1. Intraship Integration Control Instructor: TV Prabakar

2. Functional Requirement  150 sensors are placed on turbine.  Sensors read Temperature and Pressure every 50 ms.  Monitoring System checks for threshold Bounds.  If unusual, gear of the turbine is changed.  Should respond to External Control signals.  Should balance load among all four Turbines.

3. Constraints  Should not effected by Electro Magnetic Interference.  Components should be ‘MIL’ standard.  Database updates are restricted.

4. Obvious Architecture Temperature Sensors Pressure Sensors Turbine Monitor Gear Controller Data

5. Architecture Evaluation  It is Block Diagram, not an Architecture.  What is Turbine monitor Black box?  Does Turbine Control Component is readily available?  Assumed that no component fails.  Using above diagram, can development be started?  How many Network interfaces Turbine Monitor should have?

6. Improved Block Diagram Mote 1 Mote 2 Mote 3 Mote 4 Mote 5 Concen- trator Concen- trator Turbine Monitor Gear Controller Data

7. Priority of Quality Attributes  Availability.  Performance.  Usability.  Testability.  Security.  Modifiability.

8. Quality Attribute Dependency  Performance ∞ Availability.  Performance ∞ 1/Security.  Performance ∞ 1/Modifiability.  Performance ∞ 1/Usability.  Availability ∞ Security.  Availability ∞ 1/Usability.

9. Availability Driven Design Apply Removal from service Tactic to every component.

10. Sensor Fault recovery µP Active redundancy Temper ature sensor Pressure Sensor

11. Sensor Fault recovery µP

12. Tactics Applied When µP samples sensor, reads incorrect Value or no value. Fault Detection  Heartbeat.  voting  Active Redundancy  Spare  Shadow operation

13. µP Fault Recovery

14. Tactics Applied  Redundant µP informs the concentrator.  Concentrator doesn’t read data from mote. Fault Detection  Heartbeat.  voting  Exception.  Transaction.  Passive Redundancy.  Active Redundancy.  Spare.  State resynch.

15. Mote to Conc. Cable Fault Recovery Concen- trator

16. Mote to Conc. Cable Fault Recovery Concen- trator

17. Tactics Applied  µP senses the channel before sending, if not in operating voltage, chooses spare. if not in operating voltage, chooses spare.  Similarly Concentrator does. Fault Detection  Heartbeat.  Spare.  Transaction

18. Concentrator Fault Recovery Concentrator

19. Conc. To Monitor Cable Fault Recovery Concentrator Turbine Monitor

20. Conc. To Monitor Cable Fault Recovery Concentrator Turbine Monitor

21. Open the Turbine Monitor Black Box Slave processor Shared Memory Controller Slave processor Controller Turbine Monitor System Monitor Heartbeat

22. Slave Processor Fault Recovery Slave1 Slave2 Shared Memory Controller Slave2 Slave1 Shared Memory Controller

23. Tactics Applied  After Processing of operations each slave sets the their flag byte to one.  At end of deadline, controller checks the flag byte, if zero, respective slave failed. Fault Detection  Heartbeat.  Voting.  Active Redundancy.  Spare  Shadow operation.

24. Controller Fault Recovery Shared Memory Controller Slave Shared Memory Controller Slave Shared Memory Controller Slave

25. Tactics Applied  Each controller lives for fixed safe duration and initializes its spare as controller. Fault Prevention  Removal from service.  Passive Redundancy.  Spare  State resynchronization.

26. Poor Resource utilization….  Each processor can act as Controller and Slave.  Each removed Controller becomes spare of slave, acts as slave till it dies.

27. Turbine - Gear Cable Fault Recovery Gear Monitor Turbine Monitor

28. Turbine - Gear Cable Fault Recovery Gear Monitor Turbine Monitor

29. Open the Gear Monitor Black Box Slave processor Shared Memory Controller

30. Slave Processor Fault Recovery Slave1 Slave2 Shared Memory Controller Slave2 Slave1 Shared Memory Controller

31. Controller Fault Recovery Shared Memory Controller Slave Shared Memory Controller Slave Shared Memory Controller Slave

32. Performance Driven Design

33. Current Architecture µP Contro ller Slave Shared Memory Contro ller Slave Shared Memory Mote Concentrator Turbine Monitor Gear Monitor

34. Compute function Change gear Get data From ROM sense Data send Data receive Data send Bounds checking Glow LEDS Raise Alarm Check failures Store threshold Data receive Data send Data send Data receive send ACK Receive ACK Data send Data receive Data receive

35. Allocation view sense Bounds checking Data send Data receive Store threshold Data receive Check failures Raise Alarm Glow LEDS Data send Data send Data receive Data receive Data send Data send Compute function Get data From ROM Receive ACK Data receive Change gear send ACK Mote Concentrator Gear Monitor Turbine Monitor

36. Simple Range checking program, choosing high speed hardware will give high performance.

37. Hardware Specifications  8-bit µP on mote.  1 KBps cable from mote to concentrator.  16-bit µP on Concentrator.  1MBps cable from concentrator to Monitor.  Pentium Processor on Turbine Monitor.  1KBps cable from Turbine to Gear Monitor.  Pentium Processor on Gear Monitor.  Packet size from mote is 5 bytes.

38. Delay on mote.  To sense Temperature (ADC) = 1 ms  To sense Temp. thru spare (ADC) = 1 ms  To sense Pressure ( ADC ) = 1 ms  To sense Pressure thru spare = 1 ms  To sense channel (ADC) = 1 ms  To sense spare channel (ADC) = 1 ms  To send packet on channel (DAC)= 1 ms Total delay at mote = 7 ms. Total delay at mote = 7 ms.

39. Mote to Conc. Cable delay 1000 Bytes = 1 sec 1000 Bytes = 1 sec 5 Bytes = 5 ms 5 Bytes = 5 ms In µP Active redundancy mode load is 10 bytes Total network delay = 10ms Total network delay = 10ms

40. Concentrator delay  To sense the channel (ADC) = 1 ms  To sense spare channel (ADC) = 1 ms  To read 300 packets (ADC) = 300 ms  To process and aggregate data = 2 ms  To sense other channel (ADC) = 1 ms  To sense other spare channel = 1 ms  To send 150 packets (DAC) = 150 ms Total delay at concentrator = 456 ms Total delay at concentrator = 456 ms

41. Conc. To Monitor cable delay To transfer 1500 bytes = 1.5 ms Delay at Turbine Monitor Total delay = 150 + 25 + 150 Total delay = 150 + 25 + 150 = 325 ms. = 325 ms.

42. Monitor to Concentrator cable delay To transfer 1500 bytes = 1.5 ms Delay at Concentrator Total delay = 456 ms Total delay = 456 ms Conc. to mote delay Total delay = 10 ms Total delay = 10 ms

43. Total processing delay Total delay = 7 +10 + 456 + 1.5 + 325 Total delay = 7 +10 + 456 + 1.5 + 325 + 1.5 + 456 + 10 + 1.5 + 456 + 10 = 1256 ms (or 1260 ms approx) = 1256 ms (or 1260 ms approx)

44. Evaluation  Not satisfying functional requirements.  Sensitivity points are ADC and DAC.  Packet errors during transmission are not considered, may result in retransmissions.  EMI constraint is not taken considered.

45. Possible solutions  Increase redundancy in ADCs and DACs  Use checksum for transmission error detection.

46. Comments What is the redundancy number of ADC and DAC? 150 ADCs and 150 DACs may solve the Problem (infeasible). Checksum - increases network delay by 1 ms. - increases network delay by 1 ms. - calculation takes 5 ms ( min. ). - calculation takes 5 ms ( min. ). - checking takes 5 ms. - checking takes 5 ms. Retransmissions are not decreased.

47. Effective solution  Use Fiber optic cable.  Data should be sent digitally ( elimination of ADC and DAC ) of ADC and DAC )  No effect of EMI.  Retransmission free.

48. Poor Resource utilization….  For efficient usage of bandwidth, connect all components in bus Architecture.

49. µP Modified Architecture µP Contro ller Slave Shared Memory Turbine Monitor Contro ller Slave Shared Memory Gear Monitor Concentrator mote Fibre optic cableFibre optic cable Fibre optic cableFibre optic cable

50. Need of Concentrator Acts as proxy Delay calculation Total delay = 7 + 0.3 + 2 + 0.015 + 25 Total delay = 7 + 0.3 + 2 + 0.015 + 25 + 0.015 + 2 + 0.03 + 0.015 + 2 + 0.03 = 36.615 ms (or 37 ms approx) = 36.615 ms (or 37 ms approx)

51. Performance Tactics used  Concurrency  FIFO  Caching. Performance Techniques used  using ROM rather than disk to store data.  No contention of resources.

52. For four Turbines Turbine Monitor1 Turbine Monitor2 Turbine Monitor3 Turbine Monitor4 Gear Monitor1 Gear Monitor2 Gear Monitor3 Gear Monitor4 Turbine Monitor

53. Usability Techniques  Usability component is connected externally to Monitoring system.  Provides different views of logged data (graphs, thermometer reading, numbers ).

54. Security  Security component is connected external to the monitoring system.  Only read operation is provided in menu for user.  Level of security is chosen by user.

55. Testability Tactic  Separate interface from implementation. Testability Technique  LED display at concentrator.  Different alarms according to level of urgency.

56. Architecture Documentation  Chosen views - Class view. - Class view. - Process view. - Process view. - Deployment view. - Deployment view. - Allocation view. - Allocation view. - Use case view. - Use case view.

57. Logical view

58. Use case view Sensor failure Turbine Monitor Turbine Monitor Network failure Turbine Monitor Turbine Monitor

59. Use case view Abnormal Condition Turbine Monitor Gear Monitor Needs Spare Turbine Monitor

60. Technology Architecture  Atmel ATMega processor on mote and concentrator.  Pentium processor at Turbine Monitor.  ECos, Real time operating system.  HP UDO 30GB write-once disk.  C language for implementation

61. Take away  For any problem “Deployment view” is first step (if reference Arch. not available).  Attribute Driven Design is tool to achieve Quality attribute.  Design – Evaluation iteration results in good Architecture.  Current Technology (with specifications) should be known in advance to design an Architecture.