RollingArt

Nice post Joaquin!

Now this is why we have forums.

Looks like nice CAD work from your engineer buddy.

Very robust looking solution. What is the upgrade to the actual bearing itself?



Phil

juankimalo

I used an increased F125 steel bearing support. Here you can see the differences between OEM 8 mm and custom made (12 mm):




SKF hybrid steel / ceramic ball bearing:






SKF bearing = 155 € = 210 $

Lathe operator bearing = 50 € = 68 $

nick49

Lathe operator bearing = $68

12mm custom made support $63
actual bearing cost $5

I assume this is close to an actual cost breakdown. I'm basing this on doing these type of conversions myself in my line of work. I have my own machine tools and I purchase a lot of bearings also.

dresler

Holy cow! Wish I spoke Spanish..

juankimalo

Sorry if I didn't explain correctly...


I mean:

Lathe operator labour bearing support 12 mm steel F125 = $68
Hybrid bearing from SKF = 210 $

nick49

Thanks for explaining it,

WOW! that is a hybrid bearing. I thought somewhere it was mentioned that a Vespa scooter bearing was used.

Also, FWIT, if the support and housing is being made from scratch it seems a non proprieraty could be used.

Thanks!

juankimalo

My mechanic posted the whole operation in a website and as he is very funny and ironic, he told that Vespa's motor wear a similar bearing.

this is which I used: SKF reference: 6204-2RSLTN9/HC5C3WT

http://www.skf.com/skf/productcatalo...bleName=1_23_1


(You can see 3D view)







Both bearings with their support (left OEM, right custom made/SKF)

Graufuchs

An update.......Article by OP....published regarding JD Power Associates 2010 VDS.

http://www.thetruthaboutcars.com/the...bility-survey/

Interesting.....

mkaresh

Sorry, I had no idea these responses were being posted. Either automatic comment notification is not available or it is not turned on (I'll check after posting this).

The negative posts in this thread, the ones directed at me at least (it seems to have gone OT), follow from misreading / not reading what I wrote, insufficient knowledge of statistics, and more than a touch of cynicism.

--I obtained the formal permission of rennlist to post here. This is noted in the OP.

--I worked the first four years on this at a loss, and continue to put in far more hours than can be justified by how much it earns. Anyone who has started a project like this from the ground up and then worked to get thousands of people involved can imagine what I've had to go through. Others, perhaps not. Saying that I'm conducting this research for some underhanded reason without any basis for saying this is simply insulting. But then I've learned that if you don't want to be insulted without cause, don't post in forums

--I don't ask people to pay for results they've helped provide the data for. I explicitly state in the OP and on the site itself that anyone who participates gets full access to my site for free. Let me repeat this, in case it was missed the first time: participants get full access for free. That's free, as in no money. Some people manage to read the opposite, but I can only control what I write, not what people read.

--The size of the population has absolutely no relationship to the necessary sample size. It's not even part of the formula. This is statistics 101. Larger sample sizes are certainly always better. But 25, and sometimes even fewer, can certainly provide useful information.

--The survey process was designed to deal with the problem that people might join because of the problems they've had with a car. And it has been very successful in dealing with this problem. For one thing, it collects data going forward. Past problems cannot be reported on the main survey, and so cannot influence the results.

--People charge all of the time that the survey will just be based on people who want to complain. Fact of the matter is, we have some car models for which over 90% of car owners report no repairs at all in the past year, and for the average car model over half report no repairs at all in a given year. So I think we can consider the "complainer hypothesis" rejected.

--Data from dealers wouldn't be of much use unless you knew how many cars this data COMPLETELY represented. This might not be possible. Better yet would be warranty data from the manufacturer. But it doesn't matter, because dealers and manufacturers aren't going to provide this data.

--If I have to choose between some ideal that simply isn't going to happen and a possible, practical alternative, I'll go with the latter. It's the difference between talking and doing.

--I'm well aware that not everyone cares about reliability stats or wants to participate in something like this. I conduct this research for people like myself who are interested.

If anyone seeking to help has made it this far into the thread, you can read more about the survey and sign up here:

Car reliability research

s_kelly

I'm not sure why everyone if flaming mkaresh & jumping to these conclusions about validity. You can't declare a survey invalid if you don't know all the details about the goals of the survey. Sample size is not the be-all end-all of whether a survey is valid or not, depending on what they're trying to accomplish. For something like a national opinion political survey, a representative sample of all constituents from all 50 states would be important. Having such a population might not be so important if the survey was designed to measure the reliability of, say, aging porsches owned by enthusiasts who post on the internet a lot.

Wellardmac

See post #27 and you'll see why no matter what the OP says, the endeavor is still statistically flawed.

EastBay

AND YET, it is the entire business model of Consumer Reports auto reports. And most people believe CR to be the gospel truth!

I say good luck to the survey.

BruceP

+1

The main issues here are that respondents self-select, and that they self-report over time. The latter is a bigger deal than you might think. Just ask A.C. Nielsen what happened to television ratings when they switched from diaries to people meters.

By the way, the OP's comment about sample size might have been a bit flip. If I assume a population of cars somewhere around 200,000 and a study sample of 25 people as he suggests, the margin of error would exceed 20%, 95% of the time. In other words, if five people in the sample had their ronky booboo valve fail, the projected number of failures for the total population would be anywhere between near-zero and 80,0000 units. You see the problem

RallyJon

A much better solution is legislation that requires automakers to give up their internal data. With the recent Toyota situation, combined with the current administration's tendencies, I wouldn't be surprised to see something like that in the future.

mkaresh

The self-selection is not a major factor if you collect data going forward, word the survey as objectively as possible, and keep the load light enough. Nielsen's survey required too much of participants. And, IIRC, the main thing that was discovered with the people meter is that people used the bathroom and so forth during commercials. Shocking, I know.

I'm not saying a sample size of 25 is ideal. I'm saying that at that level you do get a viable indication of how often cars require repairs. I would not attempt to measure the failure rate of a specific part with 25 people.

It's easy to trot out hypotheticals. Here's what our current results look like for the Boxster, in terms of repair trips per 100 cars per year:

2008: 67, small sample size

2005: 103

2003: 125, small sample size

2002: 167, small sample size

2001: 125

2000: 129

1999: 133

Those labeled "small sample size" have fewer than 25 responses, so they're asterisked and only visible to members on the site. Even then there's only one result out of line with the others, the 2002. The 1999, 2000, 2001, and 2003 are all very close together.

I see this over and over in the survey. There are some results that are clearly off, and are commented as such, but these are easily exceptions and not the rule, and don't happen once the sample size is over 50.

Do you think this happens by chance?

I can guarantee one thing: claiming that the survey isn't viable because of the sample size tends to be a self-fulfilling prophesy.