Aspergillus oryzae (3042)mycelium pellets cultured in liquid were immobilized by cross-linking with gelatin and formaldehyde.Based on orthogonal design L 16 (4 5)

a four-layer artificial neural network (ANN) model with back-propagation of error (BP) was used to optimize the immobilization conditions.The optimal condition was obtained.The conditions were:1g wet pellets were added into 8ml immobilization solution

both the concentration of gelatin and formaldehyde were 5g·L -1

and the immobilization time was 1.5h.The relative activity of immobilized preparation under the optimal condition was assayed.The relative activity was 83%

which was in good agreement with that predicted by ANN.Finally

the stabilization of immobilized preparation was studied.It illustrated that the optimum reaction condition differed from that of free pellet.Compared with free pellet

the optimum pH of reaction system changed from 7.0 to 8.0

temperature changed from 52℃ to 63℃

and the proper reaction condition range was increased.When the immobilized pellet preparation was used for L-alanine continuous production in a φ10mm×100mm packed bed reactor

operational stabilization was markedly increased

and the half-life period was 77d.The immobilized preparation had high stability

and artificial neural network can find the better conditions.