Optimization of Spacer Rheology Using Neural Network Technology

Abstract

Abstract This paper presents a new method to design the optimum composition of a spacer fluid used for mud removal in a cement job. The method integrates two steps: determining the ideal spacer properties from the knowledge of well conditions, mud and cement properties; and designing the fluid composition to achieve these properties. The first step is based on well-accepted requirements for spacers, including friction pressures and density-hierarchy criteria. The second step is based on a neural network model developed from an extensive database of fluid properties, where all key variables were moved: composition, spacer density, and temperature. A software module simulates and displays the mud-removal performance of the proposed design. Two field examples demonstrate successful application of the method. They compare the uses of the new method with the conventional one, demonstrating the timesaving of the new method. They also show the efficiency of the new method in designing optimal spacers for the full range of pumping conditions. Several years of field experience demonstrate this integrated method has advantages in two areas. First, displaying mud-removal performance graphically enables better and faster job design. Second, the method achieves fluids with the required properties in significantly less time than conventional methods because fewer fluid samples are mixed.