Predictive Maintenance 4 - PwC
June 2017Predictive Maintenance 4.0Predict the unpredictablePdM 4.0
ContentsContentsForeword. 3Summary. 4Chapter 1: Introduction. 6Chapter 2: Key findings. 10Case Infrabel. 12Case Sitech. 18Chapter 3: Recommendations . 20Chapter 4: Call to action. 24About the survey. 26Contacts. 28Acknowledgements. 302 PdM 4 Predict the unpredictable
ForewordForewordPwC and Mainnovation have joined forces in the field of maintenance and asset management.We are both convinced that maintenance can be brought to a new level by combining the powerof new digital technologies with a deep understanding of maintenance. We believe predictivemaintenance with big data analytics can be a tremendous source of new value for asset ownersand maintenance service providers.To deepen our understanding and sharpen our insights, we have jointly carried out a marketsurvey on predictive maintenance. This involved surveying 280 companies from Belgium,Germany and the Netherlands about their current use of, and future plans for, predictivemaintenance, and conducting interviews with leading companies in the field.This report presents the results of this research and our approach to successfully implementingpredictive maintenance with big data. Our findings should be of interest to those responsible forthe maintenance and asset management of fleets, factories and infrastructure, who are lookingfor new ways to increase the reliability of their assets.We are proud to share these findings with you and look forward to fruitful discussions with youon this topic.Michel MuldersPartner at PwC NetherlandsMark HaarmanManaging Partner at MainnovationPdM 4 Predict the unpredictable 3
SummarySummaryPredictive maintenance is surely one of the most talked-about topics in maintenance and assetmanagement. In order to find out where companies currently stand regarding predictive maintenance,and where they plan to be in the near future, we surveyed 280 companies in Belgium, Germany andthe Netherlands.In order to assess current practices, we have used aframework that identifies four levels of maturity inpredictive maintenance. As companies move throughthese levels, there is an increase in how much data theyuse to predict failures. Visual inspections representlevel 1 in this framework; instrument inspectionsand real-time condition monitoring are associatedwith levels 2 and 3. At level 4 big data analytics starts4 PdM 4 Predict the unpredictableto drive decision-making. This is where the digitalrevolution meets maintenance. This level involvesapplying the power of machine learning techniques toidentify meaningful patterns in vast amounts of dataand generate new, actionable insights for improvingasset availability. We call this Predictive Maintenance4.0, or PdM 4.0. PdM 4.0 offers you the potential topredict failures that had been unpredictable up to now.
SummaryKey findings from the surveyWe found that two thirds of survey respondents arestill at maturity levels 1 or 2. Only 11% have alreadyachieved level 4. The resources, capabilities and toolsrespondents use match their maturity levels: skilledtechnicians, standard software tools and maintenancelogs play a dominant role in their current predictivemaintenance processes. Only a few companies alreadyemploy the people and tools needed for PdM 4.0:reliability engineers and data scientists, statisticalsoftware packages and external data sources.We also found that respondents are quite ambitiousabout improving their predictive maintenancematurity. Around half said they have plans to use PdM4.0 at some point in the future. Taking into accountrespondents who are already working on PdM 4.0 andthose who plan to do so within the next five years,around one in three companies will be using PdM4.0 in some form within five years, provided they cansuccessfully implement it. We conclude that PdM 4.0is widely recognized as a potential improvement overcurrent maintenance practices, but that the market isstill in the very early stages of adopting this technology.Uptime improvement is the main reason whyrespondents have plans for PdM 4.0. Other importantreasons relate to other traditional value drivers inmaintenance and asset management such as costreductions, lifetime extension for aging assets and thereduction of safety, health, environment and qualityrisks. Respondents also identified a number of criticalsuccess factors for PdM 4.0 implementation. Theavailability of data was mentioned most often as acritical success factor, followed by technology, budgetand culture. We conclude that, at this early stage inthe PdM 4.0 lifecycle, companies still see considerabletechnical obstacles to its implementation. However,they recognize that PdM 4.0 implementation is not apurely technical challenge.Our approach to successful PdM 4.0implementationThe second half of this report highlights our approachfor implementing PdM 4.0, which considers technicalas well as organisational aspects. We have provideda framework for the step-by-step implementationof technical components in the PdM 4.0 model, in amanner that supports business strategy. Our approachalso covers the technical infrastructure - data analyticsplatform, IoT infrastructure - needed to sustainPdM 4.0. Organisational aspects are also importantif PdM 4.0 is to be successful. We have focused ontwo such aspects: building skills and capabilitiesneeded for PdM 4.0, and building a digital culture. Itis not enough to simply attract and develop talent inreliability engineering and data science. Companiesmust also create circumstances in which these peoplecan flourish, and challenge and complement each otherto generate valuable and actionable new insights forimproving maintenance and asset management.Digital culture is the final aspect to be addressed inour approach. In other words, a culture that embracesnew, cross-functional ways of working, which allowcompanies to capitalize on the power of digitaltechnologies. A culture where everyone from theboardroom to the shop floor understands the powerof data analytics. Companies with a robust digitalculture possess the confidence and ambition to becomeincreasingly data-driven in their decision-making.PdM 4 Predict the unpredictable 5
Chapter 1 Introduction The next level in predictive maintenanceChapter 1 IntroductionThe next level in predictive maintenancePredictive maintenance is a bit of hype these days. It is being proclaimed as the ‘killer app’ for the Internet ofThings. Machine learning and predictive analytics - the main technologies that enable predictive maintenance- are nearing the ‘Peak of Inflated Expectations’ in Gartner’s Hype Cycle. At the same time, Google Trend datareveals increased interest in the subject, as do articles that have started to appear in the mainstream and businesspress.“predictive maintenance”Monthly queries(reative, (relative,max 100) max 100)PREDICTIVE -08mei-082008nov-08mei-08aug-08Number of maonthy Google querles Google trend data2017A historical frameworkA historical perspective may help clear up some of thehaze that surrounds predictive maintenance. Althoughit may be a bit of a hype, it is not an entirely newconcept. Without really using the term, people havebeen doing predictive maintenance for many years.Over time, different levels of maturity have evolved.When a technician performs a visual inspectionand selects - based on his knowledge, experienceand intuition - the best time to shut down a piece ofequipment so repairs can be carried out, he is in factperforming predictive maintenance.The next level of maturity involves augmentingthe inspector’s expertise with periodic instrumentinspections that provide more specific and objectiveinformation about the condition of the asset inquestion. The next step in sophistication involvesusing real-time condition monitoring, where sensorscontinuously collect data about the state of an assetand send alerts based on pre-established rules or whencritical levels are exceeded.One thing that has changed over the years isthe amount of data that goes into making thesepredictions. The enhanced use of data correspondswith increasing levels of maturity, and these areaccompanied by improvements in maintenanceperformance. By collecting more and more data,maintenance staff are able to make better informeddecisions that lead to increased reliability, higher uptime, fewer accidents and failures, and lower costs.Condition mPredicIntelligSenEncrypted computer pEquipmAn6 PdM 4 Predict the unpredictable
Chapter 1 Introduction The next level in predictive maintenanceThe next step: big data analyticsWirelessmonitoringThe current buzz about predictive maintenance stemsfrom new opportunities to capitalize on the digitalrevolution, and more specifically on advances indecision support tools powered by big data analytics.like spreadsheets or relational databases, but can alsobe unstructured, like maintenance logs or thermalimages which can be ‘unlocked’ through text miningand pattern recognition software respectively.In our increasingly digitized world, where virtuallyevery activity creates a digital trace, there has beenexponential growth in how much data can be usedfor predictive maintenance. Data sets can be obtainedfrom both internal and external sources. Consider,for example, the vast pools of sensor data that can becollected from entire factories, transportation fleets orinfrastructure networks and distributed via Internet ofThings technology. In terms of external data, considerenvironmental data about temperature, humidity andwind speeds, or operator profiles or specifications ofmaterials being processed at the time of failure. Datasets used for predictive maintenance may be structured,One could easily drown in this sea of data. Fortunately,rapid advances in artificial intelligence techniqueshave enabled us to make sense of all this data. Machinelearning algorithms are particularly crucial in thisrespect (see text box ‘Machine beats human: the powerof self-learning machines). These algorithms are notconstructed as a predefined set of rules, as in traditionalsoftware programming. Instead, these algorithms areself-learning. They infer rules by performing a series oftrials on a set of training data and thus construct theirown model of the world. Every subsequent amount ofdata is then used to refine that model and improve itspredictive powers.AnalyticsIndustry 4.0BigInfrastructureMaintenanceDataData driven decisions makingctive maintenancegent machinesAvailability PrognosticsnsorsCloudFailureAssetsDataTechnology drivenprocesses PredictmentComputerization failuresnalysisAdvanced analyticsData collectionTransmitters Costs autonomouslyPdM 4 Predict the unpredictable 7
Chapter 1 Introduction The next level in predictive maintenanceMachine beats human: the power of self-learning machinesIf you are somewhat sceptical about what artificial intelligence (AI) canachieve in your field of expertise, you are in good company. When we werewriting this report, Google’s AlphaGo had just defeated Ke Jie, the world’sbest player of the ancient Chinese board game Go. Ke Jie, who had boastedprior to the match he would never be beaten by a computer, lost 3-0 despiteplaying almost perfectly.So why is this considered a groundbreaking achievement for AI? Essentiallybecause the incomputably large number of options in a Go game make itimpossible to ‘calculate’ your advantage a few moves ahead. Unlike chessgrandmasters, top Go players do not ‘calculate’ their next move. Instead, theyrely on experience, intuition and the ability to learn.This difference matters when trying to teach computers how to play suchgames. In chess, you can use rule-based programming where humanknowledge is coded into a set of instructions. This approach doesn’t get youvery far with Go. Professional Go players rely on so-called tacit knowledge:they know more then they can tell. The same is true for ordinary humanswho drive a car through traffic, instantly recognize a face or who can tell if apicture contains a cat or not.Recent breakthroughs in AI have occurred in tasks like these, where we can’texactly explain the steps followed to carry them out. This can be attributedto rapid advances made in a particular field of AI called machine learning.It works a bit like this: a machine learning algorithm is presented with atraining set that has been classified (e.g. pictures labelled ‘cat’ or ‘no cat’) andis, after a large number of iterations, able to figure out what features to lookfor and how to weigh their importance in order to come up with the correctanswer, either ‘cat’ or ‘no cat’.How is this relevant to predictive maintenance with big data? We can presentself-learning algorithms with historical maintenance data and a failurehistory, and let the algorithm detect patterns and signals in the data thatcorrelate with failure. If it detects such patterns in the future, the algorithmwill predict an increased likelihood of failure and will give an early warning.When self-driving trucks can deliver a cargo of beer, when computersoutperform humans in speech recognition and when self-learning patternrecognition algorithms can detect malignant cells that pathologists overlook,it is time to investigate whether this technology can also be applied to predictfailures that had been unpredictable up to now.8 PdM 4 Predict the unpredictable
Chapter 1 Introduction The next level in predictive maintenancePredictive Maintenance 4.0The application of big data analytics in maintenancerepresents the fourth level of maturity in predictivemaintenance, as shown in the PdM maturity growthmodel below. We call this fourth level PredictiveMaintenance 4.0, which is abbreviated as PdM 4.0.PdM 4.0 is about predicting future failures in assets andultimately prescribing the most effective preventivemeasure by applying advanced analytic techniques onbig data about technical condition, usage, environment,maintenance history, similar equipment elsewhereand in fact anything that may correlate with theperformance of an asset.Level 1 Visual inspections: periodic physicalinspections; conclusions are based solely on inspector’sexpertise.Level 2 Instrument inspections: periodic inspections;conclusions are based on a combination of inspector’sexpertise and instrument read-outs.Level 3 Real-time condition monitoring: continuousreal-time monitoring of assets, with alerts given basedon pre-established rules or critical levels.Level 4 PdM 4.0: continuous real-time monitoring ofassets, with alerts sent based on predictive techniques,such as regression analysis.Al intenance organisationhas been fused into larger units in order toreap synergy-related benefits. At Infrabelheadquarters in Brussels, a central DataCell has been created where increasingvolumes of data generated by thesetools are collected and analysed. A widerange of home-made IT applications formaintenance is being replaced by a singletool where data from various systems isintegrated and standardized. A number ofpilot projects to test predictive analyticsin maintenance have been started, andInfrabel is currently recruiting datascientists to take its maintenance operationsto the next level.On the eve of a new era inmaintenanceBy making these preparations, Infrabelhas put itself in an excellent positionfor the large scale application of dataanalytics in maintenance. Even though thisimplementation could face a few regulatoryhurdles - stemming from strict safetyrequirements and current regulationsthat prescribe a minimum number ofvisual inspections per year, Infrabel is stillexpected to make progress in this area.That would be a major step along theway of what Infrabel, describes as “acomplete transformation of Infrabel intoa digital enterprise in which ‘basic’ assetsare replaced by smart assets that areintegrated in an Internet of Things. Thistransformation enables Infrabel to becomeincreasingly data-driven in its decisionmaking.”PdM 4 Predict the unpredictable 13
Chapter 2 Key Findings Towards PdM 4.0: ambitions and capabilitiesFuture plans for PdM 4.0Now that we know where the market currently standson predictive maintenance, we want to know whatcompanies’ future plans are for PdM 4.0.Almost half of the respondents (132 out of 280) haveplans to eventually implement PdM 4.0, and one in five(54/280) have already started to implement it.20%Yes,currentlyworking on it6%Yes,we startnext year6%Yes,we startwithin 3 years17%Yes,no startdate51%NoOnly 58 out of 132 respondents currently workingon, or with plans to work on, PdM 4.0 were able and willing - to indicate a budget for their futureinvestments in PdM 4.0. A similar number ofrespondents said they “have no idea”, or that no specificbudget would be set aside for implementing PdM 4.0.Discussion: Ample ambition to advance to PdM 4.0Assuming it takes two years to implement PdM 4.0, and that all such implementation projects will be successful,almost one in three companies will be using PdM 4.0 five years from now. This would represent a major increasefrom the 11% currently at maturity level 4, and also leave significant potential for further implementation.Why do companies want to adopt PdM 4.0?Knowing that almost one in three companies have ambitions to adopt PdM 4.0 in thecoming years, it’s worthwhile taking a closer look at what drives companies to implementPdM 4.0.Respondents expect PdM 4.0 to contribute to further improvements in all ‘traditional’value drivers in maintenance and asset management. Uptime improvement is clearly themost important in this regard, with almost half of the companies in our survey identifyingit as their primary goal for implementing PdM 4.0.Primary goal for adoption of PdM 4.0Higher customer satisfaction 8%New revenue stream 1%16% Lifetimeextension of aging asset47% Uptime improvement17% Cost reduction11% Reductionof safety, health,environment & quality risks14 PdM 4 Predict the unpredictable
Chapter 2 Key Findings Towards PdM 4.0: ambitions and capabilitiesDiscussion: Secondary motives for implementing PdM 4.0Low scores for primary goals like ‘New revenue stream’, ‘Higher customer satisfaction’ and ‘Better product design’can be attributed to the relatively small share of respondents that provide maintenance services to externalcustomers. If we zoom in on the survey results, we see that maintenance service providers think these goals areequally important as the traditional value drivers in maintenance.‘Dealing with employee turnover from an aging workforce’ was never mentioned as the primary goal forimplementing PdM 4.0. However, it is worth noting that both Infrabel and Sitech, companies that we portray asPdM 4.0 front-runners in this report, mention PdM 4.0 as a possible substitute for employees they expect to retirein the coming years.How are companies placed to make the step towards PdM 4.0?Successful implementation of PdM 4.0 requires capabilities in several key domains to be on par with theconcerned level of maturity. (The next Chapter provides a more in-depth recommendation about how toapproach such an implementation). The capability matrix below shows how mature a company’s Processes,Content, Performance measurement, IT and Organisation need to be if it wants fully exploit the potentialof PdM 4.0.PdM Maturity StageCapability1. Visual Inspections2. Instument InspectionsProcesses-p eriodic inspection(physical)- checklist- paper recording- periodic inspection (physical) - continouos inspection (remote) - c ontinuous inspection- instruments- s ensors(remote)- digital recording-d igital recording- s ensors and other data-d igital recordingContentPerformanceMeasurementIT3. R eal TimeConditions Monitoring- paper based condition - digital condition datadata- single inspection points- multiple inspectionpoints- digital condition data- multiple inspection points- visual norm verification- paper based trendanalyses- prediction by expertopinion- MS Excel/MS Access- automatic norm verification- digital trend analyses- monitoring by CM software-a utomatic norm verification- digital trend analyses- prediction by expert opinion-e mbedded instrumentsoftwareOrganisation - experienced craftsmen - trained inspectors4. PdM4.0- digital condition data- multiple inspection points- digital environment data- digital maintenance history-a utomatic norm verification- digital trend analyses-p rediction by statisticalsoftware- advanced decision support- condition monitoring software - condition monitoring softwa
Predictive maintenance is a bit of hype these days. It is being proclaimed as the ‘killer app’ for the Internet of Things. Machine learning and predictive analytics - the main technologies that enable predictive maintenance - are nearing the ‘Peak of Inflated Expectations’ in Gartner’s Hype Cycle. At the same time, Google Trend data
PWC Driving Licence In NSW it is compulsory for every person driving a PWC to hold a current PWC driving licence. There are two types of PWC driving licence: 1. PWC driving licence for those aged 16 years and over. 2. Young Adult PWC driving licence for people aged from 12 to less than 16 years. A Young Adult PWC driving licence
Predictive Maintenance Predictive maintenance lets you estimate time-to-failure of a machine. Knowing the predicted failure time helps you find the optimum time to schedule maintenance for your equipment. Predictive maintenance not only predicts a fu-ture failure, but also pinpoints problems in your complex
MAINTENANCE STAKEHOLDERS UNIFIED PREDICTIVE MAINTENANCE CONCEPT Definition of UPTIME unified predictive maintenance concept ISO 13374 as implemented by MIMOSA OSA-CBM, RAMI4.0 for compliance with Industry 4.0 standards Phases of predictive maintenance and proactive computing Phases of industrial analytics maturity
On May 12, at approximately 2:30 pm, two personal watercraft (PWC) were operating in Biscayne Bay. The PWC were jumping the wakes of other vessels in the area. PWC #1 jumped the wake of a vessel and . Boating Accidents Statistical Report PWC (private) 128,319 98% PWC (rental) 2,838 2% PWC O WNERSHIP BY R EGISTRATION Private vessels 694 / 77% .
Initial Temp of PWC was(27 ), and Electric Heater exchanges its thermal energy to PWC, a PWC heated up to melting Temp(saving energy as a sensible heat). After that, the heat stored as latent heat, thus the PWC melts and becomes liquids phase. Then the energy saved as sensible heat as a liquids phase PWC. The Temp PWC is registered at a period of
In this document, "PwC" refers to PricewaterhouseCoopers Priv ate Limited (a limited liability com MS 219-September 2011 S&R .indd Designed by: PwC Brand and Communications, India www.pwc.in Contacts Shashank Tripathi Executive Director 91 98196 78900 shashank.tripathi@in.pwc.com Anurag Garg Senior Manager 91 9711701799 anurag.garg@in.pwc .
predictive analytics and predictive models. Predictive analytics encompasses a variety of statistical techniques from predictive modelling, machine learning, and data mining that analyze current and historical facts to make predictions about future or otherwise unknown events. When most lay people discuss predictive analytics, they are usually .
in advanced mathematics used in US universities are also popular in Australian universities for students studying engineering and some areas of applied sciences. However, the advanced mathematics .
