中(zhong)(zhong)控(kong)技(ji)術(shu)基于(yu)其在(zai)流程(cheng)(cheng)工業(ye)(ye)深耕30年(nian)的實踐經驗和技(ji)術(shu)積累，在(zai)新(xin)(xin)加坡(po)面(mian)向(xiang)全球用(yong)戶發布了(le)流程(cheng)(cheng)工業(ye)(ye)首款時(shi)(shi)序(xu)大(da)模(mo)(mo)型TPT(Time-series Pre-trained Transformer)，開啟了(le)產業(ye)(ye)落(luo)地應用(yong)的進程(cheng)(cheng)。與傳統的大(da)語言模(mo)(mo)型不同，它基于(yu)最新(xin)(xin)的時(shi)(shi)間序(xu)列數(shu)據(ju)分(fen)析架構，更注重理解生(sheng)產過程(cheng)(cheng)中(zhong)(zhong)基于(yu)時(shi)(shi)間序(xu)列數(shu)據(ju)的變(bian)化(hua)(hua)規律。TPT順利上(shang)線后(hou)能有效精(jing)準預測，并(bing)處(chu)理裝置運行過程(cheng)(cheng)中(zhong)(zhong)的信號波動；得(de)益于(yu)TPT強(qiang)大(da)的數(shu)據(ju)融合和回歸能力，用(yong)戶可以實時(shi)(shi)監測潛在(zai)優(you)化(hua)(hua)點(dian)，并(bing)根(gen)據(ju)實際需(xu)求調整操(cao)作參數(shu)；此外，企業(ye)(ye)借助TPT能減少操(cao)作員閱讀(du)數(shu)據(ju)、查(cha)找潛在(zai)風險的時(shi)(shi)間，更能輕松獲取自動生(sheng)成的各類(lei)生(sheng)產匯報、分(fen)析報告。（美通社）