Time　delay　is　an　important　studying　aspect　in　real-time　substructure　testing　with　shaking　table.　Time　delay　can　affect　accuracy　and　stability　of　a　testing　system,　and　even　cause　divergence　of　calculation　results.　In　modern　control　theory,　neural　network　is　an　effective　way　to　compensate　time　delay.　Here,　it　was　used　to　predict　a　signal　exerted　on　actuator　of　a　shaking　table.　Through　simulations,　it　was　found　that　its　effect　of　compensation　on　time　delay　is　significant.