Model Based Troubleshooting Methodology and Tools Dan Koren
Model Based Troubleshooting Methodology and Tools Dan Koren – HP Indigo March 2013
Agenda 1. Major TS methodologies 2. Model Based Reasoning (MBR) concept 3. A very simple example 4. More complex cases 5. Available tools and vendors 6. Modeling in DSI-e. Xpress 7. VSE-Diagnostician implementation 8. Demo 9. Q&A 2 HP Confidential
Troubleshooting vs. Diagnostics • • Troubleshooting – guided process to find the cause for a system symptom by isolating the failure Diagnostics – free activation of individual tests HP Confidential
Major TS Methodologies • Statistical approach (“CBR”, Bayesian Networks) • TS trees, Decision trees • Model Based Reasoning HP Confidential
Major TS Methodologies Symptom “A” • Statistical approach (“CBR”, Bayesian Networks) • TS trees, Decision trees 0% - solution A 1 22% - solution A 2 10% - solution A 3 …… Pros’ – very simple • Model Based Reasoning Cons’ – it builds up only with field experience – requires “rebuild” on updates HP Confidential
Major TS Methodologies Symptom “A” • Statistical approach (“CBR”, Bayesian Networks) • TS trees, Decision trees 0% - solution A 1 22% - solution A 2 1 10% - solution A 3 2 …… Pros’ – very simple • Model Based Reasoning Cons’ – it builds up only with field experience – requires “rebuild” on updates HP Confidential
Major TS Methodologies Symptom “A” Perform Test 1 - if pass then, else…. • • Statistical approach (“CBR”, Bayesian Networks) TS trees, Decision trees e. g. Test 1 - Pass…. . Test A - Fail …. . Test B Pros’ – very simple • Model Based Reasoning Cons’ – it is normally based on past experience – requires system level knowledge – no solution if a test can’t be performed – prompt to “little/obvious” issues overlook – requires “trees rebuild” on updates (normally NOT performed) HP Confidential
Major TS Methodologies Symptom “A” Perform Test 1 - if pass then, else…. • • Statistical approach (“CBR”, Bayesian Networks) TS trees, Decision trees e. g. Test 1 - Pass…. . Test A - Fail …. . Test B Pros’ – very simple • Model Based Reasoning Cons’ – it is normally based on past experience – requires system level knowledge – no solution if a test can’t be performed – prompt to “little/obvious” issues overlook – requires “trees rebuild” on updates (normally NOT performed) HP Confidential
Major TS Methodologies A “model” of the system is created • Statistical approach (“CBR”, Bayesian Networks) • TS trees, Decision trees Symptom “A” 1. Prepare a list of “suspects” or possible causes 2. Prepare a list of available tests for the above suspects list • Model Based Reasoning 3. Perform the “optimal test” 4. Remove “innocents” 5. Go to step 1 Pros’ – developed based on design info. – very easy to maintain – all “little/obvious” issues are included Cons’ – requires up-front time investment – “new” methodology HP Confidential
The MBR Concept Input Initial Symptom Data Get Ambiguity Group > 1? Get Best Next Test Run Test & Pass/Fail = “Suspects list” Get <Faulty Item> R/R <Faulty Item> No Matter What the Fault(s), the Logic is the Same! No Matter how Big the System is, the Program Structure is the Same! Isolates "Multiple Fault" Events HP Confidential
The MBR Concept Input Initial Symptom Data TESTS T 1 FAULTS Part 1 Output 1 Part 3 Output 1 Part 4 Output 1 Part 5 Output 1 Part 6 Output 1 Part 7 Output 1 Part 8 Output 1 T 4 P 1 P 2 X X Get Ambiguity Group X Output 2 Part 2 Output 1 T 2 T 3 X X = “Suspects list” X Ambiguity Group > 1? X X X X D-Matrix X X Get Best Next Test Run Test & Pass/Fail Get <Faulty Item> R/R <Faulty Item> No Matter What the Fault(s), the Logic is the Same! No Matter how Big the System is, the Program Structure is the Same! Isolates "Multiple Fault" Events HP Confidential
TS process – A “simple example” 12 HP Confidential
TS process – a “simple example” Plug Cable 1 Switch Cable 2 Test Lamp Test 13 HP Confidential
TS process – a “simple example” Element: Name: Cable 1 Failure modes: Broken: MTBF 20 K hours Bad connection: MTBF 5 K hours … Plug Cable 1 Switch Cable 2 Test Lamp Test 14 HP Confidential
TS process – a “simple example” Element: Name: Cable 1 Failure modes: Broken: MTBF 20 K hours Bad connection: MTBF 5 K hours … Plug Cable 1 Switch Cable 2 Test Lamp Test: Cable 1 connection – bad connection Test: Cable 1 status – broken Test: Cable 1 output voltage – bad connection & broken 15 HP Confidential
TS process – a “simple example” Element: Name: Cable 1 Failure modes: Broken: MTBF 20 K hours Bad connection: MTBF 5 K hours … Plug Cable 1 Switch Cable 2 Test Test: Cable 1 connection – bad connection Test: Cable 1 status – broken Test: Cable 1 output voltage – bad connection & broken Lamp Test Symptom: no light Source: Lamp Suspects list: Lamp Cable 2 Switch Cable 1 Plug 16 HP Confidential
TS process – a “simple example” Step #1: Sort suspects list according to MTBF Step #2: Get all relevant tests from TM Step #3: Run best test for top component (availability, reliability, type, time to run, etc. ) Step #4: If it passes – remove all relevant components from the suspects list and Plug Cable 1 Switch Cable 2 Lamp go back to step #1 Example: if there is power after the switch, remaining suspects are Cable 2 and Lamp If it fails – create a new list starting from the failure point and go back to step #1 Example: if there is no power after the switch, the new suspects list is Switch, Cable 1 and Plug Etc. 17 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 18 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 Symptom Z A B FM-C 1 FM-C 2 D E FM-G 1 F Symptom W K FM-G 2 19 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 Symptom Z A B FM-C 1 FM-C 2 D E FM-G 1 F Symptom W K FM-G 2 20 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 Symptom Z A B FM-C 1 FM-C 2 D E FM-G 1 F Symptom W K FM-G 2 21 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 Symptom Z A B FM-C 1 FM-C 2 D E FM-G 1 F Symptom W K FM-G 2 22 HP Confidential
TS process – more realistic cases Input 2 Input 3 Output 3 Input 4 Symptom Z A B FM-C 1 FM-C 2 D E FM-G 1 F Symptom W K FM-G 2 23 HP Confidential
MBR Known Vendors • QSI, Qualtech Systems Inc. , CT – Diagnostics is their main business • • • Aircraft Industry NASA Airspace Orbotech Communication VSE Corporation, VA – Diagnostics group. P&W is part of the Naval Systems Naval Air Division in a large company KLA Tencor Railway Systems Boeing NASA Etc. Volvo Honeywell US Army DSI International, CA – Diagnostics analysis is their main Saab US Navy business Sweden military USAF Airbus Sorman Information AB, Sweden – Diagnostics (RODON) is one of 2 product lines HP Confidential
Vendors’ Tools Arsenal and Capabilities • QSI Modeling with graphic user I/F − RT monitoring with some diagnostics capabilities (no UI) − Guided TS tool with integrated UI − • VSE Modeling with tabloid user I/F, variety of data upload capabilities (including FMECA, design EDIF/netlist, Diag. DL+ and existing TS trees in SGML format) − RT and non-RT guided TS tool without UI (API to user S/W) − HP Confidential
Vendors’ Tools Arsenal and Capabilities • DSI Modeling with graphic user I/F − No guided TS tools − • Sorman Modeling with graphic user I/F, some data upload capabilities (including design EDIF/netlist) − Dynamic generation of Diagnostic Trees. − HP Confidential
Sample model in e. Xpress 27 HP Confidential
Typical model in e. Xpress 28 HP Confidential
Typical model in e. Xpress 29 HP Confidential
TS & Diagnostics modules 30 HP Confidential
Questions? 31 HP Confidential
- Slides: 31