{"id":9781,"date":"2023-07-12T15:04:02","date_gmt":"2023-07-12T13:04:02","guid":{"rendered":"https:\/\/onload.reinhausen.com\/de\/?p=9781"},"modified":"2023-10-17T15:22:35","modified_gmt":"2023-10-17T13:22:35","slug":"artificial-intelligence-transformers-sensors","status":"publish","type":"post","link":"https:\/\/onload.reinhausen.com\/en\/07-2023\/artificial-intelligence-transformers-sensors\/","title":{"rendered":"The AI Lends a Hand"},"content":{"rendered":"\n
\n\n\n\n

Ener\u00adgy net\u00adwork oper\u00ada\u00adtors and asset man\u00adagers already have to com\u00adpen\u00adsate for the grow\u00ading loss of per\u00adson\u00adnel knowl\u00adedge and the increas\u00ading main\u00adte\u00adnance require\u00adments of aging oper\u00adat\u00ading equip\u00adment. At the same time, chal\u00adlenges and bur\u00addens are increas\u00ading: <\/p>\n\n\n\n

The urgent expan\u00adsion of grids, the feed-in of renew\u00adable ener\u00adgies and the decen\u00adtral\u00adiza\u00adtion of ener\u00adgy sup\u00adply in the face of ever increas\u00ading ener\u00adgy demands hard\u00adly seem man\u00adage\u00adable. A key role is there\u00adfore being played by new tech\u00adnolo\u00adgies that help mon\u00adi\u00adtor and diag\u00adnose assets. Such intel\u00adli\u00adgent data-dri\u00adven sys\u00adtems make it pos\u00adsi\u00adble to use the exist\u00ading infra\u00adstruc\u00adture more effi\u00adcient\u00adly and extend its ser\u00advice life. The per\u00adfect inter\u00adac\u00adtion of all automa\u00adtion lev\u00adels is cru\u00adcial for this: from the sen\u00adsor to the data node on the trans\u00adformer to the lev\u00adel of glob\u00adal asset man\u00adage\u00adment.<\/p>\n\n\n\n

\u201cThis is no longer a dream of the future; it is already pos\u00adsi\u00adble today,\u201d explains Tobias Gru\u00adber, MR prod\u00aduct man\u00adag\u00ader who is involved in the devel\u00adop\u00adment of algo\u00adrithms and math\u00ade\u00admat\u00adi\u00adcal train\u00ading meth\u00adods. \u201cRein\u00adhausen already offers a large port\u00adfo\u00adlio of self-learn\u00ading sen\u00adsors and sys\u00adtems. And their capa\u00adbil\u00adi\u00adties are con\u00adstant\u00adly increas\u00ading through mutu\u00adal net\u00adwork\u00ading and an ever-grow\u00ading learn\u00ading curve.\u201d The dig\u00adi\u00adtal\u00adiza\u00adtion of assets is a nec\u00ades\u00adsary step for the rea\u00adsons men\u00adtioned. But this trans\u00adfor\u00adma\u00adtion process\u2009\u2014\u2009the inde\u00adpen\u00addent fur\u00adther devel\u00adop\u00adment of the sys\u00adtems them\u00adselves \u2013 will bring about changes in all indus\u00adtries, and also bring ben\u00ade\u00adfits that are not yet imag\u00adin\u00adable today. Since it is obvi\u00adous that AI is here to stay, it makes sense to use it to sup\u00adport the ener\u00adgy indus\u00adtry at all lev\u00adels.<\/p>\n\n\n\n

<\/span>
\n
\n

\u201cRein\u00adhausen already offers a large port\u00adfo\u00adlio of self-learn\u00ading sen\u00adsors and sys\u00adtems. And their capa\u00adbil\u00adi\u00adties are con\u00adstant\u00adly increas\u00ading through mutu\u00adal net\u00adwork\u00ading and an ever-grow\u00ading learn\u00ading curve.\u201d <\/p>\nTobias Gru\u00adber, Maschi\u00adnen\u00adfab\u00adrik Rein\u00adhausen<\/cite><\/blockquote>\n<\/div><\/div>\n\n\n\n

At the process lev\u00adel, var\u00adi\u00adous sen\u00adsors record numer\u00adous sig\u00adnals\u2009\u2014\u2009reli\u00adably and pre\u00adcise\u00adly\u2009\u2014\u2009direct\u00adly at the trans\u00adformer and for\u00adward them to a cen\u00adtral com\u00admu\u00adni\u00adca\u00adtion node at the field lev\u00adel for con\u00adsol\u00adi\u00adda\u00adtion. This pro\u00advides reli\u00adable infor\u00adma\u00adtion on the main\u00adte\u00adnance and health sta\u00adtus of the respec\u00adtive trans\u00adform\u00aders. These eval\u00adu\u00ada\u00adtions then come togeth\u00ader at the con\u00adtrol lev\u00adel where main\u00adte\u00adnance strate\u00adgies for the entire fleet can be devel\u00adoped. How\u00adev\u00ader, this only works if the sen\u00adsor tech\u00adnol\u00ado\u00adgy is mod\u00adu\u00adlar\u00adly expand\u00adable and man\u00adu\u00adfac\u00adtur\u00ader-inde\u00adpen\u00addent so that it can react agile\u00adly to future needs in fleet main\u00adte\u00adnance and renew\u00adal. And with such a high lev\u00adel of data exchange between the lev\u00adels, cyber\u00adse\u00adcu\u00adri\u00adty is of par\u00adtic\u00adu\u00adlar impor\u00adtance. <\/p>\n\n\n\n

\n
There is more to the term cybersecurity than just protection against hacker attacks.<\/h5>\n\n\n\n
<\/i> Security expert Dr. Hubert Feyrer explains exactly what this means in an interview.<\/a><\/h6>\n<\/div><\/div>\n\n\n\n
<\/div>\n\n\n\n
<\/figure>\n\n\n\n
\n
\n
The intel\u00adli\u00adgent MSENSE\u00ae DGA sen\u00adsor con\u00adtin\u00adu\u00adal\u00adly learns dur\u00ading its entire life\u00adtime.<\/figcaption><\/figure>\n<\/div><\/div>\n\n\n\n
\n
With its vibroa\u00adcoustic mea\u00adsure\u00adments, the MSENSE\u00ae VAM\u00ae keeps sight of the vibra\u00adtions of the switch\u00ading process.<\/figcaption><\/figure>\n<\/div><\/div>\n<\/div>\n\n\n\n

1. INTELLIGENT SENSOR TECHNOLOGY<\/h2>\n\n\n\n

Sen\u00adsors col\u00adlect infor\u00adma\u00adtion \u2013 that much is obvi\u00adous. Smart sen\u00adsors can also eval\u00adu\u00adate it. But what does that mean in con\u00adcrete terms? Which data can be bet\u00adter eval\u00adu\u00adat\u00aded using machine learn\u00ading than with pre\u00advi\u00adous meth\u00adods? An impor\u00adtant stan\u00addard mea\u00adsure, for exam\u00adple, is the analy\u00adsis of the insu\u00adlat\u00ading oil in the gas phase (DGA \u2013 dis\u00adsolved gas analy\u00adsis). In all com\u00admon analy\u00adsis meth\u00adods, the gas is extract\u00aded, which expos\u00ades it each time to fluc\u00adtu\u00adat\u00ading exter\u00adnal fac\u00adtors such as ambi\u00adent tem\u00adper\u00ada\u00adture, air pres\u00adsure, humid\u00adi\u00adty and much more. Ensur\u00ading that these val\u00adues always remain con\u00adstant for mea\u00adsure\u00adment would con\u00adsume vast sums of mon\u00adey and sig\u00adnif\u00adi\u00adcant\u00adly increase the com\u00adplex\u00adi\u00adty of DGA sys\u00adtems.<\/p>\n\n\n\n

This is where arti\u00adfi\u00adcial intel\u00adli\u00adgence comes into play. Using train\u00ading data sets, the algo\u00adrithm in the smart DGA sys\u00adtem learns the rela\u00adtion\u00adships between the gas con\u00adcen\u00adtra\u00adtion in the insu\u00adlat\u00ading oil, the sen\u00adsor sig\u00adnal and the pos\u00adsi\u00adble dis\u00adturb\u00ading influ\u00adences. With the aid of math\u00ade\u00admat\u00adi\u00adcal-sta\u00adtis\u00adti\u00adcal meth\u00adods from the machine learn\u00ading tool\u00adbox, the DGA sen\u00adsor is thus \u201ctrained\u201d dur\u00ading its devel\u00adop\u00adment phase. The new intel\u00adli\u00adgent spe\u00adcial\u00adist is then installed on the trans\u00adformer and imme\u00addi\u00adate\u00adly knows what to do. \u201cWith the lab\u00ado\u00adra\u00adto\u00adry analy\u00adsis of the oil as a ref\u00ader\u00adence point, the DGA sen\u00adsor then also learns about its trans\u00adformer and con\u00adtin\u00adu\u00adous\u00adly recal\u00adi\u00adbrates itself,\u201d says Gru\u00adber. \u201cIn this way, effects such as sen\u00adsor drift, aging of the insu\u00adlat\u00ading oil and the like can be com\u00adpen\u00adsat\u00aded for, and con\u00adsis\u00adtent mea\u00adsure\u00adment repeata\u00adbil\u00adi\u00adty ensured.\u201d<\/p>\n\n\n\n

Gas analy\u00adsis for on-load tap-chang\u00aders looks a bit more dif\u00adfi\u00adcult, because in con\u00adtrast to the trans\u00adformer, few\u00ader data sets on gas for\u00adma\u00adtion rates and typ\u00adi\u00adcal gas con\u00adcen\u00adtra\u00adtions have been devel\u00adoped. \u201cAna\u00adlyz\u00ading on-load tap-chang\u00aders requires a lot of expert knowl\u00adedge about how the par\u00adtic\u00adu\u00adlar tap-chang\u00ader type works and how it oper\u00adates,\u201d Gru\u00adber tells us. That\u2019s why it doesn\u2019t just come down to ana\u00adlyz\u00ading the key gas\u00ades. Rather, in this regard, sta\u00adtis\u00adti\u00adcal-math\u00ade\u00admat\u00adi\u00adcal algo\u00adrithms use a much broad\u00ader data set. In addi\u00adtion to gas con\u00adcen\u00adtra\u00adtions, infor\u00adma\u00adtion about the on-load tap-chang\u00ader such as the num\u00adber of switch\u00ading oper\u00ada\u00adtions, the oil quan\u00adti\u00adty, or the time since the last oil change are also used as input vari\u00adables. The result is a diag\u00adno\u00adsis with an indi\u00adca\u00adtion of prob\u00ada\u00adbil\u00adi\u00adties, which is far ahead of pure gas analy\u00adsis.<\/p>\n\n\n\n

\n

\u201cAna\u00adlyz\u00ading on-load tap-chang\u00aders requires a lot of expert knowl\u00adedge about how the par\u00adtic\u00adu\u00adlar tap-chang\u00ader type works and how it oper\u00adates.\u201d <\/p>\nTobias Gru\u00adber, Maschi\u00adnen\u00adfab\u00adrik Rein\u00adhausen<\/cite><\/blockquote>\n\n\n\n

For a more com\u00adpre\u00adhen\u00adsive assess\u00adment of on-load tap-chang\u00aders, MSENSE\u00ae VAM\u00ae is the online diag\u00adnos\u00adtic tool that per\u00adforms a vibro-acoustic mea\u00adsure\u00adment by detect\u00ading vibra\u00adtions that occur dur\u00ading the switch\u00ading process of an on-load tap-chang\u00ader. The asso\u00adci\u00adat\u00aded algo\u00adrithm rep\u00adre\u00adsents the switch\u00ading oper\u00ada\u00adtion in a lim\u00adit val\u00adue curve, which is cal\u00adcu\u00adlat\u00aded more pre\u00adcise\u00adly with each switch\u00ading oper\u00ada\u00adtion. Thus, the sys\u00adtem iter\u00ada\u00adtive\u00adly learns dur\u00ading the switch\u00ading oper\u00ada\u00adtions what the acoustic sig\u00adna\u00adture of a cor\u00adrect\u00adly oper\u00adat\u00ading on-load tap-chang\u00ader looks like. \u201cAnd it works every\u00adwhere,\u201d Gru\u00adber explains, \u201cfor new as well as for exist\u00ading on-load tap-chang\u00aders of all man\u00adu\u00adfac\u00adtur\u00aders and designs.\u201d<\/p>\n\n\n\n

One thing that no one wants to find in their trans\u00adformer is mois\u00adture. In addi\u00adtion to many pos\u00adsi\u00adble test points, it is above all the break\u00addown volt\u00adage of the insu\u00adlat\u00ading oil that gives con\u00adcrete indi\u00adca\u00adtions of pos\u00adsi\u00adble con\u00adt\u00ada\u00adm\u00adi\u00adna\u00adtion. How\u00adev\u00ader, the meth\u00adods pro\u00adposed by cur\u00adrent stan\u00addards for ana\u00adlyz\u00ading this volt\u00adage are cum\u00adber\u00adsome and expen\u00adsive. Here, too, online mon\u00adi\u00adtor\u00ading by self-learn\u00ading algo\u00adrithms is bet\u00adter: They learn the rela\u00adtion\u00adship between rel\u00ada\u00adtive oil mois\u00adture, oil tem\u00adper\u00ada\u00adture and break\u00addown volt\u00adage. In this process, data for break\u00addown volt\u00adage lev\u00adels at dif\u00adfer\u00adent oil tem\u00adper\u00ada\u00adtures and humidi\u00adties are deter\u00admined exper\u00adi\u00admen\u00adtal\u00adly using a ref\u00ader\u00adence method, e.g. IEC 60156, and these data are then used to train a math\u00ade\u00admat\u00adi\u00adcal mod\u00adel. The mod\u00adel is val\u00adi\u00addat\u00aded and opti\u00admized using test data that is inde\u00adpen\u00addent of the train\u00ading data and again, a trained arti\u00adfi\u00adcial expert is on hand to reli\u00adably iden\u00adti\u00adfy where prob\u00adlems might be brew\u00ading.<\/p>\n\n\n\n

<\/figure>\n\n\n\n
<\/div>\n\n\n\n
\n
\n
All data flows togeth\u00ader via the Sen\u00adsor\u00adBus\u00ae in the cen\u00adtral com\u00adput\u00ading unit ISM\u00ae, where they are eval\u00adu\u00adat\u00aded and assessed by algo\u00adrithms. The entire know-how about trans\u00adform\u00aders that Rein\u00adhausen has built up over decades is bun\u00addled in this unit. They and the sen\u00adsors are installed in the ETOS\u00ae con\u00adtrol cab\u00adi\u00adnet and eval\u00adu\u00adate all sen\u00adsor data across all man\u00adu\u00adfac\u00adtur\u00aders.<\/figcaption><\/figure>\n<\/div><\/div>\n\n\n\n
\n
<\/figure>\n<\/div><\/div>\n<\/div>\n\n\n\n

2. <\/strong>INTELLIGENT DATA EVALUATION<\/h2>\n\n\n\n

Mon\u00adi\u00adtor\u00ading sys\u00adtems on pow\u00ader trans\u00adform\u00aders already play a major role, espe\u00adcial\u00adly in terms of sta\u00adble fault detec\u00adtion. How\u00adev\u00ader, this alone is not enough. They con\u00adtin\u00adu\u00adous\u00adly report infor\u00adma\u00adtion and data which must be eval\u00adu\u00adat\u00aded and inter\u00adpret\u00aded in rela\u00adtion to each oth\u00ader \u2013 and this for every trans\u00adformer. This is no easy task, espe\u00adcial\u00adly when sev\u00ader\u00adal sen\u00adsors are capa\u00adble of the same func\u00adtion. \u201cFor exam\u00adple, if tem\u00adper\u00ada\u00adture, par\u00adtial dis\u00adcharge and DGA sen\u00adsors are all capa\u00adble of detect\u00ading wind\u00ading faults, their state\u00adments are gen\u00ader\u00adal\u00adly not com\u00adpa\u00adra\u00adble which can result in con\u00adtra\u00addic\u00adto\u00adry state\u00adments, and a sim\u00adple diag\u00adno\u00adsis is not pos\u00adsi\u00adble.\u201d <\/p>\n\n\n\n

That\u2019s why at MR all sen\u00adsors are inte\u00adgrat\u00aded into the ETOS\u00ae and eval\u00adu\u00adat\u00aded by the self-learn\u00ading algo\u00adrithms of the ISM\u00ae com\u00adput\u00ading unit with\u00adin the ETOS\u00ae. \u201cA Bayesian net\u00adwork checks which fault pat\u00adterns best match the warn\u00adings that occur \u2013 and don\u2019t occur. In addi\u00adtion, a pri\u00adori prob\u00ada\u00adbil\u00adi\u00adties for typ\u00adi\u00adcal trans\u00adformer faults and sen\u00adsor reli\u00ada\u00adbil\u00adi\u00adty are con\u00adsid\u00adered.\u201d The result is a prob\u00ada\u00adbil\u00adi\u00adty esti\u00admate for all known trans\u00adformer prob\u00adlems, with the most like\u00adly prob\u00adlems dis\u00adplayed to the cus\u00adtomer along with a list of rea\u00adsons for the find\u00adings. This allows the pro\u00adfes\u00adsion\u00adal to inter\u00adpret the diag\u00adnos\u00adtic results in a sim\u00adpli\u00adfied and effi\u00adcient man\u00adner. \u201cThe Bayesian net\u00adwork Asset Intel\u00adli\u00adgence for pow\u00ader trans\u00adform\u00aders pro\u00advides a guide to min\u00adi\u00admize risk and take cor\u00adrec\u00adtive action quick\u00adly,\u201d says Tobias Gru\u00adber.<\/p>\n\n\n\n

<\/div>\n\n\n\n
<\/figure>\n\n\n\n
\n
TESSA\u00ae com\u00adbines all data from the offline and online worlds to enable con\u00adsis\u00adtent visu\u00adal\u00adiza\u00adtion of the exact sta\u00adtus of each asset down to mod\u00adule lev\u00adel at any time.<\/figcaption><\/figure>\n<\/div><\/div><\/div>\n\n\n\n

3. INTELLIGENT FLEET MANAGEMENT <\/h2>\n\n\n\n

Most trans\u00adformer fleets cur\u00adrent\u00adly con\u00adsist of mul\u00adti\u00adple prod\u00aducts that come from var\u00adi\u00adous sup\u00adpli\u00aders and dif\u00adfer in age and equip\u00adment. Asset man\u00adage\u00adment must do jus\u00adtice to them all. The best way to do this is with a cen\u00adtral solu\u00adtion that uses and eval\u00adu\u00adates all online and offline data \u2013 from sen\u00adsor data to man\u00adu\u00adal\u00adly entered mea\u00adsured val\u00adues fol\u00adlow\u00ading an on-site inspec\u00adtion. Rein\u00adhausen has devel\u00adoped TESSA\u00ae APM\u00ae for this pur\u00adpose. The soft\u00adware behind the plat\u00adform eval\u00adu\u00adates data, rec\u00adog\u00adnizes trends and inde\u00adpen\u00addent\u00adly issues rec\u00adom\u00admen\u00adda\u00adtions for action. In the process, the intel\u00adli\u00adgent soft\u00adware con\u00adtin\u00adu\u00adous\u00adly learns dur\u00ading oper\u00ada\u00adtion and \u201cgets to know\u201d its con\u00adnect\u00aded assets. Because at some point the rel\u00ade\u00advant spe\u00adcial\u00adist may be miss\u00ading and it will be TESSA\u00ae APM\u00ae that says, for exam\u00adple, \u201cTake a look at trans\u00adformer 5A, it\u2019s a bit sen\u00adsi\u00adtive with the oil\u201d. <\/p>\n\n\n\n

The fact that the soft\u00adware com\u00adbines all data, can be used remote\u00adly and warns pro\u00adphy\u00adlac\u00adti\u00adcal\u00adly in the event of mal\u00adfunc\u00adtions, saves asset man\u00adagers valu\u00adable time and also fills the per\u00adson\u00adnel sup\u00adply gap. A tri\u00adad of intel\u00adli\u00adgent sen\u00adsors at the process lev\u00adel, data eval\u00adu\u00ada\u00adtion at the field lev\u00adel on the trans\u00adformer, and fleet man\u00adage\u00adment at the con\u00adtrol lev\u00adel thus pre\u00adpare net\u00adwork oper\u00ada\u00adtors and OEMs for the chal\u00adlenges of today and tomor\u00adrow.\u2002<\/p>\n\n\n\n

<\/div>\n\n\n\n
\n
\n\n\n
\n
<\/figure>\n<\/div>\n\n\n

YOUR CONTACT<\/h2>\n\n\n\n

Do you have ques\u00adtions about MR\u2019s AI solu\u00adtions?<\/strong>
Tobias Gru\u00adber is there for you:
<\/i> \u2009T.Gruber@reinhausen.com<\/a><\/p>\n<\/div><\/div>\n","protected":false},"excerpt":{"rendered":"

Sys\u00adtems and sen\u00adsors that use AI-sup\u00adport\u00aded soft\u00adware increase the ser\u00advice life and oper\u00ada\u00adtional reli\u00ada\u00adbil\u00adi\u00adty of net\u00adworked assets, but this is not pos\u00adsi\u00adble with\u00adout asso\u00adci\u00adat\u00aded cyber\u00adse\u00adcu\u00adri\u00adty. Ide\u00adal\u00adly, holis\u00adtic sys\u00adtems are placed at three cen\u00adtral lev\u00adels. And this is already pos\u00adsi\u00adble today.<\/p>\n","protected":false},"author":1,"featured_media":10450,"comment_status":"closed","ping_status":"open","sticky":false,"template":"template-cover.php","format":"standard","meta":{"_acf_changed":false,"wp_typography_post_enhancements_disabled":false,"footnotes":""},"categories":[38,1],"tags":[],"class_list":["post-9781","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-07-2023","category-onload"],"acf":[],"_links":{"self":[{"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/posts\/9781","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/comments?post=9781"}],"version-history":[{"count":6,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/posts\/9781\/revisions"}],"predecessor-version":[{"id":10875,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/posts\/9781\/revisions\/10875"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/media\/10450"}],"wp:attachment":[{"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/media?parent=9781"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/categories?post=9781"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/onload.reinhausen.com\/en\/wp-json\/wp\/v2\/tags?post=9781"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}