liqueur and actively fermenting yeast are added to a stabilised base wine to form the champagne base wine, known as tirage. The tirage is bottled on the filling and transfer line, then moved to a warehouse where secondary fermentation takes place in the bottle. Fermentation is completed in 4-6 weeks when the product enters its maturation period. The minimum fermentation and maturation period is prescribed under the Food and Drugs Act in Australia to be six months.

After maturation, (six months or more, depending on the product), the tirage bottles are taken to the transfer line. Here the transfer machine pierces the crown seals and the wine is transferred under counter pressure (to preserve the wine's natural fermentation gas), to a pressure tank in the cellar.

The empty tirage bottles are decrowned and rinsed, and then transported back to the tirage filler, where they are immediately refilled with new base wine.

The wine in the pressure tanks is held for about two weeks, after which it is bottled on the expedition (finished product) line in new bottles. The finished product is stored in the warehouse for six weeks. The transference and filling of tirage bottles is an integrated process, performed on the one line.

Seasonal demand patterns, and constraints on bottling capacity made it difficult to manage stock holdings at a level below 12 months.

A basic time staged linear programming model is effective in scheduling production with the objective of reducing holding stock.

Spreadsheet optimization software has played an important role in recent years in bringing these methods within easy reach of decision makers.

In order to meet the peak demand for finished product in December an equivalent volume in tirage must be filled in the previous January, to be transferred in October, allowing for two months settling in the expedition bottle.

Normally this would present no difficulty were it not for the necessity to immediately refill tirage bottles. In October when the transfer from tirage bottles is at its peak, the demand for filling of new tirage stock is low, giving an excess of tirage bottles, whereas in January when the tirage refilling has its peak requirement, the supply of tirage bottles is too low, since little is transferred then.

In a typical year, demand for champagne peaks in December. The champagne must have been transferred in October, and the tirage bottles must have been filled in January.

Production graph

Previously proposed alternative scheduling strategies were found to have shortcomings for various reasons. A linear programming model was adapted to minimise variable production costs.

The linear programming model is effective in scheduling production with the objective of reducing holding stock. A bottle handling model provided a quantitative basis for recognizing that although the setting aside of tirage bottles was a good strategy in principle, it was not cost effective with current handling costs.

MI*Net consultant: Dr David Panton
University of South Australia