キャンセル
次の結果を表示 
表示  限定  | 次の代わりに検索 
もしかして: 

JMP Blog

A blog for anyone curious about data visualization, design of experiments, statistics, predictive modeling, and more
%3CLINGO-SUB%20id%3D%22lingo-sub-324081%22%20slang%3D%22zh-TW%22%20mode%3D%22NONE%22%20mode%3D%22NONE%22%20mode%3D%22NONE%22%3E%5BEstudio%20de%20caso%5D%20%7C%20Aplicaci%C3%B3n%20de%20un%20algoritmo%20inteligente%20para%20predecir%20el%20ancho%20natural%20del%20laminado%20en%20caliente%20y%20acabado%20del%20Baosteel%201580%3C%2FLINGO-SUB%3E%3CLINGO-BODY%20id%3D%22lingo-body-324081%22%20slang%3D%22zh-TW%22%20mode%3D%22NONE%22%20mode%3D%22NONE%22%20mode%3D%22NONE%22%3E%3CP%3ELa%20fabricaci%C3%B3n%20inteligente%20es%20la%20tendencia%20inevitable%20del%20desarrollo%20tecnol%C3%B3gico%20de%20las%20sociedades%20an%C3%B3nimas%20en%20la%20nueva%20situaci%C3%B3n%20y%20tambi%C3%A9n%20es%20una%20garant%C3%ADa%20importante%20para%20ganar%20la%20iniciativa%20y%20permanecer%20invencible%20en%20la%20feroz%20competencia%20del%20mercado.%20La%20construcci%C3%B3n%20del%20taller%20inteligente%20de%20laminaci%C3%B3n%20en%20caliente%201580%20es%20el%20primer%20proyecto%20inteligente%20de%20demostraci%C3%B3n%20de%20Baosteel%20dirigido%20por%20el%20Ministerio%20de%20Industria%20y%20Tecnolog%C3%ADa%20de%20la%20Informaci%C3%B3n.%20Mediante%20la%20introducci%C3%B3n%20y%20actualizaci%C3%B3n%20de%20la%20tecnolog%C3%ADa%20y%20los%20equipos%20existentes%2C%20Baosteel%20ha%20tomado%20medidas%20para%20mejorar%20la%20industria%20del%20acero%20inteligente.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22pexels-photo-373543.jpeg%22%20style%3D%22width%3A%20940px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22pexels-photo-373543.jpeg%22%20style%3D%22width%3A%20940px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22pexels-photo-373543.jpeg%22%20style%3D%22width%3A%20940px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22pexels-photo-373543.jpeg%22%20style%3D%22width%3A%20940px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22pexels-photo-373543.jpeg%22%20style%3D%22width%3A%20940px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27520iA7EEBF27C25F8534%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22pexels-photo-373543.jpeg%22%20alt%3D%22pexels-photo-373543.jpeg%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EEste%20art%C3%ADculo%20presentar%C3%A1%20en%20detalle%20la%20pr%C3%A1ctica%20del%20proyecto%20de%20c%C3%B3mo%20Baosteel%20aplica%20el%20algoritmo%20inteligente%20basado%20en%20JMP%20Pro%20para%20predecir%20el%20ancho%20natural%20del%20laminado%20en%20caliente%20y%20del%20laminado%20de%20acabado%201580.%20Tambi%C3%A9n%20es%20un%20complemento%20al%20discurso%20pronunciado%20por%20el%20Sr.%20Zhu%20Jianqin%2C%20ingeniero%20jefe.%20del%20Departamento%20de%20Equipos%20de%20Modelado%20Digital%20de%20Baosteel%20Co.%2C%20Ltd.%20en%20la%20JMP%20Discovery%20Summit%20China%202020.%20.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CFONT%20size%3D%225%22%3E%3CSTRONG%3Euno%3C%2FSTRONG%3E%3CSTRONG%3E%3C%2FSTRONG%3E%3CSTRONG%3E%20Antecedentes%20del%20proyecto%3C%2FSTRONG%3E%3CSTRONG%3E%20%E2%80%94%3C%2FSTRONG%3E%3CSTRONG%3E%20por%20qu%C3%A9%20es%20necesario%3C%2FSTRONG%3E%3CSTRONG%3E%20JMPPro%3C%2FSTRONG%3E%3CSTRONG%3E%20%3F%3C%2FSTRONG%3E%3C%2FFONT%3E%3C%2FP%3E%0A%3CP%3EEn%20el%20proyecto%20de%20construcci%C3%B3n%20del%20taller%20inteligente%20de%20laminaci%C3%B3n%20en%20caliente%20Baosteel%201580%2C%20Primetals%20Technologies%20Germany%20GmbH%20(PT)%20desarroll%C3%B3%20el%20sistema%20de%20control%20din%C3%A1mico%20de%20ancho%20de%20laminaci%C3%B3n%20de%20acabado%20(DWC)%20utilizando%20su%20predicci%C3%B3n%20del%20ancho%20natural%20de%20laminaci%C3%B3n%20de%20acabado%20y%20su%20modelo%20de%20red%20neuronal%20autoajustable.%20En%20el%20sistema%2C%20a%20trav%C3%A9s%20de%20la%20entrada%20de%20los%20par%C3%A1metros%20del%20proceso%20L2%2C%20se%20calcula%20el%20valor%20previsto%20del%20ancho%20natural%20del%20laminado%20de%20acabado%2C%20que%20se%20utiliza%20para%20determinar%20el%20ancho%20objetivo%20del%20laminado%20en%20bruto%20y%20el%20autoajuste%20posterior%20al%20laminado%2C%20y%20se%20transmite%20a%20L1%20como%20referencia.%20valor%2C%20y%20L1%20realiza%20el%20ancho%20de%20laminado%20de%20acabado.%20Control%20din%C3%A1mico%2C%20mejorando%20as%C3%AD%20significativamente%20la%20precisi%C3%B3n%20del%20control%20del%20ancho%20de%20salida%20del%20laminado%20de%20acabado%20de%20laminado%20en%20caliente%201580.%20La%20desviaci%C3%B3n%20promedio%20entre%20el%20valor%20de%20predicci%C3%B3n%20del%20modelo%20y%20el%20valor%20real%20es%20de%20solo%201%2C38%20mm%20(de%20enero%20a%20septiembre).%202019)%2C%20ver%20Figura%201.%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE%201..png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE%201..png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE%201..png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE%201..png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE%201..png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27513i1C579340E8890208%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE%201..png%22%20alt%3D%22%E5%9B%BE%201..png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3EFigura%201%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3ESin%20embargo%2C%20el%20modelo%20de%20red%20neuronal%20autoajustable%20y%20de%20predicci%C3%B3n%20de%20ancho%20natural%20de%20precisi%C3%B3n%20proporcionado%20por%20Pratt%20en%20Alemania%20es%20un%20modelo%20de%20caja%20negra.%20Su%20c%C3%B3digo%20de%20programa%20original%20est%C3%A1%20encriptado%20y%20no%20se%20puede%20leer%20ni%20modificar.%20No%20hay%20descripci%C3%B3n%20de%20la%20estructura%20del%20modelo%2C%20por%20lo%20que%20no%20se%20puede%20realizar%20Ajuste%20de%20par%C3%A1metros%2C%20ajuste%20del%20modelo%20y%20otros%20trabajos%20de%20mejora.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EPor%20lo%20tanto%2C%20el%20equipo%20espera%20desarrollar%20un%20modelo%20inteligente%20aplicando%20otras%20herramientas%20de%20aprendizaje%20autom%C3%A1tico%2C%20aprendizaje%20profundo%20y%20otras%20herramientas%20de%20modelado%20de%20inteligencia%20artificial%20de%20la%20familia%20de%20productos%20de%20an%C3%A1lisis%20estad%C3%ADstico%20JMP%20para%20predecir%20el%20ancho%20natural%20del%20laminado%20final%20en%20el%20laminador%20en%20caliente%20Baosteel%20Co.%2C%20Ltd.%201580.%20Reemplaza%20el%20modelo%20de%20caja%20negra%20de%20la%20empresa%20alemana%20PT%20y%20hace%20que%20la%20desviaci%C3%B3n%20entre%20el%20valor%20previsto%20y%20el%20valor%20real%20del%20modelo%20de%20desarrollo%20propio%20alcance%20el%20valor%20promedio%20mensual%20de%20%E2%89%A41%2C7%20mm%2C%20alcanzando%20as%C3%AD%20b%C3%A1sicamente%20el%20nivel%20de%20control%20del%20modelo%20de%20caja%20negra%20alemana.%20modelo%20de%20caja.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EEn%20base%20a%20esto%2C%20el%20equipo%20comenz%C3%B3%20a%20realizar%20m%C3%A1s%20intentos%20y%20exploraciones%20con%20la%20ayuda%20de%20herramientas%20de%20modelado%20predictivo%20en%20JMP%20Pro.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CFONT%20size%3D%225%22%3E%3CSTRONG%3Edos%3C%2FSTRONG%3E%3CSTRONG%3E%20Establecimiento%20de%20modelos%20de%20predicci%C3%B3n%20inteligentes.%3C%2FSTRONG%3E%3C%2FFONT%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E1%3C%2FSTRONG%3E%3CSTRONG%3E%20Recopilaci%C3%B3n%20y%20limpieza%20de%20datos.%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3EEl%20equipo%20recopil%C3%B3%201.580%20datos%20de%20producci%C3%B3n%20de%20laminado%20de%20acabado%20laminado%20en%20caliente%20durante%20un%20total%20de%20nueve%20meses%20desde%202019.01%20hasta%202019.09%2C%20y%20compil%C3%B3%20un%20programa%20SAS%20para%20limpiar%20e%20integrar%20los%20datos%20de%20enero%20a%20agosto%20de%202019.%20Se%20extrajeron%20y%20resumido%20Forme%20un%20archivo%20de%20datos%20de%20Excel.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E2%3C%2FSTRONG%3E%3CSTRONG%3E%20Filtrado%20de%20datos%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E01%3C%2FSTRONG%3E%3CSTRONG%3E%20Utilice%20algoritmos%20inteligentes%20para%20detectar%20factores%20clave%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3ESe%20utiliz%C3%B3%20el%20algoritmo%20de%20bosque%20Bootstrap%20en%20el%20software%20de%20an%C3%A1lisis%20estad%C3%ADstico%20JMP%20Pro%20para%20realizar%20un%20entrenamiento%20de%20ajuste%20en%20este%20conjunto%20de%20datos%20(enero%20a%20agosto%20de%202019%2C%2085423%20registros%2C%20206%20factores%20influyentes)%2C%20y%20luego%20se%20bas%C3%B3%20en%20los%20resultados%20del%20entrenamiento%20en%20funci%C3%B3n%20del%20ancho%20natural%20real.%20contribuci%C3%B3n%20(basado%20en%20la%20suma%20de%20cuadrados%20de%20la%20regresi%C3%B3n)%2C%20se%20seleccionaron%20factores%20de%20influencia%20con%20una%20suma%20de%20cuadrados%20de%20varianza%20%26gt%3B%203000%2C%20un%20total%20de%2031%20factores%2C%20ver%20Figura%202.%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE2.png%22%20style%3D%22width%3A%20780px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE2.png%22%20style%3D%22width%3A%20780px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE2.png%22%20style%3D%22width%3A%20780px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE2.png%22%20style%3D%22width%3A%20780px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE2.png%22%20style%3D%22width%3A%20780px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27514iCC2AD329343C05BC%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE2.png%22%20alt%3D%22%E5%9B%BE2.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3CBR%20%2F%3EFigura%202%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E02%3C%2FSTRONG%3E%3CSTRONG%3E%20Pratt%20(%3C%2FSTRONG%3E%3CSTRONG%3E%20PT%3C%2FSTRONG%3E%3CSTRONG%3E%20)%20Herencia%20de%20experiencia%20de%20la%20empresa%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3ELa%20interfaz%20de%20entrada%20%2F%20salida%20del%20sistema%20de%20an%C3%A1lisis%20y%20diagn%C3%B3stico%20del%20modelo%20de%20red%20neuronal%20de%20ampliaci%C3%B3n%20natural%20proporcionado%20por%20Pratt%20(PT)%20de%20Alemania.%20No%20es%20dif%C3%ADcil%20ver%20que%20el%20modelo%20de%20caja%20negra%20de%20la%20red%20neuronal%20requiere%2033%20par%C3%A1metros%2C%20como%20la%20composici%C3%B3n%20qu%C3%ADmica%20y%20el%20ancho%20de%20entrada%20del%20acabado.%20y%20ancho%20de%20salida.Variables%20de%20entrada%2C%20consulte%20la%20Figura%203.%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE3.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE3.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE3.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE3.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE3.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27515iECC98F64208B339F%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE3.png%22%20alt%3D%22%E5%9B%BE3.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3Eimagen%203%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E03%3C%2FSTRONG%3E%3CSTRONG%3E%20S%C3%ADntesis%20de%20datos%3C%2FSTRONG%3E%3CSTRONG%3E%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EEl%20equipo%20compar%C3%B3%20los%20factores%20clave%20seleccionados%20por%20el%20algoritmo%20forestal%20Bootstrap%20de%20JMP%20Pro%20con%20las%20variables%20de%20entrada%20requeridas%20por%20el%20modelo%20de%20red%20neuronal%20de%20Pratt%20(PT)%20y%20descubri%C3%B3%20que%204%20factores%20de%20influencia%20eran%20iguales%20y%20estaban%20precalculados%20y%20eran%20precisos.%20ancho%20del%20objetivo%20y%20composici%C3%B3n%20qu%C3%ADmica%20silicio%20(Si)%2C%20niobio%20(Nb)%2C%20etc.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAl%20mismo%20tiempo%2C%20seg%C3%BAn%20la%20experiencia%20en%20el%20an%C3%A1lisis%20de%20la%20calidad%20de%20la%20producci%C3%B3n%2C%20no%20es%20dif%C3%ADcil%20ver%20que%20existe%20una%20autocorrelaci%C3%B3n%20entre%20factores%20como%20el%20ancho%20real%20del%20laminado%20en%20desbaste%2C%20la%20segunda%20mitad%20real%20del%20ancho%20del%20laminado%20en%20desbaste%20y%20el%20ancho%20objetivo%20del%20laminado%20en%20desbaste%2C%20y%20Estas%20variables%20son%20consistentes%20con%20el%20ancho%20natural%20real%20objetivo%20que%20el%20equipo%20necesita%20ajustar.%20Spread%20(spread_ACT)%20tambi%C3%A9n%20tiene%20una%20correlaci%C3%B3n%20lineal.%20Al%20aplicar%20la%20herramienta%20de%20an%C3%A1lisis%20de%20correlaci%C3%B3n%20lineal%20multivariante%20de%20JMP%20Pro%20para%20analizar%20estos%20ocho%20factores%20de%20influencia%20y%20los%20valores%20reales%20del%20ancho%20natural%2C%20se%20puede%20ver%20que%20existe%20una%20fuerte%20correlaci%C3%B3n%20entre%20los%20ocho%20factores%20de%20influencia%20(coeficiente%20de%20correlaci%C3%B3n%20r%26gt%3B0%2C99)%2C%20y%20son%20estrechamente%20relacionado%20con%20el%20ancho%20natural.%20La%20correlaci%C3%B3n%20del%20valor%20real%20del%20diferencial%20tambi%C3%A9n%20es%20relativamente%20fuerte%20(r%26gt%3B55%25).%20Ver%20Figura%204.%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE4.png%22%20style%3D%22width%3A%20784px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE4.png%22%20style%3D%22width%3A%20784px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE4.png%22%20style%3D%22width%3A%20784px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE4.png%22%20style%3D%22width%3A%20784px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE4.png%22%20style%3D%22width%3A%20784px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27516i103369C00425E1FF%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE4.png%22%20alt%3D%22%E5%9B%BE4.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3EFigura%204%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAdem%C3%A1s%2C%20hay%20m%C3%A1s%20de%201.000%20marcas%20de%20roscado%20(c%C3%B3digo%20de%20aleaci%C3%B3n).%20Aunque%20tienen%20un%20impacto%20en%20el%20ancho%20natural%20y%20son%20las%20que%20m%C3%A1s%20contribuyen%2C%20no%20se%20pueden%20ajustar%20porque%20las%20dimensiones%20son%20demasiado%20grandes.%20Se%20puede%20utilizar%20un%20m%C3%A9todo%20de%20tabla%20para%20el%20aprendizaje%20de%20autoajuste.%20.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EPor%20lo%20tanto%2C%20despu%C3%A9s%20de%20la%20discusi%C3%B3n%2C%20finalmente%20se%20decidi%C3%B3%20eliminar%20estas%209%20variables%2C%20m%C3%A1s%20las%204%20variables%20repetidas%2C%20se%20eliminaron%20un%20total%20de%2013%20factores%20influyentes.%20De%20esta%20manera%2C%20quedan%2018%20de%20los%2031%20factores%20influyentes.%20Sumando%20los%2033%20factores%20heredados%20de%20la%20experiencia%20de%20Pratt%2C%20finalmente%20se%20determinan%2051%20factores%20influyentes.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E3%3C%2FSTRONG%3E%3CSTRONG%3E%20preparaci%C3%B3n%20de%20datos%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3EPara%20los%2051%20factores%20de%20influencia%20finales%20que%20afectan%20el%20ancho%20natural%20del%20laminado%20de%20acabado%2C%20el%20equipo%20recopil%C3%B3%20datos%20durante%20un%20total%20de%20tres%20meses%20de%20junio%20a%20agosto%20de%202019%20(un%20total%20de%2030%2C533%20tiras)%2C%20utiliz%C3%B3%20JMP%20Pro%20para%20leer%20los%20datos%20y%20el%20valor%20real.%20del%20ancho%20natural%20del%20laminado%20de%20acabado%20se%20utiliza%20como%20valor%20objetivo%20(Y%2C%20variable%20de%20respuesta)%2C%20se%20utilizan%2051%20factores%20de%20influencia%20como%20datos%20de%20ajuste%20(X%2C%20cita)%20y%20el%2080%25%20del%20conjunto%20de%20datos%20de%20ajuste%20se%20establece%20aleatoriamente%20como%20conjunto%20de%20entrenamiento.%20y%20el%2020%25%20se%20utiliza%20como%20conjunto%20de%20verificaci%C3%B3n.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CFONT%20size%3D%225%22%3E%3CSTRONG%3Etres%3C%2FSTRONG%3E%3CSTRONG%3E%3C%2FSTRONG%3E%3CSTRONG%3E%20Entrenamiento%20de%20modelos%20inteligentes%20y%20adaptaci%C3%B3n%20fuera%20de%20l%C3%ADnea%3C%2FSTRONG%3E%3C%2FFONT%3E%3C%2FP%3E%0A%3CP%3EEl%20equipo%20intent%C3%B3%20aplicar%20dos%20m%C3%A9todos%20de%20modelado%20inteligente%20en%20el%20software%20de%20an%C3%A1lisis%20JMP%20Pro%3A%20red%20neuronal%20y%20m%C3%A9todo%20de%20bosque%20Bootstrap%20para%20entrenar%20y%20verificar%20el%20conjunto%20de%20datos%20anterior%20respectivamente%2C%20y%20obtuvo%20los%20siguientes%20resultados%3A%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E1%3C%2FSTRONG%3E%3CSTRONG%3E%20Aplicar%20el%20m%C3%A9todo%20de%20modelado%20de%20redes%20neuronales.%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27517i1EC03FCCA0F5F286%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20alt%3D%22%E5%9B%BE5%20%E5%9B%BE6.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3EFigura%205%20Figura%206%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EComo%20puede%20verse%20en%20las%20Figuras%205%20y%206%2C%20el%20modelo%20obtenido%20despu%C3%A9s%20del%20entrenamiento%20y%20ajuste%20utilizando%20el%20algoritmo%20de%20red%20neuronal%20tiene%20un%20coeficiente%20de%20determinaci%C3%B3n%20R2%20de%200%2C858%2C%20un%20RMSE%20de%201%2C506%20y%20un%20promedio%20de%20desviaciones%20absolutas%20de%201%2C16%20despu%C3%A9s%20de%20la%20verificaci%C3%B3n%20utilizando%20datos%20de%20verificaci%C3%B3n.%20En%20teor%C3%ADa%2C%20se%20han%20logrado%20los%20objetivos%20marcados%20por%20el%20equipo.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3ELa%20Figura%206%20muestra%20el%20diagrama%20de%20dispersi%C3%B3n%20de%20los%20valores%20predichos%20de%20los%20tres%20modelos%20(de%20izquierda%20a%20derecha%2C%20el%20modelo%20JMP%2C%20el%20modelo%20PT%20y%20el%20modelo%20existente%20de%20Baosteel)%20y%20los%20valores%20reales.%20Se%20puede%20ver%20que%20el%20modelo%20ajustado%20por%20JMP%20tiene%20el%20menor%20grado%20de%20dispersi%C3%B3n.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CSTRONG%3E2%3C%2FSTRONG%3E%3CSTRONG%3E%20solicitud%3C%2FSTRONG%3E%3CSTRONG%3E%20Oreja%3C%2FSTRONG%3E%3CSTRONG%3E%20M%C3%A9todo%20de%20modelado%20del%20m%C3%A9todo%20forestal.%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27518i079CCBC58B215CC6%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20alt%3D%22%E5%9B%BE7%E5%9B%BE8.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3EFigura%207%20Figura%208%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EComo%20se%20puede%20observar%20en%20las%20Figuras%207%20y%208%2C%20el%20modelo%20obtenido%20despu%C3%A9s%20del%20entrenamiento%20y%20ajuste%20mediante%20el%20m%C3%A9todo%20del%20bosque%20Bootstrap%20tiene%20un%20coeficiente%20de%20determinaci%C3%B3n%20R2%20de%200%2C863%20y%20un%20RMSE%20de%201%2C48%20despu%C3%A9s%20de%20la%20verificaci%C3%B3n%20utilizando%20los%20datos%20de%20validaci%C3%B3n%2C%20logrando%20una%20mayor%20precisi%C3%B3n%20de%20predicci%C3%B3n.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EDe%20manera%20similar%2C%20la%20Figura%208%20muestra%20el%20diagrama%20de%20dispersi%C3%B3n%20de%20los%20valores%20predichos%20de%20los%20tres%20modelos%20(de%20izquierda%20a%20derecha%2C%20el%20modelo%20JMP%2C%20el%20modelo%20PT%20y%20el%20modelo%20existente%20de%20Baosteel)%20y%20los%20valores%20reales.%20Se%20puede%20ver%20que%20el%20modelo%20ajustado%20JMP%20tiene%20el%20menor%20grado%20de%20dispersi%C3%B3n.%20)%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3ESe%20puede%20ver%20en%20la%20aplicaci%C3%B3n%20de%20los%20dos%20algoritmos%20de%20modelo%20inteligente%20anteriores%20que%3CSTRONG%3E%20Oreja%3C%2FSTRONG%3E%3CSTRONG%3E%20El%20algoritmo%20forestal%20tiene%20un%20mejor%20efecto%20de%20ajuste%20y%20una%20mayor%20precisi%C3%B3n%20de%20predicci%C3%B3n%20que%20la%20red%20neuronal.%20Ambos%20m%C3%A9todos%20de%20modelado%20han%20logrado%20los%20objetivos%20del%20equipo%20del%20proyecto.%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3ESin%20embargo%2C%20esto%20es%20s%C3%B3lo%20el%20resultado%20del%20an%C3%A1lisis%20te%C3%B3rico%20de%20los%20datos.%20El%20efecto%20real%20de%20la%20aplicaci%C3%B3n%20a%C3%BAn%20debe%20determinarse%20mediante%20los%20datos%20de%20control%20reales%20en%20el%20sitio.%20C%C3%B3mo%20aplicar%20los%20modelos%20establecidos%20por%20estos%20algoritmos%20inteligentes%20al%20control%20real%20es%20el%20trabajo%20verdaderamente%20significativo.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3ELa%20funci%C3%B3n%20de%20exportaci%C3%B3n%20de%20modelos%20del%20software%20JMP%20Pro%20tiende%20un%20puente%20entre%20la%20teor%C3%ADa%20y%20la%20pr%C3%A1ctica%20para%20el%20equipo%2C%20cierra%20la%20brecha%20entre%20el%20an%C3%A1lisis%20fuera%20de%20l%C3%ADnea%20y%20la%20aplicaci%C3%B3n%20en%20l%C3%ADnea%20y%20aclara%20el%20abismo%20natural.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CFONT%20size%3D%225%22%3E%3CSTRONG%3Ecuatro%3C%2FSTRONG%3E%3CSTRONG%3E%3C%2FSTRONG%3E%3CSTRONG%3E%20Verificaci%C3%B3n%20online%20de%20modelos%20inteligentes.%3C%2FSTRONG%3E%3C%2FFONT%3E%3C%2FP%3E%0A%3CP%3E1%3CSTRONG%3E%20Exportaci%C3%B3n%20de%20modelo%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3ESe%20utiliz%C3%B3%20el%20software%20JMP%20Pro%20para%20convertir%20estos%20dos%20algoritmos%20de%20modelos%20inteligentes%20en%20c%C3%B3digos%20de%20lenguaje%20C%2C%20dando%20as%C3%AD%20un%20paso%20cr%C3%ADtico%20para%20la%20aplicaci%C3%B3n%20pr%C3%A1ctica%20de%20modelos%20de%20control%20inteligentes%20en%20el%20campo.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E2%3CSTRONG%3E%20Verificaci%C3%B3n%20online%20de%20modelos.%3C%2FSTRONG%3E%3C%2FP%3E%0A%3CP%3EEl%20c%C3%B3digo%20de%20lenguaje%20C%20convertido%20a%20partir%20de%20los%20dos%20algoritmos%20de%20modelo%20inteligente%20se%20compil%C3%B3%20e%20integr%C3%B3%20en%20el%20proceso%20preestablecido%20de%20la%20m%C3%A1quina%20anal%C3%B3gica%20fuera%20de%20l%C3%ADnea%20de%20la%20computadora%20del%20proceso%20de%20laminaci%C3%B3n%20en%20caliente%201580%2C%20y%20luego%20se%20prob%C3%B3%20utilizando%20los%20datos%20en%20septiembre%20de%202019%2C%20es%20decir%2C%20los%20factores%20de%20entrada%20son%20los%20Datos%20reales%20de%20la%20banda%20de%20acero%20en%20septiembre.%20El%20modelo%20inteligente%20se%20utiliza%20para%20predecir%20el%20ancho%20natural%20del%20laminado%20de%20acabado.%20Luego%2C%20se%20utiliza%20la%20misma%20f%C3%B3rmula%20de%20c%C3%A1lculo%20de%20desviaci%C3%B3n%20est%C3%A1ndar%20de%20Pratt%20PT%20Company%20para%20calcular%20la%20desviaci%C3%B3n%20est%C3%A1ndar%20del%20modelo%20Baosteel%2C%20el%20modelo%20PT%20Company%20y%20JMP.%20modelo%20inteligente%20respectivamente.%20Consulte%20la%20Figura%209%20a%20continuaci%C3%B3n%3A%3C%2FP%3E%0A%3CP%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%20lia-image-align-inline%22%20image-alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3CSPAN%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cspan%20class%3D%22lia-inline-image-display-wrapper%22%20image-alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20style%3D%22width%3A%20999px%3B%22%3E%3Cimg%20src%3D%22https%3A%2F%2Fcommunity.jmp.com%2Ft5%2Fimage%2Fserverpage%2Fimage-id%2F27519i876978A1CD879F24%2Fimage-size%2Flarge%3Fv%3Dv2%26amp%3Bpx%3D999%22%20role%3D%22button%22%20title%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20alt%3D%22%E5%9B%BE9%E7%A5%9E%E7%BB%8F%E7%BD%91%E7%BB%9C%E6%A8%A1%E5%9E%8B%20VS%20%E6%A3%AE%E6%9E%97%E6%B3%95.png%22%20%2F%3E%3C%2Fspan%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FSPAN%3E%3C%2FP%3E%0A%3CP%3EFigura%209%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EComo%20se%20puede%20ver%20en%20la%20Figura%209%2C%3CSTRONG%3E%20A%20partir%20de%20los%20resultados%20de%20la%20predicci%C3%B3n%20real%2C%3C%2FSTRONG%3E%3CSTRONG%3E%20JMP%3C%2FSTRONG%3E%3CSTRONG%3E%20Los%20resultados%20de%20predicci%C3%B3n%20del%20modelo%20ajustado%20por%20el%20m%C3%A9todo%20forestal%20han%20alcanzado%20los%20resultados%20de%20predicci%C3%B3n%20del%20modelo%20existente%20de%20Baosteel.%3C%2FSTRONG%3E%20!%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3E%3CFONT%20size%3D%225%22%3E%3CSTRONG%3Ecinco%3C%2FSTRONG%3E%3CSTRONG%3E%3C%2FSTRONG%3E%3CSTRONG%3E%20Resumen%20del%20proyecto%20y%20experiencia.%3C%2FSTRONG%3E%3C%2FFONT%3E%3C%2FP%3E%0A%3CP%3EAl%20utilizar%20los%20m%C3%A9todos%20inteligentes%20de%20an%C3%A1lisis%20y%20modelado%20del%20software%20de%20an%C3%A1lisis%20estad%C3%ADstico%20JMP%20Pro%2C%20el%20equipo%20entren%C3%B3%2C%20verific%C3%B3%20y%20ajust%C3%B3%20los%20valores%20previstos%2C%20los%20valores%20reales%20y%20los%20par%C3%A1metros%20relacionados%20del%20ancho%20natural%20del%20acabado%20laminado%20en%20caliente%201580%20y%20obtuvo%20la%20red%20neuronal.%20y%20La%20precisi%C3%B3n%20de%20la%20predicci%C3%B3n%20te%C3%B3rica%20y%20la%20precisi%C3%B3n%20de%20la%20verificaci%C3%B3n%20de%20los%20dos%20modelos%20de%20control%20inteligente%20del%20bosque%20Bootstrap%20han%20alcanzado%20completamente%20la%20precisi%C3%B3n%20del%20modelo%20de%20red%20neuronal%20de%20la%20empresa%20alemana%20Pratt.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAl%20mismo%20tiempo%2C%20con%20la%20ayuda%20de%20la%20funci%C3%B3n%20de%20publicaci%C3%B3n%20de%20modelos%20del%20software%20JMP%20Pro%2C%20el%20equipo%20obtuvo%20su%20c%C3%B3digo%20de%20programa%20en%20lenguaje%20C%20y%20realiz%C3%B3%20pruebas%20reales%20del%20proceso%20preestablecido%20en%20la%20m%C3%A1quina%20anal%C3%B3gica%20controlada%20por%20el%20programa%20en%20el%20sitio%2C%20y%20logr%C3%B3%20muy%20buenos%20resultados.%20resultados%2C%20proporcionando%20as%C3%AD%20la%20base%20para%20el%20siguiente%20paso.%20Las%20pruebas%20y%20verificaci%C3%B3n%20de%20producci%C3%B3n%20en%20l%C3%ADnea%20en%20un%20solo%20paso%20sentaron%20una%20base%20s%C3%B3lida.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAdem%C3%A1s%2C%20durante%20el%20proceso%20de%20an%C3%A1lisis%20y%20modelado%2C%20el%20equipo%20tambi%C3%A9n%20resumi%C3%B3%20las%20siguientes%20experiencias%20para%20su%20referencia%3A%3C%2FP%3E%0A%3CUL%3E%0A%3CLI%3ELa%20duraci%C3%B3n%20de%20los%20materiales%20de%20capacitaci%C3%B3n%20no%20deber%C3%ADa%20ser%20demasiado%20larga%2C%20preferiblemente%20de%203%20a%204%20meses%3B%3C%2FLI%3E%0A%3CLI%3ELa%20cantidad%20de%20datos%20para%20capacitaci%C3%B3n%20y%20verificaci%C3%B3n%20no%20debe%20ser%20demasiado%20grande%2C%20siendo%20apropiado%20entre%2030.000%20y%2040.000%20registros%3B%3C%2FLI%3E%0A%3CLI%3ESi%20est%C3%A1%20equipado%20con%20una%20funci%C3%B3n%20de%20aprendizaje%20autoajustable%2C%20la%20precisi%C3%B3n%20de%20la%20predicci%C3%B3n%20puede%20ser%20mayor.%3C%2FLI%3E%0A%3C%2FUL%3E%0A%3CP%3E%3C%2FP%3E%0A%3CP%3ESe%20puede%20ver%20que%20con%20la%20propuesta%20y%20el%20avance%20de%20la%20estrategia%20Made%20in%20China%202025%2C%20la%20industria%20manufacturera%20se%20est%C3%A1%20transformando%20y%20actualizando%20gradualmente%20de%20la%20fabricaci%C3%B3n%20a%20la%20%22fabricaci%C3%B3n%20inteligente%22%2C%20y%20es%20imperativo%20que%20la%20fabricaci%C3%B3n%20inteligente%20potencie%20las%20futuras%20plantas%20sider%C3%BArgicas.%20El%20intento%20y%20la%20exploraci%C3%B3n%20del%20proyecto%20de%20ensanchamiento%20natural%20de%20laminado%20en%20caliente%201580%20de%20Baosteel%20proporciona%20una%20buena%20referencia%20y%20base%20para%20otras%20l%C3%ADneas%20de%20productos%20de%20Baosteel%20en%20el%20camino%20hacia%20la%20fabricaci%C3%B3n%20inteligente.%20Las%20herramientas%20de%20an%C3%A1lisis%20cient%C3%ADfico%20seguramente%20ser%C3%A1n%20indispensables%20para%20que%20la%20industria%20manufacturera%20avance%20cada%20vez%20m%C3%A1s.%20motor%20poderoso.%3C%2FP%3E%3C%2FLINGO-BODY%3E%3CLINGO-TEASER%20id%3D%22lingo-teaser-324081%22%20slang%3D%22zh-TW%22%3E%3CP%3EUna%20introducci%C3%B3n%20detallada%20a%20la%20pr%C3%A1ctica%20del%20proyecto%20sobre%20c%C3%B3mo%20Baosteel%20aplica%20algoritmos%20inteligentes%20basados%20en%20JMP%20Pro%20para%20predecir%20el%20ancho%20natural%20del%20laminado%20en%20caliente%20y%20del%20laminado%20de%20acabado%201580%20tambi%C3%A9n%20es%20un%20complemento%20al%20discurso%20pronunciado%20por%20el%20Sr.%20Zhu%20Jianqin%2C%20ingeniero%20jefe%20de%20modelado%20digital%20de%20Baosteel.%20Co.%2C%20Ltd.%20Departamento%20de%20Equipos%20en%20la%20JMP%20Discovery%20Summit%20China%202020.%3C%2FP%3E%3C%2FLINGO-TEASER%3E
言語を選択 翻訳バーを非表示
Ritalv
Community Manager Community Manager
【案例研究】 | 應用智慧演算法預測寶鋼1580熱軋精軋自然寬展

智慧製造是股份公司在新形勢下技術發展的必然趨勢,也是在激烈的市場競爭中贏得主動、立於不敗的重要保證。1580 熱軋智慧車間建設是寶鋼第一個作為國家工信部試點的示範智慧化專案,通過引進和升級現有技術和裝備,寶鋼邁出了鋼鐵工業智慧化升級的步伐。

 

pexels-photo-373543.jpeg

 

本文將詳細介紹寶鋼如何應用基於JMP Pro的智慧演算法,預測1580熱軋精軋的自然寬展的項目實踐,也是2020 JMP Discovery Summit China寶鋼股份設備部數模主任工程師朱健勤先生當天演講內容的補充。

 

   專案背景為何需要JMP Pro

寶鋼 1580 熱軋智慧車間建設工程中,精軋動態寬度控制(DWC)系統由德國普瑞特公司(Primetals Technologies Germany GmbH,簡稱 PT)將其精軋自然寬展預測及自我調整的神經網路模型嵌入到 L2 系統中,通過 L2 過程參數的輸入,計算精軋自然寬展的預測值,用於確定粗軋目標寬度及軋後自我調整,並作為參考值傳送至 L1,由 L1 實現精軋寬度的動態控制,從而顯著地提高了 1580 熱軋精軋出口寬度的控制精度,其模型預測值與實際值的平均偏差僅有 1.38mm(2019 年 1 月~9 月),參見圖 1。

图 1..png

圖1

 

然而,德國普瑞特公司提供的精軋自然寬展預測及自我調整神經網路模型是一個黑箱模型,其原始程式碼是經過加密的,無法閱讀或修改,也沒有模型結構說明,因而無法進行參數調整、模型調優等改進工作。

 

因此,團隊希望能夠通過應用JMP統計分析產品家族的其他機器學習、深度學習等人工智慧建模工具,開發出一個智慧模型,用於寶鋼股份1580熱軋廠的精軋自然寬展預測,籍以替代德國PT公司的黑箱模型,並使該自主開發模型的預測值與實際值的偏差達到月平均值≤1.7mm,從而基本達到德國黑箱模型的控制水準。

 

基於此,團隊開始借助於JMP Pro中的預測建模工具作進一步的嘗試和探索。

 

 

智慧化預測模型的建立

 

資料的收集與清理

團隊採集了 1580 熱軋 2019.01~2019.09 共九個月的精軋生產資料,並編制 SAS 程式對其中 2019 年 1~8 月的資料進行清理與整合,共提取出有效資料 85423 塊帶鋼,並匯總形成 Excel 資料檔案。

 

數據篩選

 

01  運用智慧演算法篩選關鍵因數

應用JMP Pro統計分析軟體中的 Bootstrap 森林演算法對該資料集(2019 年1~8 月,85423 條記錄、206 個影響因數)進行擬合訓練,然後從訓練的結果中根據對實際自然寬展貢獻度(依據回歸平方和)的大小,選取方差平方和>3000 的影響因數,共31 個,參見圖 2。

图2.png
圖2

 

02  普瑞特(PT)公司經驗傳承

由德國普瑞特(PT)公司提供的自然寬展神經網路模型分析診斷系統的輸入/輸出介面,不難看出該神經網路黑箱模型需要化學成分、精軋入口寬度、出口寬度等 33 個輸入變數, 參見圖 3。

图3.png

圖3

 

03  資料合成 

 

團隊將 JMP Pro的Bootstrap森林演算法篩選的關鍵因數與普瑞特(PT)公司神經網路模型所需的輸入變數進行對比,可以發現有 4 個影響因數是相同的,它們是預計算的精軋目標寬度及化學成分矽(Si)、鈮(Nb)等。

 

同時,根據生產品質分析經驗,不難看出粗軋實際寬度、粗軋後半寬度實際值、粗軋目標寬度等因數之間存在自相關性,並且這些變數與團隊需要擬合的目標實際自然寬展(spread_ACT)也存在線性相關性。應用JMP Pro 的多元線性相關性分析工具對這八個影響因數及自然寬展實際值進行分析,可以看到八個影響因數之間呈強相關關係(相關性係數 r>0.99),且與自然寬展實際值的相關性也比較強(r>55%)。參見圖 4。

图4.png

圖4

 

除此以外,出鋼記號(alloyCode)有 1000 多個,雖然其對自然寬展有影響,且貢獻度最大,但由於維度太大,無法進行擬合,可以另行採用表格方式進行自我調整學習。

 

因而經過討論,最終決定將這 9 個變數刪去,加上重複的 4 個變數共刪去 13 個影響因數。這樣 31 個影響因數還剩餘 18 個,加上普瑞特公司經驗傳承的 33 個因數,最終確定 51 個影響因數。

 

3  數據準備

針對最終確定的影響精軋自然寬展的 51 個影響因數,團隊採集了 2019 年 6 月至 8 月共三個月的資料(共 30533 塊帶鋼),應用 JMP Pro讀取這些資料,並將精軋自然寬展實際值作為目標值(Y,回應變數),51 個影響因數作為擬合數據(X,引數),並隨機設置該擬合資料集的 80%作為訓練集,20%作為驗證集。

 

 

  智慧模型的訓練及離線擬合

團隊嘗試應用 JMP Pro分析軟體中的兩種智慧建模方法:神經網路及 Bootstrap 森林法分別對上述資料集進行訓練和驗證,得到如下結果:

 

應用神經網路建模方法

图5 图6.png

                                  圖5                                                                  圖6

 

由圖5和圖6可知,應用神經網路演算法訓練擬合後得到的模型,用驗證資料驗證後的決定係數 R2 為 0.858,RMSE 為 1.506,絕對偏差的平均值為 1.16,理論上已經達到了團隊設定的目標。

 

圖6給出了三種模型的預測值(從左至右依次是 JMP 模型、PT 模型及寶鋼現有模型)與實際值的散佈圖,從中可見 JMP 擬合的模型其離散程度是最小的。

 

2  應用Bootstrap森林法建模方法

图7图8.png

                                         圖7                                                                圖8

 

由圖7和圖8可知,應用 Bootstrap 森林法訓練擬合後得到的模型,用驗證資料驗證後的決定係數 R2達到了 0.863,RMSE 為 1.48,獲得了更高的預測精度。

 

同樣,圖8給出了三種模型的預測值(從左至右依次是 JMP 模型、PT 模型及寶鋼現有模型)與實際值的散佈圖,從中可見 JMP 擬合的模型其離散程度是最小的。)

 

由上述兩種智慧模型演算法的應用可以看出,Bootstrap 森林演算法的擬合效果比神經網路要好,預測精度更高,這兩種建模方法都達到了專案團隊的目標。

 

但是,這僅僅是對資料的理論分析結果,究竟實際應用效果如何,還是要用現場的實際控制資料來說話。如何將這些智慧演算法建立的模型應用於實際控制才是真正有意義的工作。

 

JMP Pro軟體的模型匯出功能為團隊架起了理論到實踐的橋樑,跨越了離線分析到線上應用的鴻溝,使得天塹變通途。

 

  智慧模型的線上驗證

模型的匯出

應用JMP Pro 軟體將這兩種智慧模型演算法轉換成了 C 語言代碼,從而為智慧控制模型在現場的實際應用,跨出了關鍵性的一步。

 

模型的線上驗證

將兩種智慧模型演算法轉換成的 C 語言程式碼經過編譯後集成到 1580 熱軋粗軋過程電腦離線類比機的預設定進程中,然後應用 2019 年 9 月份的資料進行測試,即輸入因數是 9 月份的帶鋼實際資料,運用智慧模型預測精軋自然寬展,然後採用普瑞特 PT 公司相同的標準差計算公式分別計算寶鋼模型、PT 公司模型和 JMP智慧模型的標準差,參見下圖 9:

图9神经网络模型 VS 森林法.png

圖9

 

由圖 9 可見,從實際預測結果看,JMP 森林法擬合得到的模型,其預測結果已經達到了寶鋼現有模型的預測結果

 

  專案小結及經驗

通過運用JMP Pro統計分析軟體中的智慧分析及建模方法,團隊對 1580 熱軋精軋自然寬展的預測值、實際值以及相關參數進行了訓練、驗證和擬合,得到了神經網路和Bootstrap 森林兩種智慧控制模型,其理論預測精度和驗證精度完全達到了德國普瑞特公司的神經網路模型精度。

 

同時,借助於 JMP Pro軟體的模型發佈功能,團隊獲得了其 C 語言的程式碼,並在現場程式控制電腦的類比機上進行了預設定進程實際測試,取得了非常好的效果,從而為下一步開展線上生產試驗和驗證奠定了充實的基礎。

 

除此以外,在分析和建模過程中,團隊也總結了以下幾點經驗供大家參考:

  • 訓練資料的時間跨度不宜過長,以 3~4 個月為宜;
  • 訓練驗證的資料量不宜過大,以 3~4 萬條記錄為宜;
  • 如果配以自我調整學習功能,可以使預測精度更高。

可以看到的是,隨著中國製造2025戰略的提出和推進,製造業逐步從製造向“智造”轉型與升級,智慧製造賦能未來鋼廠勢在必行。寶鋼1580熱軋精軋自然寬展項目的嘗試與探索,為寶鋼其他產品線的智慧製造之路提供了很好的參考和依據,科學的分析工具必將成為製造業走向詩和遠方不可或缺的強大引擎。

最終変更日: Oct 20, 2020 5:05 AM