Intelligence at the Edge for Industrial IoT 1
IntelligenceattheEdgeforIndustrialIoT
1
Proprietary and Confidential
FogHorn Background
Purpose-BuiltEdge
Platform
Experienced,SuccessfulTeam
KeyIndustrialPartnerships
SiliconValleyStart-UpEst.2014
EdgeIntelligenceSoftwareforIndustrialIoT
2
SeriesAinQ22016SeriesBinQ42017
Proprietary and Confidential
IIoT andAIIndustryRecognition
3Proprietary and Confidential
#1 HotIoT StartuptoWatchin2017
The10CoolestTechStartupsof2016
1. Qualcomm2. Cisco3. Intel4. FogHornSystems5. AmazonWebServices6. Microsoft7. Everythng8. Google9. Tesla10. IBM
Proprietary and Confidential
AscendancyofEdgeComputing
4
Proprietary and Confidential
TransportationWearables
Connected Cars
Connected Homes
Connected Cities
Industrial Internet
Manufacturing
Utilities
Oil&Gas
Healthcare
TheRiseoftheInternetofThings
…IoTdeviceswillgrowtoasmanyas30billiondevicesby2020.McKinsey&Company.Image:GoldmanSachs.
5
Proprietary and Confidential
IndustrialIoTDataVolumeOverwhelming
EdgeintelligencewilldriverealvalueinIndustrialIoTLessthan1percentofthedatabeinggeneratedbythe30,000sensorsonanoffshoreoilrigiscurrentlyusedtomakedecisions.McKinsey
EdgesolutionsarecriticalforIoTCloudmodelsarenotdesignedforthevolumeofdataIoTgenerates.Cisco “Things”generate
moredataeveryday11PB Mining
480TB Jetengine24TB Automatedmanufacturing1TB Large refinery
0.8TB Large retail shop0.5TB USSmartmeters
6
Proprietary and Confidential
Maximizeinsightbyanalyzingreal-timeassetdata• StreamingML• Cleandiverse/noisyOTdataformaximuminsight• Determinesensorhealthinreal-time
ApplyyourbestintelligencetotheEdge• Updatemodelson-the-fly• Deploywithconfidence
OptimalEdgeperformance• Submilliseconddecisioningenablesnewapplications• Compact,commodityhardware/softwarefoundation,NoFPGA
EdgeAdvantage
Edgecomputingshiftsprocessingfromcentralserversoracloudtotheasset.Thisenablesricherdata,fasterreactions,andlowerbandwidth requirements.
Proprietary and Confidential
Monitor Manage
FogHorn EdgeIntelligencefor IndustrialIoT
FogHorn Differentiators• Tinyfootprint• OT-centric• Cloudagnostic
EdgeNode(PLC/DCS,Gateway)
8
FogHornEdgeIntelligence
MachineLearning
Enrichment
CEPAnalytics
Applications/SDK LocalHistorian
Transport/Publication
KeyCustomer Benefits• Lowersbandwidth/hostingcosts• Triggersreal-time insights• Enablesproactiveusecases• Maximizessecurityandprivacy
EdgeManagement/Monitoring/Configuration
ClosedLoopAnalytics/MLEdgeCloudFogHorn Manager
DataIngestion
MainUseCases• ConditionMonitoring• PredictiveMaintenance• AssetPerformanceManagement• IndustrialProcessOptimization
Proprietary and Confidential
EdgeProcessingAdvantagesforAnalytics/ML
9
Significanteventsmissed
Insignificantevents(sensorfluctuations)appearsignificant
Controllogic,streaminganalyticsandMLinferencesachievefarhigherfidelityonlivedataEd
ge(GB’s)
Clou
d(M
B’s)
Edgeprocessingdelivers:• Higherquality,cleanerdata• Reductioninfalsepositives• Maximuminsight• Fasterresponse• Betterinferences• Faulttolerance
Proprietary and Confidential
ClosedLoopMachineLearning
10
MLModels
DeepLearningModels
Filtered,Normalized,EnrichedData
EdgeMLTM
VELTM CEP Analytics
DataEnrichmentEnterpriseReferenceData
OperationalInsights
Business InsightsMES
Proprietary and Confidential
FogHornIIoTPartnerEcosystem
11
IndustrialSolution Providers CloudInfrastructureandAI/MLCompanies
IIoT ConsultantsandSIs
IIoT GatewaySuppliers
IIoT SemiconductorDevelopers
Proprietary and Confidential
IndustrialIoTUseCases
ManufacturingAPMandProcessIntelligence
DrillingEquipmentPredictiveMaintenance
Pipeline LeakandCorrosionDetection
Compressor/ValvePredictiveAnalytics
PumpConditionMonitoringandPredictiveMaintenance
RenewableEnergyOutputForecasting
WindTurbine OptimizationanfPredictiveMsintenance
Mining EquipmentTrackingAndAssetOptimization
LocomotiveFuelConsumptionandRemoteMonitoring
SmartCitiesandSmartBuildings
IntelligentReal-TimeHealthMonitoring
12
TurbinePerformanceMonitoringandOptimization
CHALLENGE
FOGHORNSOLUTION
BENEFITS
Proprietary and Confidential
ImprovingCapacitorProductionYield
• Hard-to-detectfailureconditionsreducingyieldandincreasingscrap
• Noreal-timemonitoringoflargeamountsofsensordata
• NoOT-centricanalyticsformanufacturing teammembers
13
• FogHorn VEL™:Real-timeanalyticsonwindingmachinesensordata
• EdgeML™:MLonnormalizeddatastreamsforrealtimefailurealerts
• IterativerefinementofVELanalyticsandMLmodelstoassistoperators
MANUFACTURING
Improveyield,reducescrap
Deliverreal-timeanalyticstoOTstaff
Smart,notscheduled,maintenance
ANALYTICS/MLONWINDINGMACHINEDATADETECTSEARLYDEFECTS
CHALLENGE
FOGHORNSOLUTION
BENEFITS
Proprietary and Confidential
AutomatedFlareStackMonitoring
• Monitorlargenumberofflarestacks
• Limitedcommunications/computeresources
• Ensurecompliancewithenvironmental/regulatory requirements
• Reducelargespendonmaintenanceandcompliance
14
• FogHorn installedintoexistinggateways(<1Gb)
• Realtimeaudio/videoanalysisofflarefeeds
• Convolutionalneuralnetworks(CNN)fordeeplearning
• Sensorfusioncorrelateflarestatewithcompressoraudio
OIL&GAS
LowerOpexandmaintenancecosts
Broadcompliancemonitoring
Improvedsafety
REALTIMEVIDEOANALYTICSANDROOTCAUSECORRELATIONANALYSIS
CHALLENGE
FOGHORNSOLUTION
BENEFITS
Proprietary and Confidential
LocomotiveOperationalEfficiency
• Optimizefuelusage
• Detectsub-optimaloperatingconditions
• Reducemobilenetworkingcostsofmonitoring
15
• FogHorn installedintoon-boardhardeneddatacollectionsystems
• RTanalyticsonidling&throttledatabasedonlocation,speed&time
• Proactivealertssenttocommandcentersforoperationaloptimization
• Videoonlysentonabnormalconditionsreducingcellularcosts
TRANSPORTATION
Reductioninfuelandcellularcosts
Optimizecrewandtrainperformance
Ensuresafeoperatingconditions
ON-BOARDANALYTICSDRIVECENTRALIZEDOPERATIONALOPTIMIZATION
CHALLENGE
FOGHORNSOLUTION
BENEFITS
Proprietary and Confidential
OptimizingElevatorPerformance
• Monitor1.5M+elevators/escalatorsdeployedglobally
• Limitedcommunications/computeresources
• Minesensorinformationforactionableinsights
• Reduceinspection/repairfeesof~$2K/event
16
• FogHorn installedonexistingmotionsensorkits,<1Gbfootprint
• CEPtime-alignsstateandactivitydatain<20linesofcode
• 40+MLmodelsgeneratepredictivemaintenancealerts
SMARTBUILDINGS
Smart,notscheduled,maintenance
Reducecostlyrepairandservicing
Newmanagedservicerevenue
50+MLMODELSONTINYCONTROLLERSDELIVERPREDICTIVEMAINTENANCE
CHALLENGE
FOGHORNSOLUTION
BENEFITS
Proprietary and Confidential
WindFarmOutputForecasting
• Monitorlargevolumesofwindmills
• Limitedcommunications/computeresources
• Accuratelypredict,reportandmeet24hourpowergenerationgoals
17
• FogHorn installedintoexistinggateways
• Modelstrainedon20+attributestopredictpowergeneration
• Real-timescoringonpowergenerationwithalertsforproblems
• Enablestechniciantuningofsettingsorrevisedforecast
WINDENERGY
Alertswith90minutesleadtime
Constantlyupdatedpowerforecasts
Ensuregovernmentcompliance
REALTIMETURBINECONTROLS-DRIVENMACHINELEARNINGFORECASTS
IntelligenceattheEdgeforIndustrialIoT
18