Finalbhgggggggggggggggggggggggggggggggggggg_PPT.pptx

Bb

1. PREDICTIVE MAINTANENCE OF A MOTOR USING ENSEMBLE LEARNING Guided By K. Penyameen, Assistant Professor/ECE, VCE. Presented By 1) P. Akash, 2) B.R.K. Balaji, 3) K. Magalingam, Final Year ECE.

2. Agenda Introduction Abstract Work schedule Literature survey Existing system Proposed system Requirements Block diagram Complete setup Results Applications Advantages References

3. Introduction Predictive maintenance is a technique to predict the future failure point of a machine component. So that the component can be replaced based on a plan, just before it fails, Thus, equipment downtime is minimized and the component lifetime is maximum.

4. Abstract Now a days industrial plants increasingly make use of preventive maintenance to minimize downtime (i.e., failure) and thereby increase efficiency, Develop a system that identifies the failure of rotating equipment (for example motors, turbines) based on a stream of sensor data. This can be done by using open source datasets or by simulating the sensor data.

5. Work Schedule S No DURATION WORK STATUS 1 JAN first Week Look out for problem statement in Start-up India Challenge Completed 2 JAN Second Week Go through base paper & research about project requirements Completed 3 JAN Third Week Components collection, Vibration Sensor Temperature sensor Speed sensor Motor Completed 4 JAN Fourth Week Components collection ESP8266 Learnt Ml basics about classification and regression Completed

6. S No DURATION WORK STATUS 5 FEB First week Learnt Keras to build ML models , Studying about various analysis Completed 6 FEB Second week Collecting historical data of machine to build classifiers and regression models Completed 7 FEB Third week Build classifiers and regression models using different algorithms Completed 8 FEB Fourth week Study about LSTM, SVM, Xgb for classification. And setup for simulation Completed 9 MAR First week Result analysis with trained model Completed 10 MAR Second week Overall completion of project Completed

7. Literature Survey SI. NO TOPIC YEAR Description 1 Machine learning approach for predictive maintenance in industry 4.0 2018 August The system was tested on a real industry example, by developing the data collection and data system analysis, applying the Machine Learning approach and comparing it to the simulation tool analysis. 2 Machine learning for predictive maintenance, a multiple classifier approach 2014 August A multiple classifier machine learning (ML) methodology for predictive maintenance ( PdM ) is presented. 

8. SI. NO TOPIC YEAR Description 3 Optimized neural network pf predictive maintenance for Air Booster Compressor(ABC) motor failure. 2019 May Predictive Maintenance of Air Booster Compressor (ABC) Motor Failure using Artificial Neural Network trained by Particle Swarm Optimization.  4 Predictive maintenance for motor based on vibration analysis with Compact Rio 2018 February It enables spectral analysis of the different frequencies present in the vibrations of a motor. 

9. Existing system Time based maintenance, Preventive maintenance, Real time monitoring of machine parameters in dashboard. From this analysis they will schedule maintenance for the machines to minimize the downtime there by increase in efficiency.

10. Proposed system I t’s proposed to predict the remaining useful life (RUL) of a machine then the state of a machine can displayed in the dashboard. By implementing this system which identifies the future failure point of an machine based on the stream of sensor data, with help of historical data we can able to classify the failure points of an machine. According to that we can schedule maintenance. This system helps in reduce in cost due to major issue and helps to predict the RUL of a machine. It will minimize the downtime thereby increase in efficiency.

11. Ensemble Learning Ensemble learning  is the process by which multiple models, such as classifiers or experts, are strategically generated and combined to solve a particular computational intelligence problem

12. Process of making ML model

13. NodeMCU The ESP8266 is a low-cost Wi-Fi microchip with full TCP/IP stack and microcontroller capability . And used as IoT gateway Hardware requirements

14. Vibration Sensor The ADXL335 is a small, thin, low power complete 3-axis accelerometer with signal conditioned voltage outputs. It can measures the static as well as dynamic acceleration resulting from motion , vibration

15. Speed Sensor A tachometer is a sensor device used to measure the rotation speed of an object such as the engine shaft in a car, and is usually restricted to mechanical or electrical instruments.

16. Temperature Sensor The DHT11 is a commonly used temperature and humidity sensor. The sensor is also factory calibrated and hence easy to interface with other microcontrollers .

17. Single Phase Induction Motor Specification 1. 230V AC, 2. 1425 RPM, 3. 2HP 4. Current : 1. With out load 12.0Amps. 2. Initial current 17.0Amps.

18. Software Requirement Jupyter notebook The Jupyter Notebook  is an open-source web application that allows you to create and share documents that contain live code, equations, visualizations and narrative text. Uses include: data cleaning and transformation, numerical simulation , data visualization and much more. Using Keras library to build classifiers and regression models

19. Arduino IDE The Arduino is a cross-platform application that is written in functions from C and C++. It is used to write and upload programs to Arduino compatible boards, and other vendor development boards

20. Block Diagram MCU Rotary Machine

21. Complete setup

22. Dataset used The dataset used for training the models can be found NASA. It includes sensor readings recorded during turbofan engine degradation simulation. Data Set: FD002 Train trjectories: 260 Test trajectories: 259 Conditions: SIX Fault Modes: ONE (HPC Degradation) Data Set: FD001 Train trajectories: 100 Test trajectories: 100 Conditions: ONE (Sea Level) Fault Modes: ONE (HPC Degradation)

23. Results MODEL ACCURACY LSTM 97 LGBM 92 XGB 92 For training the models, we used a window of size 50 i.e , consecutive 50 sensor readings in each input sample. The results obtained for classification problem on test dataset 1 are tabulated below

24. Plot of true and predicted RUL values

25. Dashboard

26. Applications Detect a temperature decline in a steam pipeline, indicating a potential pressure leak Capture increased temperatures in electrical panels to prevent component failures Measure supply-side and demand-side power at a common coupling point for monitoring power consumption Locate overloads in electrical panels

27. Advantages Reduction or near elimination of unscheduled equipment downtime caused by equipment or system failure, Increased production capacity, Reduced  maintenance   costs, Increased equipment lifespan.

28. References 1. Clarisa García Novoa, Gerardo Antonio, Guzmán Berríos and Rodrigo Abrego Söderberg. ”Predictive maintenance for motors based on vibration analysis with compact Rio” 2017 IEEE Central America and Panama Student Conference (CONESCAPAN )   2 . Marina Paolanti , luca and Romeo Andrea Felicetti   “Machine Learning approach for Predictive Maintenance in Industry 4.0” 2018 14th IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications (MESA)   3. Gian Antonio Susto and Andrea Schirru and Simone Pampuri Machine learning for predictive maintenance, a multiple classifier approach IEEE Transactions on Industrial Informatics  ( Volume: 11 ,  Issue: 3  , June 2015  )   4. Nurfatihah Syalwiah Binti Rosli   and Rosdiazli Bin Ibrahim and Idris Ismail Optimized neural network pf predictive maintenance for Air Booster Compressor (ABC) motor failure.   2019 IEEE International Instrumentation and Measurement Technology Conference (I2MTC)