Preformed Line Products (PLP), a leading manufacturer in the overhead power lines industry, uses @RISK and its GoalSeek function to develop an intelligent and tailored method for bringing together different hardware components to work together in a working and robust system. The method has saved time and money for PLP and enabled outside vendors to create parts more efficiently and produce less reject components.
Background
Preformed Line Products (PLP) is a worldwide designer, manufacturer and supplier of high quality cable anchoring and control hardware and systems, fiber optic and copper splice closures, and high-speed cross-connect devices. PLP Australia has a fully equipped Engineering department to design and test such fittings for both the Australian and South East Asia markets. PLP Australia often has to bring together products manufactured by specialist manufacturers with their own methods, standards and specifications, which in some cases makes it difficult to combine hardware parts from two different suppliers.
Phil Timbrell, Manager of Engineering Services at PLP Australia, faced this problem with the process for manufacturing spreader rods, which are used to hold low-voltage lines in windy conditions. The two primary parts of this component are a pultruded rod, and a spring that clamps the rod to a conductor.
For the spreader rods to be functional, the rod diameter has to be bigger than the spring. This puts restrictions on the manufacturing specifications for both the rod vendor and the spring vendor. But Timbrell wanted to determine how much ‘wiggle room’ each vendor could get away with and still have a system that was robust and effective.
Traditional Tolerance-Stacking: Splitting the Difference
When product manufacturers combine several different components, they do ‘tolerance stacking’, which analyzes the accumulated variation allowed by specified dimensions and tolerances. Every component may have a flaw or deviation from the specifications that can be ‘tolerated’ by the overall design to form a functional whole.
For the spreader rods, a spring maker makes one component with nominal dimensions that might be skewed in certain ways “that we as a non-experts don’t truly understand,” says Timbrell. “The same goes for the rod made by a pultruder—this is created in a totally different process using totally different raw materials, with its own set of tolerances.”
Previously, the industry put what Timbrell calls ‘dumb tolerances’ on products. “We tell our vendors that we can’t afford for the interference to be more than a half a millimeter,” he says. “And then we split the difference of that variation between the two part manufacturers, and ask them to comply with two equal sets of tolerances.”
However, this arbitrary ‘down-the-middle’ approach is potentially inefficient. Timbrell explains that, for one manufacturer, the tolerance may be very difficult to comply with, while for the other manufacturer may find it completely easy.
“Designers would simply look at the maximum allowable interference fit and spread this between the two external vendors, virtually forcing them to try to comply with tolerances that may not be suited their manufacturing process,” says Timbrell. “The result was several occasions arose in which products had to be rejected at incoming inspection because it was out of tolerance. The root cause of the problem was the tolerancing forced on the supplier did not match their process.“
Manager of Engineering Services, Preformed Line Products Australia
@RISK Provides a More Sophisticated Approach
Timbrell decided to use existing statistical process control (SPC) data from the external suppliers to model a probabilistic approach to tolerance stacking with @RISK. “This would result in tolerances that were both within the scope of the external supplier and also could be quantified into the correct tolerance stacking,” he says.
Random-sample measurements from the rod and spring manufacturer formed the inputs to a Distribution Fitting cell (a skewed triangular distribution which is typical of the appropriate manufacturing process). The two Distribution Fitting Cells from both pieces are then subtracted to provide an interference fit that can be run as a simulation. “From those inputs, I can then get the probability of two pieces either fitting together or being off,” says Timbrell. “I can also look at the tornado graph to see which of the processes is giving me the most problems.”
Then, using Advanced Analyses and Goal Seek, Timbrell and PLP are able to see what maximum or minimum value are required of the most sensitive component in order to keep the two mating parts within specification. Those maximum or minimum values are then overwritten into the raw SPC data and highlighted, and the simulation is then run again. The new graph shows a correct interference fit, and that new tolerance is advised to the vendor. “If the vendor cannot achieve this tolerance, we may repeat the process for the other vendor,” says Timbrell. “The ultimate outcome is that, if the vendor keeps to within certain limits, then we will have no tolerance stacking issues, and zero rejects at incoming inspection.”
Timbrell says that the most useful feature of @RISK for this work has been Goal Seek, which uses multiple simulations to find an input value that achieves a specified target simulation. “It can make any part of the model the subject of the equation whilst running a probabilistic spreadsheet,” Timbrell adds. “It’s to Palisade’s credit that they built Goal Seek into @RISK—it really makes this software appealing to engineering departments.”
Distributions Used
Results from @RISK’s Goal Seek function
This new approach has been very successful for manufacturing the spreader rod at PLP Australia. Now, rather than applying a one-size-fits all approach when ensuring both vendors meet the tolerance-stacking requirements, Timbrell is able to give them specific guidance that works for them. “I can ease off of one vendor who found the restrictions too tough, and give stricter requirements to the other who can easily accommodate that new restriction,” he says. Timbrell says this sophisticated new method could benefit those in any manufacturing or engineering field where different parts from different manufacturers need to work together. “This technique allows one to shift the tolerances and share them more equitably,” he says. “It’s really reduces the probability of having a non-working product.”
This novel approach has worked so well with the spring-rod manufacturing process, PLP is looking to integrate @RISK into other aspects of its processes. “@RISK is a wonderful piece of software,” says Timbrell. “I can’t sing the praises of this software enough—it is so easy to use, and there is so much support.”