Asset performance management using Druid by Eric Lim, Bistel
Download as PPTX, PDF1 like417 views
This document discusses asset performance management (APM) and how to use Druid for APM. It contains the following key points: 1. APM involves monitoring industrial assets like production lines, conveyor belts, and semiconductor manufacturing equipment to predict failures and perform preventative maintenance. 2. Druid is an open source analytics database that can be used with tools like Kafka, Flink, and Metatron for real-time health monitoring, analytics, and predictive maintenance of industrial assets. 3. The Metatron APM system powered by Druid allows users to configure asset templates, build data pipelines from IoT sensors to Druid, perform fault detection and prediction, and generate dashboards and
Asset performance management using Druid by Eric Lim, Bistel
- 3. 산업군 전반의 1. What is APM? (2) 생산라인의 조립 로봇 (3) 물류센터의 컨베이어 벨트 모터 (1) 반도체 제조라인의 반송기 자산의 잔존수명 예측 & 예지정비 2. What is Asset?
- 5. AdvancedAPM IndustrialIoTreal timehealth monitoring& analytics Stream Flows Metatron Asset Performance Management with IIoT CloudService CloudInfrastructure DockerContainer (K8S) Sensor data Motor sensing data Network device Robot 3rd-PartyIIoTEDGE Devicesensordatacollect/datatransformation/monitoring … IIoT EDGE APM Cloud Env. Health Alarm & monitoring Model Manager Predictive analytics Asset meta management Collaborated Prescription Docker App Management Container Operation & Management Metatron eco. Asset meta Connector Job monitoring Flow(Pipeline) Management Flink Pipeline Flink Pipeline Flink Pipeline Flink Pipeline1. Data Stream Analysis Discovery Data Source Management Data Source Explorer(Search) Notebook Connector Cube(druid) 3. Visualization for raw & analyzed data2. Request for raw & analyzed data metatron ecosystem APIGateway
- 6. APM… 뭘 할 수 있나요?Part.2
- 7. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics • Rule-based Model • Machine Learning based FD Model • Domain-based Model (Current data) • Multivariate Model (Statistical Model) # Index Data Source Real-Time Summary 1 Fault Detection ● 2 SPC ● 3 Variation Detection ● 4 Parameter RUL ● Phase1 Threshold by Machine Learning Alarm Message Phase2 Phase3
- 8. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics • Parameter • Asset Score • Asset Alarm
- 9. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics < Summary View > • Total Asset Status • Latest Alarm History • Worst Performing Assets • Worst Performing Parameters < Asset Statistics > • Total Asset Status • Asset List • Asset Status • Parameter Analysis • Remaining Useful Life (RUL) • Total Alarm Count < Alarm Statistics > • Alarm Status Distribution • Alarm Count • Top5 Assets • Top5 Parameters • Top5 Subscribers
- 10. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics Fault Classification using Vibration Spectrum Data RMS Trend Chart Spectrum Chart Time Wave Chart Vibration Data
- 11. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics • Root Cause Analysis • Asset Matching Motor # 7 Motor # 9 Motor # 2 Motor # 8
- 12. Fault Detectio n Fault Classification Dashboard RUL Predictio n Real-Time Monitoring Data Analytics Failure Now Prediction for Failure RUL: 10 days Red : Linear Regression Model Blue : Prediction Line Green : Upper & Lower prediction interval
- 14. Asset Asset Asset Template Type Custom attribute (property) Asset Model (for score) Parameter Group (Sensor) Parameter Configuration Create fromTemplate MetatronIIoTEDGE Source (Kafka) Sink (Kafka) Transform Window Aggregation Model Notification Alarm (Kafka) Score (Kafka) Data(Stream) FlowPipeline metatron Discovery Alarm Score Trace Data Source Mgmt. 1. Request tocollecttheasset data 2. Create stream pipeline 3. Create datasourcefromkafka Asset Kafka indexingservice Kafka indexingservice Kafkaindexingservice
- 15. Asset template A Asset template B * Asset A-1 * Asset A-2 … * Asset B-1 * Asset B-2 …
- 16. 3rd-PartyIIoTEDGE Source (RAW DATA) Transform Model #1 Alarm Alarm (Kafka) Asset Score Data(Stream) FlowPipeline metatron Discovery (Druid) Alarm Score Trace Trans Set #1 Transform Trans Set #N Model #2 Model #N 3rd-PartyApp Model #N TransformRuleTask Model EvaluationTask AlarmRule Task Kafka indexingservice Kafka indexingservice Kafka indexingservice
- 18. AlarmDatasource
- 20. AlarmDetail (Alarm Group) AlarmGroup -MariaDB
- 21. Confidential Copyright © 2018 BISTel,Inc. All rights reserved. Thank You BISTel Tower, 128 Baumoe-ro, Seocho-gu, Seoul, Korea 82 2 597 0911 marketing@bistel.com www.bistel.com