Cable Splice Failures

 

What Is The Issue?

Electrical cable splices are causing major failures at an important chemical plant.  The failures result in significant losses of gross margin!

 

Why Do We Have The Issue?

Experienced maintenance engineers clearly believed the cable splice failure mode was due to chance failures.  Thus failures came “from out of the blue” at random times apparently without rime or reason.

 

How To Resolve The Issue?

Use failure data to find a maintenance strategy for avoiding future failures.

 

What Are The Facts?

         63 splices were installed between 1969 and 1980 as the plant site was expanded.

         19 splices have failed--44 have survived.

         Each planned replacement costs ~$40,000.

         Each unplanned failure costs ~$500,000, or (19 failures thus far)*($500,000/failure) = $9,500,000 costs incurred through 1994!

         Common knowledge says expect the failure mode is a chance failure since it’s an electrical device. 

         The repair strategy has been “fix it when it breaks” starting with the first failure in 1977 and the subsequent failures through 1994.

No one wants cable splices.  However, when you’re expanding the size of a very large chemical plant cable splices are inevitable.  Electrical loads on the cable splices have large and sudden changes in currents.

 

What Does The Data Tell?

Notice there are three parts to the failure data:
            1) age-to-failure for the failure data (literally failure data),
            2) suspended (censored) ages for old splices which have not yet failed, and
            3) suspended ages for new replacement splices.

This data is listed below in Figure 1 and converted into a Weibull probability plot using SuperSMITH Weibull.

 

Figure 1: Weibull Plot Of Age-To-Failure

 

Two important statistics come from the Weibull plot of components (b and h, along with the goodness of fit criteria r2).  The Weibull plot says the cable splices are failing from a wear-out failure mode, i.e., line slope b is greater than 1, and the characteristic age to failure, h, is 32.72 years!  The failure mode IS NOT a chance failure mode described by the plant experts.

 

When Will The Next Failure Occur?

You can approach the forecast of failures two ways: 
            1) Make an Abernethy risk of failure from the probability plot, or
            2) build a Crow-AMSAA plot. 

Both techniques are explained in The New Weibull Handbook.

 

The Abernethy risk forecast will used the data in Figure 1 (excluding the infant mortality failure mode that was suspended at year 2 as they decided an error was made in construction of the splice and it failed after 2 years of service) and SuperSMITH Weibull software to make the prediction.  Note in Figure 2 the credibility check of the present risk of failures is 18.4 versus 19 actual failures.

 

Figure 2: Abernethy Risk Forecast Of Future Splice Failures (~2.4 failures/year)

 

A second way to forecast failures is by use of a Crow-AMSAA reliability growth plot.

 

Figure 3:  Crow-AMSAA Reliability Growth Plot Forecast Made In 1998

 

Why is the X-axis in cable years of service?  For many years they were expanding the plant and increasing the number of cable years of service.  Thus a better common denominator was cable years of service.  Could the X-axis be given in years now the plant has stopped expanding, the answer is yes.

 

Why was a forecast made for year 1998, as shown by the circle on Figure 3?  This company reviewed some key demonstration projects completed by students who participated in the Reliability Engineering Principles course.  This case study was the demonstration project of the Manager of Maintenance and Engineering.

 

Transforming the data in Figure 3 you will get the cumulative MTBF plot shown in Figure 4.  This plot clearly shows reliability has been declining up until 1994 when corrective action took place to turn the curve in the positive direction by avoiding failures.

 

Figure 4: Cumulative MTBF With A Favorable Cusp Showing Positive Improvements

 

The green trendline, in Figure 4, shows an improvement in MTBF.  The Cum MTBF vs Cum Cable-Yrs of service directly shows reliability is improving.  As MTBF improves by avoiding failures the reliability of the cable splice systems improve.

 

What Has Happened Since 1998?

Accepting the failure mode as wear-out was a difficult pill to swallow.  The first thought was to consider optimum replacements as shown in Figure 5 where the data is generated in SuperSMITH Weibull and plotted in SuperSMITH Visual.

Figure 5: Optimum Replacement Cost Curve For Different Failure Costs

 

In Summary-

The brain lock was broken when the wear-out failure mode was clearly accepted as demonstrated by the Weibull plot.  Cable splices can flex with violent changes in current which would flex the mechanical joint and let air under the insulation which resulted in oxidized surfaces.  The oxidation in turn builds heat in the splice until it gets so hot it explodes.  Then the Ah Ha! occurred as to why a wear-out mode could exist. 

 

Recognizing existence of a wear-out failure mode then suggested the quickest and easiest task would be to search for hot splices using infrared technology.  The hottest connections were then replaced with more rigid and better insulated splices which has much longer life.

 

I visited this plant in October of 2009.  The scorecard continues to be no in-service cable splice failures since the new maintenance response was implemented in 1994.  This has been a very valuable project for this chemical company with savings of millions of $’s with one training program.  The bottom line:  Use your numbers to make important decisions.

 

Now here comes the rub.  Too many important RCM decisions about failure modes and cost effective alternatives are based on opinion and not on the numbers!  For small cost failures, opinions may be OK.  However, for expensive failures always make your data talk! 

 

Balance your risks against your corrective/preventive actions!  Let your analytical Weibull’s tell you the failure modes (save your opinions for your bar room stories where they make great conversations).  In the maintenance and engineering world we have too many opinions and we fail to make our numbers work for a technological solution.  Engineers, use your scientific tools.

Return to Barringer & Associates, Inc. homepage

July 14, 2010
© Barringer & Associates, Inc., 2010