Gabatarwa zuwa Hankali na Artificial a cikin Kayan Lantarki da Injiniyan Sadarwa: Iyakoki, Tabbatarwa, Nauyi da Da'a

Riba:

• Ikon rarrabewa inda AI ke adana ainihin lokaci a cikin ayyukan lantarki / sadarwa, kuma inda bin ka'ida da alhakin tsaro ya kasance tare da ƙwararren injiniya, dangane da matakin haɗari.
• Ikon yin amfani da horo na tabbatarwa mai nau'i-nau'i da yawa wanda ke gwada kowane fitarwa na AI akan aunawa, kwaikwayi, daidaitaccen rubutu, da lissafi mai zaman kansa.
• Shiga cikin al'ada na ɓoye sunan mahallin da zabar amintattun kayan aiki don kare fayilolin ƙira, bayanan cibiyar sadarwar abokin ciniki, da dukiyar ilimi

Ayyukan injiniyoyin lantarki da injiniyan sadarwa shine fassara adadin jiki marasa ganuwa (voltage, halin yanzu, mita, filin) ​​zuwa samfuran aiki da haɗin kai masu dogaro. Zayyana da'ira, sanya allon da'ira da aka buga (PCB: Board Printed Board, allon rufewa wanda aka sanya abubuwan da aka gyara), tacewa da daidaita siginar, fitar da shi daga eriya, rubuta software da aka saka (firmware: software da aka rubuta a cikin ƙwaƙwalwar dindindin wanda ke aiki da kayan aikin na'urar kai tsaye), kafa ka'idoji, sa ido kan aikin hanyar sadarwa da yin rikodin duk wannan daidai da ƙa'idodi. Alamar gama gari na waɗannan ayyuka suna aiki tare da ma'auni da bayanai da yawa, amma kowane yanke shawara yana ƙayyade ko samfurin zai yi aiki, ko watsa shirye-shiryen zai kasance a cikin iyakokin doka, ko tsarin yana da aminci ko a'a. Hankali na wucin gadi (AI a takaice; tsarin software wanda ke aiki akan rubutu, lambobi, lamba da hotuna tare da manyan bayanai da ƙirar harshe) taimako ne mai ƙarfi a cikin wannan ƙasa mai mahimmanci amma yanke shawara. Wannan rukunin yana koya muku inda za ku yi amfani da AI cikin aminci a cikin aikin lantarki / sadarwa, inda yake da haɗari, kuma me yasa dole ne ku inganta kowane fitowar AI ta hanyar aunawa, kwaikwaiyo, daidaitaccen rubutu, da hukuncin injiniya.

Babban ƙa'idar da za ku yi amfani da ita a cikin wannan tsarin ita ce: AI mataimaki ne, ba mai yanke shawara ba. A halin yanzu wadata, ƙimar impedance, mitar yanke tacewa, iyakan watsa wutar lantarki ko daidaitaccen magana ba zai iya zama gaskiya ba kawai saboda "AI ya faɗi haka"; Hasashe ne kawai har sai an tabbatar da shi ta takardar bayanai, aunawa, kwaikwayi, daidaitaccen rubutu na hukuma, da ingantaccen hukumcin injiniya. Za ku sake ganin wannan jumla a kowace naúrar tsarin, domin a cikin kayan lantarki, lambar da ba ta dace ba sau da yawa tana nufin kati ya ƙone, samfurin ya kasa takaddun shaida, ko hanyar haɗin sadarwa ta rushe.

Menene LLM, menene yake yi a cikin kayan lantarki / sadarwa?

LLM (Babban Harshe Model) software ce da ke koyon ƙira daga manyan gungu-gungu na rubutu da lamba tare da tsinkaya "kalmar gaba" bisa la'akari. Wannan shine injin kayan aikin kamar ChatGPT, Claude, Gemini. Maganar ita ce: LLM shine "masanin harshe da lambar", ba "na'urar aunawa" ko "tushen ilimi ba". Yana magana da kalmomin injiniya da kyau, amma baya auna rashin ƙarfi; yana samar da lambar da ta fi dacewa. Don haka, yana iya ba da amsoshi masu kwarin gwiwa amma ba daidai ba game da samar da halin yanzu na wani sashi, ƙayyadaddun ƙa'ida ko tebur na ƙa'idodi. Wannan shi ake kira hallucination (fabrication). Hallucination ba rashin aiki ba ne, amma siffa ce ta wannan fasaha; Saboda haka, tabbatarwa ba "ƙarin mataki ba ne" amma wani muhimmin sashi na aikin.

Ayyuka inda AI ke da ƙarfi a cikin lantarki / sadarwa:

• Bayanin ra'ayi da koyarwa: menene matching impedance, yadda OFDM ke aiki, menene ke haifar da aliasing, yadda EMC ke rushewa.
• Daftarin lamba da firmware: direban microcontroller, injin jiha, nazarin siginar Python, tsarin sarrafa kansa na gwaji.
• Tsaftace bayanai da tsari: canza ma'aunin ma'aunin tarwatsa zuwa tebur madaidaici, ɗaukar rashin daidaituwar naúrar.
• Samar da lissafin hasashe: jerin dalilai na tsari na tambayoyi kamar me yasa da'ira ba ta aiki, dalilin da yasa hanyar haɗi ta faɗo.
• Daftarin rubutu: shirin gwaji, rahoton kuskure, daftarin fayil ɗin fasaha, taƙaitaccen bayani.
• Saitin asusu: kafa kwarangwal na kasafin kuɗi na haɗin gwiwa, kasafin kuɗi, matakan ƙira tace.

Inda AI ke da rauni kuma mai haɗari:

• Madaidaicin ƙimar ƙima (sayar da halin yanzu, impedance, mitar yankewa, iyakar wutar lantarki).
• Na yau da ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙa'idodin/majalisar dokoki, ƙimar tebur da bayanin sakin.
• Bincike ba tare da ainihin ma'aunin kayan aikin ku ba.
• Yarda da yanke shawara mai mahimmanci na aminci: sanarwar EMC na daidaituwa, saki, sabis, iyakar aminci.

Rarraba tushen haɗari: tacewa kafin amfani da AI

Ba kowane aiki ba ne a matakin haɗari ɗaya. Rarraba aikin ta matakin haɗari kafin amfani da AI. Teburin da ke ƙasa yana ba da tsarin yanke shawara wanda za ku yi amfani da shi cikin wannan tsarin.

Matsayin haɗari

samfurin aiki

AI rawar

Tabbaci na wajibi

ƙananan

Bayanin ra'ayi, daftarin rubutu, kwarangwal na lamba

free amfani

Bita ya isa

matsakaici

Tsaftace bayanai, lissafin hasashe, saitin lissafi

Draft/marubuci

Dubawa da hannu + Tabbatar da takardar bayanan

babba

Ƙimar da'ira, ƙirar tacewa / eriya, dabaru na firmware

ra'ayi janareta

Simulation + ma'aunin samfuri

m

EMC / yarda da watsa shirye-shirye, gefen aminci, iyawar sabis

Draft/scan kawai

Ingantacciyar amincewar injiniya + gwajin gwaji

Tukwici: Idan kun yiwa aiki alama a matsayin "mahimmanci", fitowar AI ba ta zama takarda ta ƙarshe ba. Akasari yana iya zama daftarin farko ko jerin abubuwan dubawa; Injiniya ƙwararren yana ɗaukar sa hannu da alhakin.

Ƙarshe-zuwa-ƙarshen kwarara: saka AI a cikin aikin ku

Halin da ake yi na warware matsala a cikin lantarki / sadarwa shine kamar haka, tare da AI yana shiga kowane mataki amma ba rufe ɗayansu da kanta ba:

1. Ƙayyade matsalar: Menene za'a tsara, menene burin aiki, menene ƙuntatawa (ikon, farashi, mita, doka) akwai? (AI: ya fayyace tambayar.)
2. Tattara bayanai/bukatun: Taswirar bayanai, daidaitattun iyaka, ma'auni. (AI: taƙaitawa kuma yana neman rashin daidaituwa.)
3. Ƙirƙirar hasashe/tsara: Madadin topologies, yanayin bayani. (AI: yana samar da lissafin tsari.)
4. Yi nazari: Lissafi, kwaikwaiyo, lamba, nazarin sigina. (AI: yana haifar da lambobi da almara.)
5. Tabbatar da: kwaikwaiyo, samfuri, ma'auni, daidaitaccen rubutu. (AI yana gefe; aunawa kuma mutane sun yanke shawara.)
6. Takaddun shaida kuma yanke shawara: ƙwararren injiniya yana ɗaukar alhakin. (AI: marubucin marubuci.)

A kiyaye wadannan matakai guda shida; sauran tsarin yana nuna muku yadda ake amfani da AI tare da kankare kayan lantarki / ayyukan sadarwa, mataki-mataki.

uku mini lokuta

Case 1 - Ƙimar ɓangaren da bai dace ba. Wani ɗan ƙwararren yana tambayar AI ƙimar da ake buƙata don ƙaddamar da ƙimar capacitor a cikin da'ira da aka ciyar daga mai sarrafa 3.3 V; AI ta ce "100 nF ya isa" kuma yana ba da takamaiman takardar sayan magani: "ƙara 10 µF ga kowane IC". Mai horarwa ya rubuta wannan kai tsaye a cikin zane. Ganin cewa madaidaicin ƙimar ya dogara da shawarwarin da ke cikin takaddar bayanan kwakwalwan kwamfuta da aka yi amfani da su, mitar sauyawa da bayanin martaba na yanzu. A cikin wannan ƙira, guntu mai sauri yana zana ɗaruruwan miliyoyin lokutan halin yanzu a cikin sakan daya, kuma 100 nF kadai bai isa ba; Katin yana yin kuskure a farawa. Hanyar da ta dace ita ce ɗaukar ƙimar daga kowane guntu na bayanan daftarin aiki da bincike na amincin ƙarfi.

Case 2 - Ƙimar ƙayyadaddun ƙayyadaddun da ba a wanzu ba. Injiniyan ya nemi AI don madaidaicin izinin EIRP (daidai da ikon radiated isotropic) na samfurin sa a cikin rukunin 2.4 GHz. AI ya ce "20 dBm bisa ga ka'ida" kuma ya sanya lambar abu. Injiniyan ya rubuta fayil ɗin fasaha daidai da haka. A cikin dakin gwaje-gwaje, an bayyana cewa ƙimar iyaka ta bambanta dangane da band, faɗin tashar da ƙasa, kuma lambar abin da aka bayar ba gaskiya bane. Hanyar da ta dace ita ce buɗe ƙa'idar ta yanzu (misali takaddun ETSI / FCC mai dacewa da kuma shawarar hukumar ƙasa) daga tushen hukuma kuma tabbatar da ƙimar da abu.

Case 3 - Daidaitaccen amfani. Wani injiniya yana zargin rashin daidaituwa a cikin ma'aunin layin 5000 daga firikwensin zafin jiki. Yana gaya wa AI don "jera layuka a cikin wannan tebur waɗanda za su iya samun rashin daidaituwa na raka'a (°C/°F rudani), ƙimar da ba za ta iya yiwuwa ba ta jiki, ƙaru, da tantanin halitta mara komai, kuma ku rubuta abin da kuke zargin kowannensu; KAR KA gyara, kawai alama." Tutar AI tana ba da karatun lambar 999 da yawa, karatun -300 °C biyu, da spikes waɗanda ke kama da sake saitin firikwensin. Injiniyan yana bincika kuma ya gyara waɗannan daga ɗanyen bayanan. Anan an yi amfani da AI daidai: jawo hankali, ɗan adam ya yanke shawara da gyara.

Samfuran gaggawar da za a iya kwafi

Kuna iya amfani da samfuran da ke ƙasa ta hanyar daidaita su zuwa kasuwancin ku. A cikin kowannensu, akwai umarni mai hankali don "ba da ainihin lamba, nuna tushen, bayyana rashin tabbas".

ROLE DA LIMIT TEMPLATE "Rawar: Kai gogaggen mataimaki ne injiniyan lantarki / sadarwa. Aiki: [rubuta batun]. Dokokin: Lokacin ba da ƙimar ƙima (na yanzu, impedance, mita, iyakar wutar lantarki), ce cewa dole ne a tabbatar da shi daga BAYANIN DATA ko MEASUREMENT na ɓangaren da aka zaɓa; idan ba ku da tabbacin, ba da ma'auni. sigar sa, amma dole ne a tabbatar da shi daga tushen hukuma A ƙarshen amsar ku, rubuta 'abin da ake buƙatar tabbatarwa ta hanyar simulation/a'auni'.

TEMPLATE GENERATION HYPOTHESIS"Samar da jeri na yau da kullun na abubuwan da za su iya haifar da alama mai zuwa: [bayyana alamar; abin da aka lura, menene ma'auni akwai]. Ga kowane hasashe: (1) wace shaida ta goyi bayansa, (2) menene ma'auni (oscilloscope, analyzer, multimeter) ya kamata a yi don tabbatarwa/karya. YI tabbataccen ganewar asali; ƙirƙirar taswirar hanyar warware matsala."

TABBATAR DA TABBATARWA "Jerin KOWANE ƙimar lambobi da kowane ma'auni / ƙa'ida a cikin ƙira / rahoton da ke ƙasa. Ga kowane, nuna daga wane tushe (bayani, rubutu daidaitaccen rubutu, ma'aunin dakin gwaje-gwaje, simulation) ya kamata a tabbatar da shi. Tattara waɗanda ba su da tabbas a ƙarƙashin 'ba za a iya amfani da su ba tare da tabbatarwa' ba. Rubutu: [manna]."

ANONYMIZATION Pre-Check TEMPLATE "Kafin bada wannan rubutu zuwa kayan aikin AI, sirrin ciniki na tuta ko sassa masu mahimmanci na sirri: abokin ciniki / sunan aikin, bayanin fayil na PCB, maɓallin firmware / kalmar wucewa, ainihin IP / MACaddresses, bayanan biyan kuɗi, ƙimar ɓoye. Ba da shawarar yadda zan iya ɓoye waɗannan. Rubutu: [manna]."

Rauni mai ƙarfi / Ƙarfi mai ƙarfi

Yin tambaya iri ɗaya hanyoyi biyu zai haifar da sakamako daban-daban.

RAUNI GASKIYA: "Mene ne ribar wannan eriya?"

KARFIN GASKIYAR: "Rawar: Kai mataimakin injiniyan RF ne. Bayyana ta waɗanne matakai dole ne a ƙayyade ribar eriya (nau'in zaɓi, girman, simintin filin, ma'aunin ɗakin anechoic) da abin da bayanai da abin da ake buƙata don tabbatarwa a kowane mataki. KADA KA ba da lambar samun KYAUTA; jaddada cewa za'a iya ƙayyade wannan kawai ta hanyar simulation da iyakataccen ma'auni, abin da za a iya kimanta naúrar EIR da ma'auni. "

Mai rauni mai sauri yana gayyatar AI don gyarawa; Ƙarfi mai ƙarfi yana kiyaye shi a cikin aikin da ya dace (bayanin tsari) kuma yana hana hallucination na dijital daga faruwa a farkon wuri.

Kuskuren gama gari

• "Tambaya" AI don ƙimar sassa / ƙira. Kayan aiki na yanzu, impedance, mitar yankewa yana zuwa ne kawai daga takaddar bayanai da aunawa.
• Yin amfani da ma'auni / doka ba tare da tabbatar da shi ba. Lambar abu, ƙimar tebur da sakin suna akai-akai ana hasashe.
• Ƙaddamar da yarda / yanke shawara na tsaro akan fitowar AI. Yarda da EMC, izinin watsa shirye-shirye, iyawar sabis aiki ne na ƙwararrun injiniyoyi da gwajin da aka amince da su.
• Manna bayanai masu mahimmanci a cikin kayan aikin gajimare ba tare da ɓoye suna ba. PCB/firmware, abokin ciniki topology da masu biyan kuɗi bayanan sirri ne na kasuwanci da bayanan sirri.
• Ana jiran tambaya daya da amsa daya. Yi amfani da AI akai-akai: daftarin aiki, zargi, kwaikwaya, auna, tacewa.

Tsanaki: Madaidaicin ƙima da kwarin gwiwa lamba ta bayyana a cikin fitowar AI, mafi girman buƙatar tabbatarwa. Daidaitawa ba garantin daidaito ba ne.

A takaice

A cikin wannan rukunin kun ga cewa AI mataimaki ne a cikin kayan lantarki / sadarwa ba mai yanke shawara ba. LLMs masu kimanta harshe ne da lambar; Mai ƙarfi a cikin fayyace ra'ayi, ƙayyadaddun lambar/firmware, tsaftace bayanai da tsara hasashen, amma mai haɗari a samar da madaidaicin ƙimar ɓangaren da ma'auni (hallucination). Rarraba ayyuka azaman ƙananan / matsakaici / babba / mahimmanci kuma daidaita zurfin tabbatarwa daidai. Tabbatar da kowane ƙimar lambobi daga takaddar bayanai ko ma'auni, da kowane madaidaicin tunani daga tushen hukuma. Yarda da yanke shawara mai mahimmanci na tsaro koyaushe suna kasancewa tare da ƙwararren injiniya kuma suna ɓoye ƙira / bayanan abokin ciniki masu mahimmanci kafin amfani.

Aikin aikace-aikace

Zaɓi ayyuka guda biyar daga aikinku (ko aikin lantarki mai ƙima): bayanin ra'ayi, tsaftacewa bayanai, jerin hasashe na gazawa, ƙayyadaddun abubuwan / ƙimar ƙima, da ƙa'ida/ma'auni tambaya. Sanya kowane a cikin tebur mai haɗari a sama. Sa'an nan fitar da ayyuka guda biyu ƙananan / matsakaici ga AI tare da samfurin "Rawar da iyaka" kuma yi alama ga kowane nau'i na lambobi / daidaitattun kayan fitarwa tare da samfurin "Jerin Tabbatarwa". Rubuta abubuwan da ba za a iya amfani da su ba tare da tabbatarwa ba.

jerin abubuwan dubawa

• [ ] Na sanya manufa a matakin haɗari (ƙananan / matsakaici / babba / mahimmanci) kafin amfani da shi.
• [ ] Na sami ƙimar ɓangaren / ƙira daga takaddar bayanai da auna maimakon tambayar AI.
• [ ] Na tabbatar da kowace ma'auni/ka'idoji daga tushen hukuma (daidaitaccen sigar).
• [ ] Na ɓoye bayanan sirri (PCB/firmware, abokin ciniki, IP/MAC, mai biyan kuɗi).
• [ ] Na bar matakin dacewa/mahimmancin tsaro ga ƙwararren injiniya da ƙwararren mai gwadawa.
• [ ] Na duba kowane ƙimar lamba a cikin fitowar AI don "ba za a iya amfani da shi ba tare da tabbatarwa ba".