Under the background of condition-based maintenance in the arming maintenance, in allusion to problem of the less equipment data swatch, the fault prognostic method based on least squares support vector machine is studied. The idea of artificial fish swarm algorithm is used to replace the function of the employed bees in the artificial bee colony , therefrom AFS-ABC algorithm is advanced and then used to optimize the parameter of LS-SVM. The LS-SVM is trained by the front forty voltage data sequence swatches of the power supply module in avionics subsystem, and tested by the rear fifteen data sequence swatches. The simulation is done by using the MATLAB LS-SVM toolbox. The result of the simulation shows that the use of this method can prognosticate the arming fault and preferably enhance the capability of LS-SVM.