Improving the thermal stability of enzymes is a hot and difficult point in the field of biocatalysis. Compared with the traditional directed evolution, computational assisted rational design is more efficient, and is widely used in enzyme engineering. Using Bacillus subtilis LipA as the model protein, the structure cavity of the enzyme was analyzed by Rosetta-VIP design, the mutation which was beneficial to the filling of the structure cavity (ΔΔE<0) was selected, followed by the solvent accessible surface area and evolutionary conservation analysis. The thermal stabilities of six out of sixteen designed single-point mutants were improved, with a maximum ΔTm value of 3.18 °C. These six mutations were further used for iterative combination mutation, the maximum ΔTm of the two-point and three-point combination mutants were 4.04 °C and 5.13 °C, respectively. The Tm of the four-point combination mutant M11 (F17A/L114P/I135V/M137L) was increased by 7.30 °C. The Tm of the six-point combination mutant M10 (F17A/V74I/L114P/I135V/M137A/I157L) was increased by 7.43 °C. The thermal stability of mutation with lower energy value, reduced accessible surface area, while conformed to evolutionary conservatism, was more likely to be improved. Therefore, the multiple virtual screening strategy based on the enzyme structure cavity filling, solvent accessible surface area and amino acid sequence conservation analysis can effectively improve the thermal stability of enzyme.