An　intelligent　elevator　detecting　system　based　on　neural　network　is　proposed　in　this　paper.　It　can　collect　and　store　the　running　data　of　elevators　continuously.　Best　wavelet　packet　basis　is　used　to　select　the　features　of　car＇s　vertical　vibration.　Along　with　the　horizontal　signal　features,　a　BP　neural　network　is　made　to　diagnose　the　jerk　fault.　The　correction　of　the　model　is　proved.　So　the　system　can　provide　referential　data　to　the　inspectors.