Data solves core movement dilemma.

Inside This Story:

* Relying on speculation to solve problems in the metalcasting process can be dangerous without data to back up your findings.

* Detailed within is how statistical tools helped solve core movement and provided a clear solution.

Variation is present in all aspects of the metalcasting industry. Many decisions affecting the casting process are based on the interpretation of those variations. Confusion between a simple, random variation and a variation indicative of change can lead to greater variability in the process and higher costs. Yet, many of these important decisions are made with speculation and emotion, while only occasionally are they based on interpretation of actual data.

The first step to improvement is to understand what types of causes are dominating the process--special or common. Special causes are not part of the process and arise due to a specific situation. They may be fixed by making a quick correction to the process. A common cause may be the result of the whole system and occurs quite frequently, affecting many people.

Any deviation from an ideal process can represent a potential loss. Therefore, process knowledge must be gained before the fact, not after something goes wrong. This can be achieved by defining variation tolerances on all critical product and process parameters, minimizing the number of steps that comprise the process and standardizing the processes.

Also, continual use of statistical process control tools to isolate, control and eliminate variation can provide data on the problem, giving metalcasters something more than speculation on which to base vital decisions.

This article details the application of statistical tools to detect and improve core movement.

Core Movement

A complaint about core movement (rotation, tilt and vertical shift) appeared to be a problem of non-conformance in a component cast by Dalton Corp., Warsaw, Ind. This problem surfaced in the form of customer feedback because the defect was intermittent enough to pass through quality control in the metalcasting facility. Because the defect occurred only occasionally, further investigation was needed to find the cause before considering a solution.

First, focus was placed on getting the customer involved early. The metalcasting facility had its people talking to its customer's people at all levels. Then, a cross-functional team involving the metalcasting facility and customer was formed. The team consisted of those directly affected by the problem and the supervisor responsible for the line where the improvements were to be made.

Over a seven-month period, 264 castings were reported defective. Data was collected for all of them, and the frequency of defects was sorted and displayed in a Pareto Chart (Fig 1). The chart indicated that four reasons for the defects were responsible for 80% of the returned castings. The other five reasons were responsible for 20% of the returns.

[FIGURE 1 OMITTED]

The next step was to figure out what was causing the defect and where the core movement was occurring. Six specific data points were identified on the product, and their tolerances as per specification were noted. A schematic diagram of the design of the gating system also was obtained.

To better understand the process, the team met to discuss the problem. During a walking tour of the metalcasting facility, the team noticed that the process involved a manual procedure that presented limitations for quality--the possibility for error rested on a human factor. Further, the operators' inspection procedure was a visual check, when the accuracy of a coordinate measure machine (CMM) would have been ideal.

Monitor the Process

Throughout the data gathering and data analysis stages, the team used an array of tools (check sheets, charts, diagrams, histograms, stratifications, scatter plots and graphs) to gather the information to list the potential causes of variation in the processes. The team used a cause-and-effect diagram to lead a brain-stomping and discussion session to list possible causes for core movement.

Using measurements to define the severity of the core movement, the team decided to randomly select 30-50 castings and measure them. A data collection check sheet was designed to record the product number, casting date, actual measurements for the six data collection points and the overall length of the casting. Using the data, a control chart (Fig. 2) was designed to determine if the process was dominated by special or common causes. The control chart consists of the process' quality characteristics plotted on a graph and three lines. The centerline is the average of the data points while the upper and lower control limits bind the statistics due to common causes. Points outside of these limits are considered special causes.

[FIGURE 2 OMITTED]

A process in which the outputs are affected by common and special causes and in which variation is unpredictable from one occurrence to the next is unstable. When a process is unstable, it can be hard to differentiate changes to the process from special causes as well as when changes result in improvement. Based on the data collected, the six data points were out of control.

Analyze the Data

The quality control chart and cause-and-effect diagram were placed on the wall where team members who understand the process could ask questions and generate action items to define the root cause and create a process of continuous improvement.

Sources of variation were identified, and a statistical relationship between the process input and output variables were established, which provided data to support how often core movement occurred, when it occurred, indicators of where it could have occurred and why it was different (rotation, tilt, vertical shift). The question of where to begin the investigation was answered by a sorted cause-and-effect matrix, which suggested that the likely factors were operator error, difference in measurement systems, improper support/set-up, broken/damaged tooling and lack of assembly procedures (Table 1).

A pilot study was conducted, at which time changes to eliminate operator error were implemented, and castings with the improvements were sent to the customer to take measurements and provide feedback.

Engineers and operators at the customer site designed an error-detecting gage to generate a go/no-go signal to indicate whether the existing inventory of castings was free of core movement. If the casting was rejected by the gauge, the CMM expert measured the casting to develop data to confirm core movement and eliminate Type I errors (where the gauge found core movement when in fact core movement did not exist). This allowed the customer to detect castings with core movement before machining, preventing the generation of scrap and losing productivity and capacity.

In addition, the date code of the casting with core movement helped the identify inconsistencies, such as day-today, machine-to-machine, batch-to-batch and shift-to-shift operations.

The team worked with management to develop a deployment plan to monitor the process and ensure the problem does not reoccur. The improvements to optimize the outputs and eliminate or reduce core movement included:

* total process movement to ensure the core does not move in the mold line;

* a proposal to extend the locator on CMM inspection fixture to use the same radial locator as the machine fixture;

* evaluating placing a linear locator on the inspection fixture to simulate the machine fixture locator;

* tightening the core setter's guides to hold the core secure in position within the core setter for the helical gear series;

* verifying the clearance between the two cores in the corebox so that each one is assembled accurately;

* adding a stabilizer bar to the core setter on the product model;

* changing the power take off core prints to provide more hold-down for the core;

* standardizing the procedures of assembling the core to minimize or reduce human error.

Experimentation

Given the demand on the molding line, it was a challenge to conduct experiments without holding up production. Therefore, the team decided to conduct the experiment by changing one variable at a time and holding everything else fixed. The team generated the data sheet for the experiment and performed potential failure mode evaluation analysis as needed. The data collected was analyzed with a statistical advisor, and the results were presented in graphs that indicated improvements.

The team then set out to determine whether the process had or had not changed. This was done by defining the P-value. Holding all other factors constant, the P-value is the probability of the ratio of the difference between the two population parameters and the average standard deviation of the two populations based on data during core movement and data after the corrective actions were taken.

To determine if the specified improvements did reduce core movement, data was collected alter the changes to prove beyond some criteria that the specified improvement in core movement was achieved. A two-sample test on normally distributed data sets was conducted for each of the six data points. The null hypothesis ([H.sub.o]) and the alternative hypothesis ([H.sub.a]) were defined as:

* [H.sub.o]--the proposed changes have no effect on reducing core movement on all six data points;

* [H.sub.a]--the proposed changes do have an effect on reducing core movement on all six data points.

The test was conducted with an A risk equal to 5% and a B risk equal to 10%. Thirty-six samples were measured after the corrective actions, data was entered into a software program and P-value was determined for each of the six data points. After comparing the P-value to the a risk (0.05), [H.sub.o] was rejected. For example, one data had a P-value of 0.019; therefore there is a 1.9% risk in deciding H is greater than [H.sub.o] when compared to the initial 5% risk that was being made.

Statistical Confirmation

The solutions utilized to counteract causes, operating levels, tolerances and installed solutions provided statistical evidence that the solution improved the processes. As a result, the number of core movement defects decreased (Fig. 3), and customer satisfaction increased. The result is improved productivity while minimizing operating expenses, decreasing defects returned by the customer and improving reaction time by focusing on products rather than people.

[FIGURE 3 OMITTED]

Through Dalton's experience, the firm uncovered a number of reasons as to why it had a problem with core movement. The following show Dalton's findings.

1. Mechanical problems with the mold line were causing the core to move in the mold due to excessive vibration and impact as the cope and drag reached the transfer station

2. Dalton's layout fixture was adjusted to match the radial locator where the customer's fixture locates the part.

3. Coresetter guides were tightened to hold the core more secure in position within the coresetter for the helical gear series.

4. The clearance between the large and small cores at the core prints for the bores was repaired to specification to assure proper clocking.

5. A stabilizer bar was added to the core setter to help hold the core in the correct position while being placed in the mold.

6. Pins, bushings and seals were replaced as needed.

7. Dalton changed the way the core was clamped in the assembly fixture to stabilize the stock distribution in the area where drilled holes will be located on the shift bar opening in the insert core area.

8. Adequate time was alotted for the glue to cure before the cores were set in the mold to reduce the chances of the cores spreading apart.

Dalton was able to achieve the desired results because of the use of statistical tools. In the past, when problems were solved with speculation, the results were not always successful.

In the metalcasting industry, quality issues often are not addressed until the product is already completed, which is too late. Also, the "firefighting" approach to problem solving may result in short-term solutions to immediate problems but at great expense to long-term process improvement.

Making decisions based on numbers can provide significant financial results. If a company focuses on making quality everyone's business, mastering decision-making based on numbers, and implementing process monitoring and control, it will have a better chance of reducing defects and improving productivity.

Relying on speculation to reduce variation can be a dangerous approach because solutions could work in the short-term, but cause more problems over time. A more accurate approach is to focus on analyzing stability and capability using information in graphs.

Table 1. Potential Causes & Reasons for Returned Castings

Reason for Returned Casting          Cause

Guides on "nest" not tightened       Improper support, Improper set-up,
                                     broken/damaged tooling

Stabilizer bar not in proper         Improper support, Improper set-up
position on coresetter

Cores don't have correct             Improper support, Improper set-up
tolerances between them

Cores not assembled correctly        Improper support, Improper set-up

Cores not assembled with correct     Improper support, Improper set-up
amount of glue

Locator pin on mold pattern not      Improper support, Improper set-up
present

Bushings on mold pattern are warn,   Improper support, Improper set-up,

damaged, warped                      Broken/damaged tooling

Locators on inspection fixture of    Improper support, Improper set-up
casting don't coincide with
machining fixture locators

Measurement methods are not          Improper set-up
consistent with customer's

Stabilizer bar on coresetter is      Broken/damaged tooling
damaged

Corebox not to specification         Broken/damaged tooling

Locator pins on mold pattern are     Broken/damaged tooling
worn and/or damaged

Shock at end of mold is not          Broken/damaged tooling
functional

For More Information

"Economic Control of the Quality Manufactured Product," W.A. Shewhart, Reprinted by AFS, 1980.

P.L. Barker Jr., Dalton Corporation, Warsaw, Indiana

B. Bidassie, Six Sigma Black Belt/Statistician, Galesburg, Michigan

Paul Barker is the director of industrial engineering at Dalton Corp., Warsaw, Ind. B. Bidassie is a Six Sigma Black Belt/Statistician with Eaton Corp., Galesburg, Mich.