Most recently, in the processing of the topic, we came to the description of the FTA / Standards. Now let's move on to the Good / Bad parts.
Root cause analysis
Root cause analysis
FTA / BAD / OK PARTS
- Allows comparison between reality versus standard: bad versus good parts
- Real data coming directly from measurement of BAD and GOOD parts or process.
- EVERY TIME IT IS POSSIBLE, ORGANIZE MEASUREMENT WHEN THE PROBLEM IS STILL THERE:
- Ask Customer to perform measurement at Customer site jointly before send bad part back.
- Provide special instruction to Operator to measure (ex. the air pressure) when problem occurs next time
- OK Parts = Good part: value of factor on the OK parts produced in the same process than the Bad Parts
- Bad part: value of the factor on the Bad parts
FTA / NON-DETECTION
Make the process mapping of the process:
- Process description step by step
- You must start from the customer and go back to the station where the defect is created.
For each step, look at all the control points:
- Are they designed to catch the defect ?
- Which factor made this detection to fail ?
FTA / HOW TO REPRODUCE
- Asking Operator to create defective parts having same signature than “bad” parts
- Observation of process to capture facts when similar problems appears
- Exchange of components shows that problem follows one component (Example using Shanin technique: component search)
- Parts manufactured with same characteristics than “bad” parts show similar signature (Example using Shanin technique: component search)
- Parts manufactured with slightly amplified deviation on suspected process parameter show similar signature
FTA / OCCURRENCE
What happened at Part just before the Bad part (T-1)?
- How was the product just before becoming bad ?
- What can explain the phenomenon?
- Ask Operators
This is the basis for the completing the FTA:
- You need to be at the real place
- To observe the real facts
- And to gather real data
FTA / SOME RECOMMENDATIONS
- Why do you say there is no standard ? Is it the first time ?
- Why do you say no control point? How do you verify ? Do you have start up check list, please show me ?
- Is it easy to control ? Is repeatable ? known and done by all Operators?
- Is there any difference between good and bad (not only parts think at situation, timing, person, process, machine …. )?
- Did you get this measurement from the period when bad parts were manufactured ?
- Did you compare good and bad versus standard?
- Where is the data are coming from? How it was measured?
- Think 4 Ms: Material, Method, Machine, Men.
At this stage, it is key to be able to summarize the issue by saying without blah-blah:
- The problem is xxx.
- The main root cause is … This has been demonstrated by … The corrective action is … The corrective action will work because……..
- The main reason why it has not been detected is … This has been demonstrated by … The corrective action is …
Test your conclusion by explaining why no problem occurred prior to this occurrence!
- Don’t jump to conclusion without facts
- Compare Good/ bad/ Standard
- Verify after each why, based on fact test data
- Involve Operators and all related persons in the analysis
- Follow the PDCA / 8D standard formats: you don’t need anything else!
In the next post, I will write about Japanese methods closely related to error analysis.
There are no reviews yet.