Multifactor optimization of MICP base on BP model

Abstract

Microbial-induced calcite precipitation (MICP) can be used to cement soil and produce new biomaterials. The formation of this material is affected by many factors, such as physical, chemical, biological factors, many studies focused on the effect of a single factor on the strength ignoring the synergy between factors. Back-propagation neural network (BPNN) can be used as a multi-factor nonlinear prediction model and an analysis method. 140 MICP grout tests in the literature were summarized to15 factors affecting UCS and act as BPNN input data. At last, five key factors were elected based on weight analysis of well-trained BPNN. On this basis, a simple strength model composed of those factors is established, which can well predict the strength of MICP grouting soil with practical convenience. Key factors and strength prediction models help popularize MICP for engineering applications and optimize grouting experiments.