Plant Combined Simulation for Compost Production starting from Sewage Sludge
From Proceeding of Summer Computer Simulation Conference,SCSC95, Ottawa, July 24-26 1994
Pietro Giribone, Agostino G.Bruzzone, P. Cereda, R.F. Dell'Acqua, Domenico Rivarolo

THE MANAGEMENT MODEL

The decision support created is based on the development of an AI system that makes operating decisions aimed at obtaining certain results in terms of product quality and process cost.
The base configuration was developed in previous studies and uses a core based on a Neural Network trained to find correlations between the objective functions and the operating management parameters.
Therefore, based on the objectives defined, variables are set which act as a reference in each specific choice during the simulation.
The medium-high level decisions include:

Size level required for the cellulose materials
Amount of overtime required
Rental of additional equipment
Division of the areas in the yards
Normal quality controls on the product
Number of times to turn over the material in the maturing areas
Management logic for material recirculation

Instead, the specific decisions of the process are:

Unscheduled maintenance to be performed internally
Changes in the yard configuration
The load of the composting lines
Storage of materials waiting to be processed

For each of these terms, the decision is made based on Fuzzy rules in which one of the decision terms is supplied by the core based on a neural network and corresponds to the reference value for the choice to be made to obtain the required targets.
There are other low-level decisions that are managed in terms of opportunity. These correspond to the location of personnel in the plant and where they are assigned to carry out various operating procedures, as well as the configuration of the material inside the areas assigned and in the sequence used to perform the operations.
These low-level decisions are made on the basis of an order of the possible alternatives in relation to the particular objectives (e.g.: distance of a person from the activity to begin). Having performed the entire modelling process according to a two-dimensional graphic configuration that represents the real layout of the plant (see figure 1), these choices can be made simply and immediately confirm the logic implemented.
The analysis performed indicated that this type of choice does not have a great effect on the medium-term results (4 months), however it may be useful during the operating transposition to choose the rules that make yard management operations more methodical.
* * *