The　key　parts　of　ferromagnetic　materials　that　have　served　in　high　temperature　environment　for　a　long　time　are　prone　to　cracks,　and　the　growth　of　cracks　seriously　threatens　the　safe　operation　of　the　components.　Aiming　at　the　surface　crack　defects　of　ferromagnetic　materials,　based　on　numerical　simulation,　this　study　uses　wideband　ultrasonic　surface　waves　to　carry　out　nondestructive　evaluation　of　surface　crack　defects.　The　wideband　ultrasonic　surface　waves　are　generated　and　received　by　the　wideband　ultrasonic　surface　waves　transducers.　Based　on　time-frequency　analysis　method,　the　interaction　law　between　ultrasonic　surface　wave　with　different　frequency　components　and　surface　crack　defects　is　studied.　Ultrasonic　characteristic　parameters　with　different　frequency　components　are　extracted,　and　a　neural　network　model　with　ultrasonic　characteristic　parameters　as　input　and　surface　crack　defects　information　as　output　is　constructed　to　evaluate　surface　cracks　defects.　The　results　show　when　the　crack　length　is　between　the　wavelength　depth　of　high-frequency　ultrasonic　surface　waves　and　the　wavelength　depth　of　low-frequency　ultrasonic　surface　waves,　the　prediction　error　of　crack　length　is　large.　The　prediction　error　of　crack　length　can　be　reduced　and　the　ability　of　nondestructive　evaluation　of　surface　crack　defects　can　be　improved　when　using　wideband　ultrasonic　characteristic　parameters　for　crack　nondestructive　evaluation.