Abstract Optimal culture conditions were investigated for β-galactosidase production by recombinant Escherichia coli with a thermally-inducible P L promoter. β-Galactosidase production employing the optimized conditions was then computer-controlled based on culture phase identification using fuzzy logic. β-Galactosidase was not produced by temperature induction at a constant pH of 7.2, which is optimal for cell growth. Rather, the expression of β-galactosidase was strongly promoted by lowering the culture pH at the time the culture temperature was raised to the induction temperature. The expressed β-galactosidase was, however, rapidly degraded as fermentation progressed. This degradation was prevented by lowering the culture temperature and slightly increasing the culture pH above the induction pH. Thus, since different culture conditions were required to grow cells, induce production and prevent degradation, the entire time course of the culture was divided into three phases: a growth phase, a production phase and a degradation phase. The culture was then computer-controlled according to the optimal control policies assigned to each culture phase. Culture phase transitions were identified by calculating the adaptability of on-line culture data to the fuzzy production rules describing culture phase transition. The production rules were constructed using membership functions of state variables and the empirical knowledge of an experienced operator. The experimental results showed that culture phase transitions were successfully identified, leading to the effective production of β-galactosidase.