Abstract: A system to monitor impact forces on piezoelecric smartstructures is constructed. The signals of the impact stress wave in timedomain were transformed intomomentary sequential time signals using Hilbert transform. Neural networkswere established to extract different characteristics from these signalsfor automatic identification of impact location. Experimental resultsshow that the method enjoys good accuracy and stability, which mayfind applications in the health monitoring of laminated structuredamages caused by not very energetic impacts.