Robinson rubber products, Minneapolis, MN, has previously experienced an intermittent scrap rate of up to 10%, while producing challenging injection molding parts, both internal and external steel sleeves, and in
The most common problem is that this part does not go through the hard metal of the company --to-
Rubber bonding test, requires that the rubber will actually tear and not be squeezed out of the metal.
Robinson has built a lot over the yearsfactor-at-a-time (OFAT)
This process changes, but the problem continues to occur.
Robinson\'s quality assurance manager works with Stat
Rest assured, experimental design of Minneapolis, MN, supplier (DOE)
Software, toperform, an experiment to evaluate the impact of five key manufacturing variables.
The experiment determined the value of each variable to optimize the rubberto-metal bond.
As these values have been used all the time, the scrap rate has remained below 1% and the bond issue has not appeared.
Design and manufacture of customized Robinson rubber products
Provide mold assembly, extrusion profiles and precision rollers to original equipment manufacturers and industrial companies around the world.
The company focuses on complex components, including multiple insert parts, homogeneous rubber parts, and rubber bonded to metals and other substrates.
Robinson\'s main markets include the auto market.
Highways and heavy highways
Many industries such as automobiles, machinery, material handling equipment, household and commercial appliances. Tough rubber-to-
The part mentioned above uses two layers of adhesive for each metal insert.
The insert is manual. Load a16-
Mold, injection of rubber, parts stay in the mold at the specified time, temperature and pressure.
In order to verify the adhesion, the operator regularly performs destructive tests on the product, pushes the internal steel inserts out of the Assembly and visually checks the rubber-to-metal bond.
In the past, the company experienced a lot of rejections on a regular basis due to bond failures.
Engineers and operators will express their views on this issue in the process.
It\'s easy to disagree because there are a lot of variables in the manufacturing process for this part.
The injection speed, injection pressure, vulcanization temperature and residence time of the injection molding machine can be easily adjusted.
The thickness of the bonding agent and the time allowed for drying are other important variables.
The process changes dynamically before the problem disappears.
Then, after a period of time, the problem will appear again, and the trial and error process will start again.
When this happens three times in two weeks, the company knows that new ways need to be found to find the root cause of the problem.
A major problem with the OFAT method is that it cannot detect the interaction of multiple factors.
By changing a single variable, onecan finds the best value for each variable, while the others are constant.
However, when a person combines the so-called
The optimization value of each variable, the result is usually far lower than the best value, usually because of the way they interact with each other.
Experimental design evaluates the interaction between variables that have experienced similar problems in the past, suggesting that the only way to truly understand what is happening is to conduct design experiments to find the key factors needed to achieve rubberto-metal bond.
DOE provides a better way to change the values of all variables in parallel, so it reveals not only the main effects of each variable, but also the interaction between variables.
This approach makes it possible to combine to identify the best values of all variables, and also requires much more experimental operation than the OFAT method.
The experimental design only runs k once, where k is equal to the number of variables to be tested;
For example, seven factors were run eight times.
But more precisely, the results can be obtained through higher resolution designs that can assess the main impact of each variable, as well as two-
Factor interaction.
For example, 16 runs for a high
The resolution design of the five factors, at two levels, is only half of the 32 combinations of all these factors. Stat-
SpaceX was selected to train the company in experimental design and experimental device assistanceAnd analysis. Stat-
The development of personal computers
Based on the design of experimental software, the process of solving practical problems is simplified.
World industrial issues
Company software package, design-
Experts, guide users to complete the process of experimental design and operation and evaluation of results. A Stat-
The engineer showed a four.
The Department of Energy conducted an hour of training for Robinson rubber employees, and then everyone worked together to choose the factors they thought were most likely to cause the bond problem.
Robinson\'s quality assurance manager was selected to coordinate and monitor each step of the experiment.
The five factors selected for the experimental design are: * vulcanization temperature (Low to High);
* Adhesive material (two types);
* Thickness of adhesive application (
One coat for two);
* Injection pressure (Low to High);
And * bond settlement time (Low to High).
Robinson rubber engineer input this information into the design
The software creates an ideal two.
Level design for 16 runs.
The experiment was conducted in two days.
The design of the day was blocked. by-day impact.
It includes six additional curesat at the beginning, middle and end of each day, with the same conditions as controls to detect the effects of any variable other than the one measured in the experiment.
According to the design, the order of cure is random
As an insurance expert on potential variables such as machine wear.
In order to simplify the temperature change, a slight modification of the quality assurance manager was introduced.
Quality assurance technicians monitor each treatment in person to ensure that all factors are correct for each treatment.
Three different tests were used to evaluate the results.
The forcerequi that pushed the inner steel core out of the parts was measured.
The remaining rubber adhesion coverage on the steel sled is measured in the range of 1 to 10, of which 1 is barely bonded and 10 is 90% to 100% bonded.
Grade 8 or Better ratings are considered acceptable bonds according to the drawing specification.
The integrity of the cure is rated on one to five levels, one is very low, and the five levels are completely cured.
Three or more ratings are considered acceptable.
Optimize the value of each variable quality assurance technician
Test and evaluate parts together.
All the test results were fed back to us.
Enter them into the software and generate reports, including differential analysis (ANOVA)
Table, which identifies the factors that significantly affect the test results.
The results show that the bond settling and injection pressure have no effect on the results.
The Graphical optimization override shows the best value for each variable and takes into account its interaction with other variables.
Adhesive A is significantly better than adhesive B.
High temperature and medium high temperature are better than low temperature.
Finally, the coating of the adhesive is better than the thin coating.
Robinson rubber began to continuously produce the part under these conditions.
The scrap rate immediately fell below 1% and has remained low since then.
Experimental design is clearly a useful tool that can help determine the root cause and solution of difficult problems that may otherwise last for many years.
John Engler, Robinson rubber, Jerry fireman, structural information (www. strucinfo. com)