The　demand　for　butt　welding　of　stainless-steel　thin　plates　has　been　increasing　year　by　year　in　industries　such　as　shipbuilding,　pharmaceuticals,　petrochemicals,　and　food　processing.　However,　during　the　butt　welding　of　thin　plates,　the　low　structural　rigidity　causes　real-time　change　in　the　gap,　which　complicates　the　ability　of　automated　welding　to　achieve　uniform　weld　seam　geometry.　Therefore,　this　paper　aims　to　overcome　the　challenges　of　cross　coupling　and　the　sensitivity　of　hyperparameter　selection　and　realize　the　gap-adaptive　welding　during　the　high　frequency　pulse　arc　welding　process　by　implementing　to　optimize　the　number　of　trees　and　the　minimum　number　of　leaf　nodes　in　the　random　forest　model.　After　the　optimal　welding　parameters　are　then　used　to　train　the　random　forest　model,　a　gap-adaptive　control　platform,　which　take　gap,　weld　face　width,　weld　root　width,　and　face　reinforcement　as　input　features　and　peak　current,　wire　feed　speed,　and　welding　speed　as　output　features,　is　established　to　enable　real-time　measure　the　gap　by　using　a　laser　track　sensor.　The　results　demonstrate　that　under　fixed　and　gradient　gap　conditions,　the　weld　seam　geometry　is　uniformly　formed　and　no　significant　defects　can　be　found.　Under　step　gap　conditions,　gap-adaptive　welding　effectively　prevents　burn-through　defects　and　ensures　stable　weld　seam　geometry.　Furthermore,　microstructural　and　mechanical　property　characterization　indicates　that　an　increase　in　frequency　during　welding　effectively　refines　the　grain　size　in　the　seam.　The　tensile　strength　values　of　the　samples　are　similar　and　all　samples　fracture　at　the　base　metal.　©　2025