Qualitative reasoning and integrated management of the quality of stored grain: a promising new approach

Grain quality maintenance has traditionally been the responsibility of grain storekeepers who rely on measurements of grain or its milled products and on implicit knowledge gained through scientific results, common sense and job experience. Thus, managers of grain stores must comprehend the ecological, economic and technical consequences of their actions. Practical application of available knowledge by a grain-store manager may be greatly enhanced by the combination of fundamental concepts from stored-grain ecosystems with principles of expert systems (ES). The ES are computer programs with the capacity to mimic the reasoning logic of human experts when solving complex problems. They have been available as aids for stored-grain management for about a decade. The existing systems that focus mainly on pest management issues are briefly reviewed in the introductory part. Today, there is increasing pressure on the grain-handling industry to manage not only pests but also a broader range of parameters involved in grain quality maintenance. The implementation of a new type of ES to manage grain quality parameters during storage through an integrated qualitative approach is presented. In practice, most of the existing basic knowledge on progressive changes in grain quality attributes can be approximated by the statement of an initial state of quality and predictive models using only three variables: storage duration, grain temperature and water activity (or related moisture content). Combinations of qualitative variables through a logical decision network, predicting changes in different aspects of quality, may enable the prediction of quality from the initial diagnosis. Such a rule-based reasoning approach called "qualitative reasoning" has recently been developed as a decision support tool, starting from the diagnosis of quality attributes of a grain batch upon delivery at a grain store and anticipating the changes that will occur during the storage period. The new approach, based on high level reasoning methods and using a knowledge base (KB) of interactive rules about grain quality changes, is briefly presented. The implementation of this KB has focused on attributes that change very early in unsafe storage conditions such as germination capacity, micro-organism respiration, dry matter loss, visible moulding appearance, pest dynamics, cooling aeration effects and efficacy of residual protectants. Available models that may be introduced into the KB of the system have been reviewed. Most of the parameters of predictive models can be obtained by computation of existing experimental data. The prototype testing indicates that the system may give reliable quality diagnosis and prediction of optimal storage with a level of expertise and advice comparable to human experts. The additional work needed to extend the scope and domain of applicability of this new generation of ES is discussed. The procedure of qualitative reasoning applied to the whole grain ecosystem is shown to be a promising tool for detecting the weaknesses of basic knowledge about changes in cereal quality during storage.