A new intelligence optimization，Big Bang Search (BBS)，is proposed by simulating the big bang process. Inspired by the classical optimization method, the concept of “Approximate Gradient” is defined and the “Approximate Gradient Explosion” (AGE) operator is created，and an improved method called Approximate Gradientbased Big Bang Search (AGBBS) is proposed. AGBBS keeps down the excellent feature of BBS, the nice combination of uniformity and randomness of distributed candidate solutions; it fully uses the information of explosive pieces, which enhances the algorithm’s search ability. By improving some heuristic operators, the convergence of the algorithm and the accuracy of solutions are improved. The testing of 12 standard benchmark functions and a comparative analysis demonstrate the effectiveness of the new algorithm and the robustness of the AGBBS.