Duplex　stainless　steel　was　formed　through　welding　wire　and　arc　additive　manufacturing　(WAAM)　using　tungsten　inert　gas.　The　effects　of　wire　feeding　speed　(WFS),　welding　speed　(WS),　welding　current,　and　their　interaction　on　the　weld　bead　width　and　height　were　discussed.　Back-propagation　(BP)　neural　network　algorithm　prediction　model　was　established　by　taking　the　bead　width　and　height　as　the　output　layer,　and　the　network　weight　and　threshold　values　were　optimized　using　the　particle　swarm　optimization　(PSO)　algorithm　to　obtain　the　prediction　model　of　bead　width　and　height.　The　predicted　results　were　verified　by　experiments.　Results　show　that　the　weld　bead　width　increases　with　the　increase　in　WFS　and　the　welding　current　and　decreases　with　WS.　The　smaller　the　WFS,　the　faster　the　WS,　which　is　beneficial　for　the　generation　of　equiaxed　crystals.　The　smaller　the　welding　current,　the　faster　the　cooling　speed　of　the　metal　melt,　which　is　conducive　to　the　formation　of　dendrites.　The　interaction　among　WS,　wire　feed　speed,　and　welding　current　has　a　significant　effect　on　the　bead　width.　The　weld　bead　height　is　positively　correlated　with　the　wire　feed　speed　and　negatively　correlated　with　the　WS　and　current.　The　interaction　between　the　wire　feed　speed　and　WS　is　significant.　The　optimized　WAAM　process　parameters　for　duplex　stainless　steel　are　a　wire　feed　speed　of　200　cm/min,　WS　of　24　cm/min,　and　welding　current　of　160　A.　The　maximum　error　of　the　BP　neural　network　in　predicting　the　weld　bead　width　and　height　is　7.74%,　and　the　maximum　error　between　the　predicted　and　experimental　values　of　the　BP-PSO　neural　network　is　4.27%.　This　finding　indicates　that　the　convergence　speed　is　fast,　improving　the　prediction　accuracy.