Riba:
- Ability don auna adadi kamar girman hatsi, rabon lokaci da porosity daga hotunan microstructure tare da sarrafa hoto da AI
- Ability don daidaita daidaitaccen yanki, ƙofa da matakan daidaitawa da hana kurakuran sikeli.
- Ƙarfin ƙetare ingantaccen fassarar ƙayyadaddun ƙayyadaddun tsarin AI tare da daidaitaccen hanyar ƙarfe da kuma lura da ƙwararru.
Tauri, tauri da gajiyawar rayuwar karfe suna ɓoye a cikin ƙananan ƙwayoyin cuta, wanda ba a iya gani ga ido tsirara: girman hatsi, rabon lokaci, porosity (void), inclusions (sauran oxide / sulfur a cikin ƙarfe) da rarraba hazo. Metallography shine kimiyyar nazarin abu a ƙarƙashin na'urar hangen nesa ta hanyar yankan, gogewa da etching (kemikal ta hanyar sinadari da sanya matakan bayyane). Ƙididdigar ƙirar ƙira tana auna ƙididdige ƙididdiga (lambar girman hatsi, kashi na lokaci, rabon porosity) daga waɗannan hotuna. Hankali na wucin gadi da sarrafa hoto na iya rage sa'o'in kirgawa na hannu zuwa mintuna; amma ma'aunin da ba daidai ba ko rashin daidaituwa ya sa duk lambobi su yi kuskure. A cikin wannan naúrar za ku koyi yadda ake auna hotunan microstructure tare da AI da sarrafa hoto da kuma tabbatar da kowane ma'auni.
Matakan asali na ma'aunin tushen hoto
A cikin ƙididdigar ƙididdiga, kowane ma'auni yana wucewa ta sarka ɗaya; Idan zobe ya karye, sakamakon ba zai zama abin dogaro ba:
- Ingancin hoto: Mayar da hankali, haske da ingancin etching. A cikin samfurin da ba shi da kyau, iyakokin suna da duhu.
- Daidaitawa: Yana canza ma'auni zuwa pixels. Mikromitoci nawa ne kowane pixel? Idan aka tsallake wannan matakin, duk ma'aunin girman zai fito a sikelin da bai dace ba.
- Rarraba: Rarraba hoton zuwa yankuna masu sha'awa - wanda pixel shine hatsi, wanda pixel shine iyaka, wanda pixel shine porosity.
- Ƙaddamarwa: Rarrabe matakai/ sarari tare da sautin launin toka ko bakin launi. Idan matakai biyu suna da haske daban-daban, bakin kofa ya raba su.
- Aunawa da ƙididdiga: Ƙididdigar hatsi, rabon yanki, lissafin rarraba girman girman.
Kayan aikin AI masu ƙarfi (samfurin rarrabuwar horarwa-zurfafa koyo) suna yin aiki mafi kyau fiye da sauƙaƙan ƙofa, musamman akan hadaddun sifofi masu yawa. Amma sashin AI kuma hasashe ne; daidaita ma'auni da giciye sakamakon tare da daidaitaccen hanya (misali, girman hatsi ASTM E112, ƙididdigar tabo ASTM E562) ya zama tilas.
Hankali: Kuskuren da ya fi kowa kuma mafi kuskure shine mantawa don daidaita ma'auni. Hoton na iya yin kama da "mai kyau" a cikin ƙididdiga, amma idan fassarar pixel-to-micrometer ba daidai ba ne, girman hatsi zai zama 25 µm maimakon 8 µm kuma ba za ku lura ba. Fara kowane ma'auni ta hanyar tabbatar da daidaitawa.
Mataki-mataki: ma'aunin microstructure mai ƙarfin AI
- Ƙayyade manufa: Me kuke aunawa? Girman hatsi, rabon lokaci ko porosity? Da wane misali?
- Calibrate: Auna tsayin pixel na ma'aunin ma'auni, ƙididdige ƙimar µm/pixel kuma shigar da shi cikin kayan aiki.
- Pre-tsari: gyare-gyaren bambanci, rage amo. Tambayi AI wace tacewa ta dace.
- Yanki/kofa: Rarrabe matakai ko gibba; Duba da gani cewa an kama iyakokin daidai.
- Auna: Rabon yanki, adadin hatsi, ma'ana da rarrabawa.
- Tabbatar: Kwatanta sakamakon da hannu tare da daidaitaccen hanya (misali tsaka-tsakin layi vs girman hatsi); Maimaita akan hotuna da yawa.
Girman hatsi, rabon lokaci da porosity
Girman hatsi shine matsakaicin girman ƙwayar kristal na kayan; Ƙananan hatsi gabaɗaya yana ba da ƙarfi da ƙarfi (dangantakar Hall-Petch). An nuna shi ta ASTM E112 girman girman lambar G; Yayin da G ke ƙaruwa, hatsin ya zama ƙarami. Rabon lokaci shine adadin ƙarar / yanki na kowane lokaci a cikin tsari biyu- ko multiphase (misali, ma'aunin ferrite-austenite an yi niyya don ~ 50/50% a cikin bakin duplex). Porosity shine rabon ɓoyayyen da iskar gas ko raguwa ke haifarwa a simintin gyare-gyare kuma yana rage gajiya sosai.
aunawa
daidaitaccen hanya
Gudunmawar sarrafa hoto/AI
tabbaci mai mahimmanci
Girman hatsi (G)
ASTM E112 intersection / kwatanta
Gano kan iyaka ta atomatik da kirgawa
Calibration + kulawar tsaka-tsakin hannu
Rabon mataki (%)
ASTM E562 dige ƙidaya
Matsakaicin yanki tare da ƙofa
Etching uniformity, zaɓin bakin kofa
Porosity (%)
Bayanan Bayani na ASTM E1245
Rarraba sararin samaniya
Nuna wariya na kayan tarihi
hada
ASTM E45
Rarraba Siffa/ Rarraba
Kwatanta samfuri
uku mini lokuta
Case 1 - Girman hatsi ba tare da daidaitawa ba. Ma'aikacin injiniya yana ɗora hoton microstructure a cikin kayan aiki mai ƙarfi na AI kuma yana da auna girman hatsi; Sakamakon shine 6µm. Koyaya, an ɗauki hoton a girman girman 200x maimakon 500x, kuma daidaitawar zaman da ya gabata (500x) ya kasance a cikin kayan aiki. Ainihin girman hatsi shine ~ 15 µm. Ƙimar da ba daidai ba kuma tana ɓatar da ƙimar ƙarfin ta Hall-Petch. Mai fasaha ya shiga al'ada na sake daidaita ma'auni a kowane zama. Darasi: gyare-gyare samfuri ne na musamman kuma ba za a iya ɗauka ba.
Case na 2 - An yi kuskuren ɓoyayyen kayan tarihi a matsayin lokaci. Ana auna rabon ferrite/austenite a cikin bakin bakin duplex. Kuskure masu duhu duhu a cikin wuraren da aka ƙera don ferrite, yana ba da rabo na 62% ferrite; yayin da ainihin ma'auni shine ~ 48%. Yawan ferrite yana haifar da ƙararrawa na ƙarya game da lalata da tauri. Mai sarrafa ƙarfe ya sami kuskure bayan daidaita lokacin etching da duba wurare da yawa tare da kirga tabo na hannu. Darasi: matalauta samfurin shirye-shiryen wawaye har ma da mafi kyawun algorithm - datti a ciki, datti.
Case 3 - Daidaitaccen amfani, taswirar porosity. Ƙungiyar simintin gyare-gyare na son auna ƙimar porosity a ɓangaren giciye na ɓangaren aluminum. Samfurin rarrabuwa na AI da sauri ya raba gibin kuma yana ba da taswirar rarraba tare da porosity 2.3%. Tawagar ta gani da idonta cewa tarkacen da ke bayyana yayin goge-goge ba su yi kuskure ba, sun tabbatar da sakamakon bisa ga ASTM E1245, kuma suna ganin cewa porosity ya tattara a cikin yankin mai ciyarwa kuma yana inganta ƙirar simintin. AI ya kawo sauri, an yanke shawara ta hanyar daidaitattun abubuwa da kallo.
Samfuran gaggawar da za a iya kwafi
TEMPLATE MUNANAN AUNA"Rawar: Kai mataimaki ne mai ƙididdigewa metalography. Manufa: auna [girman hatsi / lokaci rabo / porosity]. Material: [...]. Girman: [...]. Standard: [misali ASTM E112]. Ba ni wani mataki-by-mataki ma'auni shirin: samfurin shirye-shiryen, etching, CALIBRATION ma'auni, Highlighting mataki, calibation matakin, resegment. a matsayin mataki na farko kuma na wajibi sakamakon Faɗa mani wace daidaitacciyar hanyar da za a bincika da hannu.
TEMPLATE KUSKUREN KYAUTA "Na raba [lokaci/rata] akan hoton ƙaramin tsari. Jera abubuwan da za su iya yin kuskure: etch artifact, polishing scratch, out-of-fige area, overlapped kan iyaka. Rubuta yadda za a bincika kowane da yadda za a gyara shi. Menene duban gani ya kamata in yi kafin karbar sakamakon?"
GRAIN GIRMAN FASSARAR TEMPLATE "Na auna ASTM E112 lambar girman hatsi G = [darajar], haɓakawa [x]. Bayyana yadda wannan ƙimar G ta dace da matsakaicin matsakaicin matsakaicin hatsi da tasirinsa akan ƙarfi cikin sharuddan Hall-Petch. Tuna da ni cewa ƙimar G ya dogara da daidaitawa; gaya mani mataki-mataki yadda za a tabbatar da hanyar haɗin gwiwa. "
PYTHON IMAGE PROCESSING SKELETON TEMPLATE"Rawar: Kai ne mataimaki na sarrafa hoto buga sakamakon. Ƙara sharhi zuwa kowane mataki Ƙidaya a cikin lambar cewa ƙima da daidaitawa dole ne a tabbatar da su da hannu."
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
RAUNI GASKIYA: "Ku gaya mani girman hatsin wannan ƙananan tsarin."
KARFIN GASKIYA: "Rawar: Kai mataimaki ne mai ƙididdige ƙididdiga. Ina so in auna girman hatsi bisa ga ASTM E112. Hoton 500x, sikelin bar 50 µm = 240pixel (watau 0.208 µm/pixel). Ka ba ni matakan da za a lissafta girman hatsi ta hanyar layin layi / nawa don ƙididdige yawan adadin hatsi ta hanyar tsaka-tsakin layi. G lambar Ba za ku iya 'karanta' sakamakon ƙarshe ba; kafa hanyar."
Mai rauni mai sauri yana tambayar AI don "karanta" hoton - amma AI bai san sikeli da etching ba kuma yana iya zuwa da lambar da aka yi. Ƙarfin faɗakarwa yana bayyana ma'auni, yana neman daidaitaccen hanya, yana buƙatar maimaitawa da tabbatarwa, kuma yana amfani da AI azaman kayan aiki na saiti, ba aunawa ba.
Kuskuren gama gari
- Manta don daidaita ma'auni ko aunawa tare da daidaitawar zaman da ya gabata.
- Dogaro da rarrabuwa ba tare da gyaggyarawa mara kyau/rashin uniform ba.
- Yin kuskuren goge goge ko tabo don porosity/lokaci.
- Ma'auni daga yanki guda ɗaya da ƙaddamarwa ga dukan samfurin; ƙididdiga na buƙatar filayen da yawa.
- Tsammanin AI ya "gani" hoton kuma yayi amfani da shi ba tare da tabbatar da ainihin lambar da yake bayarwa ba.
- Karɓar sakamakon rarraba ba tare da duba shi a gani ba.
A takaice
A cikin ƙididdigar ƙididdiga, AI da sarrafa hoto suna haɓaka girman hatsi, rabon lokaci da ma'aunin porosity da ba da damar rarrabuwa mai ƙarfi na sifofi masu rikitarwa. Koyaya, ingancin sakamakon ya dogara ne akan daidaita ma'auni, kyakkyawan shiri na samfur, da ingantaccen giciye tare da daidaitaccen hanyar (ASTM E112/E562/E1245). Sanya gyare-gyare mataki na farko, duba rarraba ta ido, maimaita a wurare da yawa, kuma amfani da AI a matsayin kayan aiki wanda ke kafa hanyar aunawa kuma ya rubuta lambar, ba wanda ya "karanta" hoton ba.
Aikin aikace-aikace
Ɗauki hoton ƙaramin tsari (samfurin ku ko samfurin lacca) da ƙimar ma'auni. Tare da samfurin "PYTHON IMAGE PROCESSING SKELETON", tambayi AI don lambar da ke ƙididdige rabon yanki; Tabbatar cewa µm/pixel calibration yana saman kuma an kunna shi a cikin lambar. Guda lambar kuma kwatanta rabon yanki da aka samu tare da kirga maki na hannu (grid 10x10) akan hoto ɗaya. Rubuta bambanci tsakanin sakamakon biyu da dalilin da zai yiwu (zabin bakin kofa, kayan tarihi) a cikin sakin layi daya.
jerin abubuwan dubawa
- [ ] Na daidaita ma'auni zuwa pixel kuma na tabbatar da shi a kowane zama.
- [ ] Na duba cewa etching/samfurin shirye-shiryen ya kasance uniform.
- [ ] Na duba sakamakon rarrabuwa da gani da cire kayan tarihi.
- [ ] Na maimaita ma'aunin aƙalla ƴan wurare daban-daban.
- [ ] Na binciko sakamakon da hannu tare da daidaitacciyar hanya (E112/E562/E1245).
- [ ] Na yi amfani da AI azaman hanya da kayan aiki na lamba, ba a matsayin "mai karatu" na hoton ba.