A　permanent　magnet　synchronous　linear　servo　motor　has　the　characteristics　of　high　thrust　density,　high　efficiency,　and　fast　control　response.　It　is　widely　used　in　rail　transit,　intelligent　manufacturing,　and　aerospace　fields.　The　electrical　or　mechanical　parameters　of　the　motor　and　nonlinear　resistance　interference　will　affect　the　high　precision　control　of　the　linear　servo.　Therefore,　the　stable　and　efficient　displacement　control　algorithm　has　important　engineering　application　value　to　improve　the　system　performance.　Servo　control　can　be　divided　into　serial　and　parallel　systems　according　to　the　relative　position　of　displacement　and　velocity　loop.　However,　whether　serial　or　parallel　control,　most　controller　parameters　need　to　be　set　in　advance　and　are　generally　fixed,　and　adjusting　the　actual　control　effect　is　impossible.　This　paper　proposes　a　parallel　control　strategy　of　variable　parameter　displacement　velocity　of　permanent　magnet　synchronous　linear　motor　based　on　composite　neural　network　reconstruction　object,　which　realizes　the　high-performance　control　of　linear　servo　motor.　Firstly,　a　parallel　controller　with　variable　parameters　is　designed　using　the　error　information　of　the　moving　displacement　and　linear　velocity.　Then,　the　displacement　output　of　the　permanent　magnet　synchronous　linear　motor　is　reconstructed　by　a　composite　radial　basis　neural　network　containing　multi-dimensional　information　of　the　control　object,　and　the　partial　derivative　of　the　displacement　to　the　control　signal　is　obtained.　Finally,　based　on　the　closed-loop　stability　condition　of　the　system,　a　complete　parameter　update　strategy　of　parallel　controller　with　displacement　velocity　is　provided　according　to　the　comparison　between　periodic　retrieval　errors　and　control　targets.　In　the　experiment,　a　permanent　magnet　synchronous　linear　servo　system　is　designed,　and　different　network　observations,　position　controls,　and　object　parameter　controls　are　compared.　The　network　comparison　shows　that　the　composite　radial-based　neural　network　has　fast　observation　speed　and　high　observation　accuracy,　and　the　observation　time　and　root　mean　square　error　are　18.63　μs　and　0.063　mm,　respectively.　The　position　control　comparison　shows　that　the　variable　parameter　parallel　algorithm　has　higher　control　precision　than　the　fixed　parameter　algorithm,　such　as　PID　control,　with　a　30%～50%　improvement　effect　in　the　displacement　and　velocity　control　error　indexes.　The　parameter　control　comparison　shows　that　in　the　variable　parameter　parallel　control　strategy,　when　the　moving　mass　changes,　the　root　mean　square　of　displacement　error　is　less　than　0.015　mm,　and　the　absolute　maximum　error　is　less　than　0.060　mm,　which　means　that　the　algorithm　has　stability　and　universality.　The　following　conclusions　can　be　drawn:　(1)　The　composite　RBF　neural　network　structure　has　better　observation　performance　than　the　traditional　RBF　neural　network.　Compared　with　the　traditional　improvement　method,　it　has　a　better　effect　with　less　computing　pressure　on　the　processor.　(2)　Compared　with　the　fixed　coefficient　algorithm,　such　as　PID,　the　proposed　variable　parameter　parallel　control　strategy　can　effectively　improve　the　control　performance　of　the　actuator　displacement.　(3)　The　variable　parameter　parallel　control　strategy　can　guarantee　the　high　precision　control　effect　under　different　displacement　settings　and　object　parameters.　©　2024　China　Machine　Press.　All　rights　reserved.