This system, deployed in December 2008, is composed of a set of planning modules developed for Aurora Alimentos, comprising the chains of poultry, swine, industrialized and dairy products, and feed formulation. Its solution architecture is based on statistical forecast models, optimization models implemented in AIMMS, a plan control modules developed in Java®, and on a central database for integration of the input data and the generated plans by its several components.
The poultry chain is the most extensive, including egg-laying cycles, incubation and hatching, one-day-old chick confinement and flock slaughtering. Planning the synchronism of this chain consists on conciliating cycles with discrepant durations: whereas egg laying cycles involve hens with yearly renovation, the incubation-hatching process lasts 3 weeks, and the average confinement-slaughter time is about one and a half month. The non-uniform effect of seasonality over egg production, hatchability and weight-gain rates, and over the flock processing demand itself, makes egg incubation stock management and one-day-old chick supply control a non-trivial process. Besides, there is the requirement that flocks scheduled for processing fulfill a weight distribution compatible with the weight ranges demanded for products – especially those from the export portfolio, which are subject to thin ranges.
In the swine chain, third-party associates control breeding cycles, and therefore, the weighting control of incoming animals is not done on a flock basis. In this case, the production planning for natural and industrialized products is generated based on a weight forecast, and the processing of the large number of parts produced from the slaughter is optimized considering their weight dependent yield and processing plants capacities and efficiencies.
In the dairy chain, the focus is to provide the best production plan for milk-derived products, taking into account the raw milk availability. The big impact of milk quality (especially fatness and dry material availability) on the industrial process, product shelf life and, consequently, the availability variations caused by natural factors, are important aspects for optimizing this chain.
As a considerable amount of total production cost is due to feeding costs, the generation of ideal feed formulations, subject to inventory levels and nutritional requirements, takes an important role on maximizing global operational margin.
In midst 2006, Aurora Alimentos invited UniSoma to provide a solution able to integrate the several planning instances, through a common environment shared amongst planners, whose main objective is to organize long and short-term decisions, as well as to maximize global results.
To do so, it was necessary to articulate decisions market-plants-field direction wise, and to deal with immediate variations on demand and field behavior while not compromising future operation conditions, as well as ensuring targeted service levels in a tactical-strategic view.
The implementation took one-and-a-half year, during which a number of modules from the architecture, several of which were components of a solution that had initially been implemented for Sadia in the 90s, were redesigned, customized and completely re-encoded onto newer platforms.
The mathematical models use statistical models as input which, based on field data, generate necessary forecasts for optimized planning. For the poultry chain, a set of models were adjusted to forecast incubable egg laying and hatchery for hens in parent flocks. The same was done for weight, feed consumption (also available for feed formulation planning) and mortality of chicken flocks, varying according to factors such as breed, sex and seasonality. The intermediate and final product yields derived from chicken and pork slaughter are also determined by forecast models according to the average slaughter weight, which express the results of extensive tests performed at each plant and for each separation processes.
The poultry chain planning activities are directed by the STALA/F module, which synchronizes egg incubation, one-day-old chick confinement and flock slaughter activities, by using incubable eggs and one-day-old chicks purchasing mechanisms to ensure a constant supply of flocks for plant demands considering possible variations that occur throughout time. The INCUBA/F, ALOJA/F and ABATE/F planning module respectively, attend to each of the involved instances – incubator’s, growing confinements and slaughterhouses – to guarantee the best tactical plan, given immediate conditions.
The planning modules PMP and PLASEMA coordinate the demand for chicken and pork flocks. These modules are based on a detailed description of separation and industrial processes made available by the BIBPRO module (Biblioteca de Produtos e Processos – Product and Process Representation), optimize production and transference plans amongst the plants and generate the best storage policy able to guarantee market supply.
The feed formulation module DYFOR meets demands arising from field activity levels with the lowest possible cost, while providing nutritional requirements defined for each feed class, and highlighting purchase opportunities of ingredient based on cost curves and storage capacities. The auxiliary module DYFORWEB receives adjustments on production plans and registers the feed production.
The PLANLEITE module is in charge of optimizing the dairy chain plan, which considers raw milk availability, transportation and supply constraints, production processes of each unit, conditions for product transfers amongst plants, maturation and shelf life periods of each product, as well as the possibilities of distribution and service level.
Solving these models is done coordinately by URSO (UniSoma Remote System for Optimization), whose role is to centralize and organize in a queue all optimization requisitions, allowing control over optimization processes and avoiding the costs that a fat-client architecture would incur. This rationalization of computer resources allows the benefits of a high-performance server especially sized for these tasks for all planners, making their workstations available for other activities while processing takes place on the server.
Module integration is accomplished through GIBI (Gerenciador Integrado de Base de Informação – Integrated Database Manager), the corporate database solution, implemented on Oracle© and integrated to the centralized data source of the company. Through views, GIBI makes available field and plant information necessary for optimization, and supports plan conciliation over several planning horizons. The support module VISUALIZADOR offers managerial follow-up of plan implementation and the results via graphics and reports.
How did AIMMS add value?
The solution contains more than ten AIMMS based Modules. Modules like STALA/F, ALOJA/F and PLANLEITE, are completely designed and developed during the project. Others (like DYFOR, BIBPRO) had been developed previously by UniSoma and were customized according to Aurora´s requests. In this challenging context AIMMS flexibility and robustness were fundamental to allow UniSoma to implement the complete solution in a record time of 18 months. Moreover, its agent-based technology helped the development of URSO that is in use in other companies like JBS Friboi and Gerdau.
About Aurora Alimentos
The Central Oeste Catarinense Cooperative has 17 affiliated cooperatives, 77,000 associates and more than 11,000 employees. Its products range from pork meat, poultry, feed to feed and diary products, with a variety in each group. These products are exported to countries in Latin America, Europe, the Middle East and Asia.