Dyfor – Dynamic Feed Formulation
In mid 2001 the Dyfor system was deployed at Sadia for optimized feed production planning at all of its plants.
The poultry and swine production at Sadia is based on the integration system. According to this scheme, the company provides its “integrated” partners with all resources necessary for animal production (such as feed, technical assistance and financial credit) until the moment of processing the animals, paying them according to their raised flocks or herds performance. This model avoids bank credit lines and/or subsides dependency throughout the production chain and is largely adopted by other poultry and swine companies in Brazil.
In 2007, Sadia had 6 thousand “integrated” partners for broiler production and 4 thousand for swine production. For poultry, 100% of the production was guaranteed by the integration system; for swine, around 40% was guaranteed, the rest being supplied by independent producers or partial Sadia partners.The table below summarizes the production levels and capacities installed at the company at that time.
|Poultry||millions of units||751.9||776.0|
|Swine||millions of units||4.4||5.0|
|Feed||thousands of tons||5,242.1||5,515.4|
The feed formulations are produced according to the nutritional requirements for the several types of animals, at their different growing stages. For instance, poultry initial and growing feeds make use of fats with high levels of unsaturated fatty acids, like those found on soybean oil. On the other hand, final feeds tend to increase unsaturated fats (e.g. suet) levels, ensuring the quality fat in the carcass.
Besides this, the levels of potassium must be controlled, in order to prevent diarrhea during the growing phase. Raw materials of vegetable origin may be contaminated with fungi, which, beyond a certain level, may favour the appearance of micro-toxins, with negative impact on feed conversion; to avoid this, organic acids are included in feed formulations. Still, the use of whole soybean bran or soybean oils improves the digestibility of “hard” fats.
In 1985, with the opportunity resulting from the quick dissemination of microcomputers, UniSoma developed the Multifor system (Microsoft DOS© platform) for “simultaneous” multiple feed formulation optimization subject to restricted availability (in the form of inventories) of several ingredients. Besides “globally” optimizing the use of scarce raw materials, the tool supported (1) sensibility analysis of the obtained solution and (2) raw material cost variation analysis (parametric). More than thirty copies of Multifor were licensed for several feed and pre-mix production companies, such as: Sadia, Perdigão, Chapecó, Agroceres e M. Cassab. In the early 90s, with the development of relational databases, a new version of the application was developed. The database for this new system was built on ZIM.
In last decade, UniSoma consolidated the adoption of AIMMS software as its standard development platform for Advanced Planning Systems (APS). In 2000, UniSoma proposed to Sadia a comprehensive refactoring of Multifor in AIMMS language.
In terms of the feed formulation model, the main innovations established for this migration were:
- Multi-Plant Programming: extension of the mathematical model allowing for multi-formula optimization among several feed production plants, making use of raw material transfer possibilities;
- Purchase Planning: alternative supply representation for each plant allowing for multiple suppliers in different locations, differentiated prices and availabilities;
- Strategic Stock Planning (multi-period): raw material stock control at plants (and third parties’ warehouses) throughout the planning periods, seeking gains due to seasonal raw material price variations at the several suppliers and according to a finite inventory capacity.
The new system was named Dyfor – Dynamic Formulation, given the complexity and dimension of the problems solved by it.
The solution obtained through Dyfor is the one which minimizes the present value of the total cost of feed formulation. With this, a compromised solution is reached among purchase, transportation (transference) and stock costs. In addition to this, the following operational and feed formulation constraints are observed:
- Availability of raw materials purchase at the several locations;
- Raw material consumption limits per plant;
- Ingredient stock capacities per plant;
- Ingredient transfer capacities per plant;
- Nutritional (ranged) requirements;
- Nutritional values proportions;
- Ingredient or ingredient group participation limits – with the possibility to set a minimal composition level for an ingredient, should the system decide to use it in the formulation.
Dyfor also allows for prediction equations calculation, which are formulas used to calculate nutritional values derived from other previously known nutritional values. With this feature it is possible to define, for example, the following formula for the corn calculation:
TotalLysine[%] = 0,064[%] + 0,024*RawProtein[%]
Dyfor has been used by Sadia since 2000 to plan feed formulation for all of its plants. In addition, it also supports the annual budget definition.
The gains effectively experienced by Sadia with the use of Dyfor haven’t been monetarily measured, since Sadia previously already used Multifor – even though this tool was not as complete and versatile as the new one. Moreover, not all of the Dyfor features have been effectively explored until today, as is the case of multi-period programming, whose studies performed by UniSoma indicated economy potentials of as high as 4%.
Even so, the use of Dyfor for practically a decade, with UniSoma’s support, clearly states the achievement of the goals which the solution aimed for.
Agribusiness and food producer, Sadia presented a liquid revue of R$ 9,8 billion in 2007. It is one of the world’s largest natural meat (chicken, pork and turkey) and industrialized products (margarine, pasta, frozen meals, deserts and animal protein derived products) producer and is one of Brazil’s leaders in poultry exportation. Its operational infrastructure consists of 14 industrial units, 2 agriculture and animal production centers, 16 distribution centers and several sales affiliates, present in 14 Brazilian states and Europe. Sadia was elected 4 times (2001, 2003, 2004 and 2005) the most valuable brand in the Brazilian food sector (Interbrand).
In the 40s, George B. Dantzig introduced the Simplex algorithm, for the solution of problems represented in the form of linear programming mathematical models. The minimal cost diet problem was the first example used to apply the method. Nowadays this single-formulation problem is easily solved through the use of Excel worksheets. However, such solver is not robust enough to solve problems of increased complexity, such as the ones solved by Dyfor.