Process Knowledge and Product Characteristics Gained Using Casting Process Optimization
Figure 3
The U.S. foundry sector is a $30-billion/year industry, however as a group metalcasters lose $1.5-3 billion annually due to internal and external failures in processes. Do foundries operate their processes at optimum levels? Does a technology gap in process knowledge and control exist in the industry?
Cast metal components
Metal components are designed by mechanical engineers of original equipment to perform certain functions. Using their experience, they may consider casting to be the right process for the shape they have given to the metallic component. Usually, they carry out stress analysis on the casting and determine the various dimensions of the part. In a few cases, the company that designs the casting will also have a casting facility.
However, in most cases, the company that designs the casting obtains the cast part from its vendor foundries. The task of the foundry is to ‘design’ the casting again and supply it while meeting the design specifications. For these foundries, “casting design” means the design of the gating and risering systems and heat treatment procedures to produce the casting with the properties as intended by the designer.
The designer does not have any specific preference for any particular process to select as the manufacturing process for the metallic component that he or she designs. Metalcasting processes should compete with other manufacturing processes, such as machining from wrought materials or forging. It is possible that sometimes the designers may not have adequate experience in the progress that is made in the casting process technologies. A close interaction between casting designers and foundry managers will result in a cast product that is manufacturable by the casting process. A slight alteration of the design to suit manufacturability by the casting process with suggestions from foundries will help the casting designers and also the foundries to meet the specifications from the designers of the cast parts.
Process knowledge
Initially, the foundries make “first run” castings, followed by the production castings. It is relatively easy to produce acceptable first run castings. However, it is challenging to make production castings of consistent quality with minimal scrap and rework.
The reason for high losses in foundries is the lack of the necessary process knowledge. Process knowledge can be defined as the list of process variables and their ranges in order to meet the product specifications with the least amount of scrap and rework. Process knowledge is derived from process data and process information. Process information is obtained from process data by appropriate analysis. Process knowledge is the actionable process information.
In the foundry industry, we have different foundries making different types of castings, using different processes. Even in the same foundry the relevant process knowledge is different for different castings they produce. The foundries need process engineers to create the appropriate process knowledge for every part they produce, and use it to produce parts profitably.
The foundries that produce individual parts can develop process knowledge by collecting appropriate data on the process and the product characteristics. We find some foundries collecting enormous amounts of process data, as well as data on product characteristics, but they cannot establish the traceability of the process data with the product characteristics.
At the other end of the end of the spectrum, we also have foundries that do not collect relevant data.
Appropriate Data Collection
Some foundries collect enormous amount of data, use statistical process control (SPC) and plot control charts, and determine the process capability values — and always insist that variability in the process parameters is a very big problem. It is agreed that variability is a problem, however if such variability does not have any effect on product characteristics, it should not be a concern There is a need to accurately determine the process variable tolerances based on their relationship to the product characteristics rather than by intuition or previous experience. People calculate the process capability (CPk) values based on the process variable tolerances and concluded that we should have high CPk values. But, we need to ask how these process variable tolerances have been arrived in the first place. Every foundry should determine the process variable tolerances based on the analysis of production process data with the product characteristics.
Process optimization
Process optimization is the identification and control of the input process parameters (factors) to achieve the desired output (response) in any process. Metalcasting is a complex process with several sub-processes, such as patternmaking, mold and coremaking, melting and pouring, heat treatment, and cleaning and finishing. Six Sigma methodologies have been attempted in steel foundries to minimize the casting defects and improve profitability. Six Sigma uses the DMAIC (define, measure, analyze, improve, control) methodology to improve the relevant processes.
Figure 4
Six Sigma focuses extensively on statistical analysis, as it is a data-driven and methodical approach that drives the process improvements through statistical measurements and analyses. In view of the large number of factors that are responsible for casting defects, the general statistical approach is not always the best. An alternate and more elegant pattern recognition approach is found to be appropriate for the metalcasting-related issues. It is suggested that the foundries follow the Six Sigma methodology, with the exception that a pattern recognition process optimizer is used instead of the conventional design of experiments (DOE).
In the course of DOE, the number of experiments needed depends on several factors, including:
• The number of factors in a metalcasting process is generally high (in excess of 20-30).
• The need to carry out controlled experiments to collect the required data interrupts regular production.
• There needs to be considerable difference in the levels of factors in order to have meaningful results, hence the results could be biased.
• The use of statistical techniques to design and interpret results requires significant training and expertise, making it expensive and time-consuming.
Process optimization software
MetaCause process optimization software is pattern recognition program that is designed to explain the variability in response values (e.g. % rejection rate or variation in the tensile strength of the cast material) using optimal and avoid regions of one or more factor values.
This revolutionary concept was developed at Swansea University with a $1.5-million research effort spread over 10 years. This collaborative industrial research was initially funded by the U.K.’s Engineering and Physical Research Council1 and later developed by MetaCause Solutions Ltd. MetaCause has been identified as an example of Greatest User Collaboration2 achieved through its funded program.
The core novelty of MetaCause is best explained with a foundry example. For an investment casting process, humidity in the shell room is closely monitored as it is related to inclusion defects. Figure 1 shows the variation in response values for defect inclusions. The main goal for process engineers is to identify optimal and avoid regions, e.g., top 25%, top 50%, bottom 50%, bottom 25%, or middle 50% regions in one or more factor values that explain this variation best. An example of factor values for final humidity is in Figure 2.
MetaCause develops a penalty function and gives maximum penalty to undesired response values (e.g. inclusion rate > 3%) and zero penalty if the rejection rate is zero. Penalty is scaled linearly for the rest of the observations.
The MetaCause approach differs from existing Six Sigma tools in that this penalty is shown as the diameter of a circle in the corresponding bubble diagram (Figure 3). It is clear that top 50% region is optimal as it has fewer large circles as compared to the bottom 50%. Such bubble diagrams can be plotted in MS Excel, either for every factor response combination or every combination of two/three factors and a response. This is a more elegant and reliable way of visualizing main effects and interactions than taking blind decisions based on ‘p-values’. It is a common observation in the foundry world that most process engineers do not trust ‘p-value’ based statistics.
MetaCause’s patented method (US Patent No. 7,440,928) generates a ranked list of ‘optimal’ and ‘avoid’ regions among all bubble diagrams and outputs in an easy to read Excel format (Figures 4 and 5). Figure 6 shows the format of data collection sheet for an iron foundry, and a sample MetaCause summary sheet is shown in Figure 7.
Figure 7
Advantages of Process Optimization Software
For metalcasting operations, the advantages of process optimization include:
1. It uses the actual production data collected during the operation of the process. It uses all the factors and all the responses, rather than the filtered factors in conventional statistical techniques.
2. It can handle up to 200 factors and 40 responses at a time, which is not practical with the existing statistical tools.
3. The results are based on recognizing patterns in the bubble diagrams rather than trying to fit one or more statistical distribution onto the data.
4. It creates the process knowledge that can be documented for future use on the specific components.
A process is considered as optimal only if there is no further scope for improvement in the cost of production for the specific part. The foundries need to have Pareto charts for different parts showing the loss of revenue due to scrap or rework (not meeting the product specifications) and identifying the parts that adversely affect the profitability of the company. There is no use of making money on some parts and lose on others.
Once the parts are identified, the foundries need to identify all the process variables that could be related to the defects and collect data with traceability. The analysis of such data should be able to give the ranges of process variables that have positive effect, negative effect, and no effect on the product characteristics. Such analysis should result in the validation experiment to verify the results of the analysis. It would also lead to eliminating the need to collect irrelevant data. The documentation of process knowledge in a suitable format for each part that is manufactured in the foundry should be a regular feature in metalcasting operations. Over a period of time, the foundry personnel would be more experienced in creating dashboards of process variables and their ranges for specific parts in their foundries.
The format of process knowledge is expected to have the following components:
• Process flow charts with data collection points.
• List of all the process variables and their ranges specific to the individual parts. This is obtained from the results of analysis carried out using MetaCause Optimizer.
• Data collection checklists for the process variables.
• Control charts for the process parameters.
• Process capability values for all the relevant process variables.
1.Visit www.epsrc.ac.uk/newsevents/casestudies/success/Pages/scrapwaste.aspx
2. Visit www.epsrc.ac.uk/SiteCollectionDocuments/Publications/Other/LandscapeDigital.pdf
Dr. Hathibelagal Roshan is president of Intellectual Capital Exchange LLC. Contact him at tel. 414-333-4134, or [email protected].
Dr. Rajesh S. Ransing is associate professor at Swansea University in Swansea, Wales, and leads an affiliated company, MetaCause Solutions Ltd. Contact him at +44 (0) 7968 205 178, or [email protected].