. "0018-8069" . "4"^^ . "Kv\u00ED\u010Dala, Miroslav" . "67" . . "RIV/61989100:27360/14:86092123!RIV15-MSM-27360___" . "4"^^ . . . "Zimn\u00FD, Ond\u0159ej" . . . "Hutnick\u00E9 listy" . . "neural networks, defects, steel"@en . . "38214" . "CZ - \u010Cesk\u00E1 republika" . . "[DEF5119E34B2]" . "27360" . "Pou\u017Eit\u00ED neuronov\u00FDch s\u00EDt\u00ED p\u0159i p\u0159edpov\u011Bdi v\u00FDskytu vnit\u0159n\u00EDch defekt\u016F v bloc\u00EDch z Cr-Mo oceli po v\u00E1lcov\u00E1n\u00ED" . "4"^^ . "S" . . "Pou\u017Eit\u00ED neuronov\u00FDch s\u00EDt\u00ED p\u0159i p\u0159edpov\u011Bdi v\u00FDskytu vnit\u0159n\u00EDch defekt\u016F v bloc\u00EDch z Cr-Mo oceli po v\u00E1lcov\u00E1n\u00ED" . . "Using neural networks in predicting the occurrence of internal defects in blocks of Cr- Mo steel after rolling"@en . . "P\u0159edlo\u017Een\u00E1 pr\u00E1ce je v\u011Bnov\u00E1na aplikaci neuronov\u00FDch s\u00EDt\u00ED p\u0159i \u0159e\u0161en\u00ED velmi komplexn\u00EDho probl\u00E9mu souvisej\u00EDc\u00EDho s vnit\u0159n\u00EDmi vadami, kter\u00E9 se vyskytuj\u00ED ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch z vanadu micro-legovan\u00E9 oceli 25CrMo4 za tepla. Vzhledem k tomu, vnit\u0159n\u00ED vady jsou zji\u0161\u0165ov\u00E1ny p\u0159i ultrazvukov\u00E9 kontrole v ji\u017E zchlazen\u00FDch sochorech, lze \u0159\u00EDci, \u017Ee \u0159e\u0161en\u00FD probl\u00E9m je velmi slo\u017Eit\u00FD. P\u0159\u00EDsp\u011Bvek ukazuje, \u017Ee neuronov\u00E1 s\u00ED\u0165 aplikace m\u016F\u017Ee b\u00FDt velmi u\u017Eite\u010Dn\u00FDm n\u00E1strojem pro \u0159e\u0161en\u00ED slo\u017Eit\u00FDch v\u00FDrobn\u00EDch probl\u00E9m\u016F, jako je nap\u0159\u00EDklad v\u00FDskyt trhlin ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch za tepla. Pou\u017Eit\u00ED um\u011Bl\u00E9 neuronov\u00E9 s\u00EDt\u011B (ANN) p\u0159edstavuje distribuovan\u00E9 paraleln\u00ED zpracov\u00E1n\u00ED informac\u00ED, co\u017E znamen\u00E1, \u017Ee z\u00E1znam informace, zpracov\u00E1n\u00ED a p\u0159enos se prov\u00E1d\u00ED prost\u0159ednictv\u00EDm cel\u00E9 neuronov\u00E9 s\u00EDt\u011B a pomoc\u00ED jednotliv\u00FDch pam\u011B\u0165ov\u00FDch m\u00EDst. Z\u00E1kladem matematick\u00E9ho modelu neuronov\u00E9 s\u00EDt\u011B je form\u00E1ln\u00ED neuron, kter\u00FD p\u0159edstavuje zjednodu\u0161en\u00FD zp\u016Fsob popisu funkc\u00ED biologick\u00E9ho neuronu pomoc\u00ED matematick\u00FDch vztah\u016F. Neuronov\u00E9 s\u00EDt\u011B s nejlep\u0161\u00EDmi v\u00FDsledky u\u010Den\u00ED byly vybr\u00E1ny pro predikci vad. Jedn\u00E1 se t\u0159\u00EDvrstv\u00E9 perceptronov\u00E9 s\u00EDt\u011B s topologi\u00ED 22-18-2. To znamen\u00E1, \u017Ee vstupn\u00ED vrstva obsahuje dvacet dva neuron\u016F; skryt\u00E9 vrstvy osmn\u00E1ct a v\u00FDstupn\u00ED vrstva dva neurony. Na neuronov\u00FDch s\u00EDt\u00EDch byla provedena rovn\u011B\u017E tak\u00E9 anal\u00FDza citlivosti. Tato anal\u00FDza vyjad\u0159uje vliv jednotliv\u00FDch vstupn\u00EDch prom\u011Bnn\u00FDch na dan\u00E9m syst\u00E9mu. I kdy\u017E \u0159e\u0161en\u00FD probl\u00E9m v\u00FDskytu vnit\u0159n\u00EDch vad ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch z vanadu micro-legovan\u00E9 oceli 25CrMo4 za tepla byl vysv\u011Btlen za pou\u017Eit\u00ED tradi\u010Dn\u00EDch materi\u00E1lov\u011B in\u017Een\u00FDrsk\u00FDch postup\u016F, aplikace neuronov\u00FDch s\u00EDt\u00ED byla schopna p\u0159edv\u00EDdat vznik vnit\u0159n\u00ED vady s pravd\u011Bpodobnost\u00ED na osmdes\u00E1t \u0161est procent. Aplikace neuronov\u00E9 s\u00EDt\u011B p\u0159edstavuje velmi rychl\u00E9, levn\u00E9 a efektivn\u00ED \u0159e\u0161en\u00ED r\u016Fzn\u00FDch metalurgick\u00FDch probl\u00E9m\u016F."@cs . "Jan\u010D\u00EDkov\u00E1, Zora" . "The scope of the presented paper is dedicated to neural network application in solution of a very complex problem linked to internal defects that occur in hot rolled billets from vanadium micro-alloyed 25CrMo4 steel. Since internal defects are indicated during an ultrasonic inspection in already cooled billets, it can be said that the studied problem is very complex. The paper demonstrates that neural network application may be a very useful tool for solving complicated production problems such as the occurrence of cracks in hot rolled billets. Artificial neural networks (ANN) use distributed parallel processing of information during the execution of the calculations, which means that information recording, processing and transferring are carried out by means of the whole neural network, and then by means of particular memory places. The basis of a mathematical model of the neural network is a formal neuron which uses a simplified way to describe a function of a biological neuron by means of mathematic relations. Neural networks with the best learning results were selected for the prediction of defects. They included a three-layer perceptron network with 22-18-2 topology. This means that the input layer contained twenty-two neurons; the hidden layer eighteen and the output layer two neurons. This neural network also included a sensitivity analysis. This analysis expresses the impact of the individual input variables on the given system. Although the studied problem of internal defect present in hot rolled billets from vanadium micro-alloyed steel 25CrMo4 was fully explained using traditional material engineering procedures, neural network application was able to predict internal defect occurrence with eighty-six percents probability. Neural network has a potential to be very fast, inexpensive and useful tool in solving a wide scale of similar metallurgical problems."@en . "RIV/61989100:27360/14:86092123" . "Meca, Roman" . "1" . . "Pou\u017Eit\u00ED neuronov\u00FDch s\u00EDt\u00ED p\u0159i p\u0159edpov\u011Bdi v\u00FDskytu vnit\u0159n\u00EDch defekt\u016F v bloc\u00EDch z Cr-Mo oceli po v\u00E1lcov\u00E1n\u00ED"@cs . . . "P\u0159edlo\u017Een\u00E1 pr\u00E1ce je v\u011Bnov\u00E1na aplikaci neuronov\u00FDch s\u00EDt\u00ED p\u0159i \u0159e\u0161en\u00ED velmi komplexn\u00EDho probl\u00E9mu souvisej\u00EDc\u00EDho s vnit\u0159n\u00EDmi vadami, kter\u00E9 se vyskytuj\u00ED ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch z vanadu micro-legovan\u00E9 oceli 25CrMo4 za tepla. Vzhledem k tomu, vnit\u0159n\u00ED vady jsou zji\u0161\u0165ov\u00E1ny p\u0159i ultrazvukov\u00E9 kontrole v ji\u017E zchlazen\u00FDch sochorech, lze \u0159\u00EDci, \u017Ee \u0159e\u0161en\u00FD probl\u00E9m je velmi slo\u017Eit\u00FD. P\u0159\u00EDsp\u011Bvek ukazuje, \u017Ee neuronov\u00E1 s\u00ED\u0165 aplikace m\u016F\u017Ee b\u00FDt velmi u\u017Eite\u010Dn\u00FDm n\u00E1strojem pro \u0159e\u0161en\u00ED slo\u017Eit\u00FDch v\u00FDrobn\u00EDch probl\u00E9m\u016F, jako je nap\u0159\u00EDklad v\u00FDskyt trhlin ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch za tepla. Pou\u017Eit\u00ED um\u011Bl\u00E9 neuronov\u00E9 s\u00EDt\u011B (ANN) p\u0159edstavuje distribuovan\u00E9 paraleln\u00ED zpracov\u00E1n\u00ED informac\u00ED, co\u017E znamen\u00E1, \u017Ee z\u00E1znam informace, zpracov\u00E1n\u00ED a p\u0159enos se prov\u00E1d\u00ED prost\u0159ednictv\u00EDm cel\u00E9 neuronov\u00E9 s\u00EDt\u011B a pomoc\u00ED jednotliv\u00FDch pam\u011B\u0165ov\u00FDch m\u00EDst. Z\u00E1kladem matematick\u00E9ho modelu neuronov\u00E9 s\u00EDt\u011B je form\u00E1ln\u00ED neuron, kter\u00FD p\u0159edstavuje zjednodu\u0161en\u00FD zp\u016Fsob popisu funkc\u00ED biologick\u00E9ho neuronu pomoc\u00ED matematick\u00FDch vztah\u016F. Neuronov\u00E9 s\u00EDt\u011B s nejlep\u0161\u00EDmi v\u00FDsledky u\u010Den\u00ED byly vybr\u00E1ny pro predikci vad. Jedn\u00E1 se t\u0159\u00EDvrstv\u00E9 perceptronov\u00E9 s\u00EDt\u011B s topologi\u00ED 22-18-2. To znamen\u00E1, \u017Ee vstupn\u00ED vrstva obsahuje dvacet dva neuron\u016F; skryt\u00E9 vrstvy osmn\u00E1ct a v\u00FDstupn\u00ED vrstva dva neurony. Na neuronov\u00FDch s\u00EDt\u00EDch byla provedena rovn\u011B\u017E tak\u00E9 anal\u00FDza citlivosti. Tato anal\u00FDza vyjad\u0159uje vliv jednotliv\u00FDch vstupn\u00EDch prom\u011Bnn\u00FDch na dan\u00E9m syst\u00E9mu. I kdy\u017E \u0159e\u0161en\u00FD probl\u00E9m v\u00FDskytu vnit\u0159n\u00EDch vad ve v\u00E1lcovan\u00FDch \u0161palc\u00EDch z vanadu micro-legovan\u00E9 oceli 25CrMo4 za tepla byl vysv\u011Btlen za pou\u017Eit\u00ED tradi\u010Dn\u00EDch materi\u00E1lov\u011B in\u017Een\u00FDrsk\u00FDch postup\u016F, aplikace neuronov\u00FDch s\u00EDt\u00ED byla schopna p\u0159edv\u00EDdat vznik vnit\u0159n\u00ED vady s pravd\u011Bpodobnost\u00ED na osmdes\u00E1t \u0161est procent. Aplikace neuronov\u00E9 s\u00EDt\u011B p\u0159edstavuje velmi rychl\u00E9, levn\u00E9 a efektivn\u00ED \u0159e\u0161en\u00ED r\u016Fzn\u00FDch metalurgick\u00FDch probl\u00E9m\u016F." . "Using neural networks in predicting the occurrence of internal defects in blocks of Cr- Mo steel after rolling"@en . . "Pou\u017Eit\u00ED neuronov\u00FDch s\u00EDt\u00ED p\u0159i p\u0159edpov\u011Bdi v\u00FDskytu vnit\u0159n\u00EDch defekt\u016F v bloc\u00EDch z Cr-Mo oceli po v\u00E1lcov\u00E1n\u00ED"@cs . .