Statistical Process Control (13+15 = 28 Questions) - ENGG381-14A Engineering Statistics (Task A, B, C) - 15 Questions done twice
ENGG381-14A Engineering Statistics Due: Monday 9th June Assignment 5
1
Produce a report which clearly presents your responses to the following tasks. Brevity is appreciated but explain what needs to be explained! You may discuss suitable approaches with classmates but must work independently on your own report. You should submit your report using Moodle by 10.50 pm Monday 9th June.
TASK A: SPC for Rod Thickness rodthickness2014.mtw [23 marks]
A plant manufactured approximately 12,000 connecting rods per day for use in an engine assembled in the plant. The rod, illustrated above, connects the piston (at the small or pin end of the rod) to the crankshaft (at the large or crank end of the rod). The plant received forged blanks and machined the rods in a large number of process steps. Management identified the rod line for a variation reduction project because the overall scrap cost was greater than budget. The yearly scrap cost was excessive, and the scrap rate had been 3.2% over the previous four months. Looking at scrap records, the team found that scrap occurred at several stages in the process and for several reasons. The results showed that 65% of the scrap occurred at a grinding operation. At this operation, the team discovered that about 90% of the scrap was due to rods with their crank end thickness less than specification. The team focused their attention on reducing variation in rod thickness. Their first step was to set up control charts to see whether the grinding step produced rods with a stable crank end thickness. The worksheet contains the output from the first five days of charting. Thicknesses were recorded from subgroups of 5 consecutive rods chosen at 8 times spread through each day. As in normal operation, each rod was measured at 4 positions (white circles in figure); the worksheet contains their mean thickness as a deviation from 0.900 inches in thousandths of an inch (i.e. mil). In that scale the specification range is [10, 60] mil and in practice if any of the 4 deviation measurements was less than 10 mil the rod would be scrapped. Those with any over 60 mil were reworked. In this assignment, however, we will assume that the specification interval applies to mean thickness. To do this, you will need to create a new variable, MnThick, which is the average thickness across the four positions for each rod.
based on a Steiner and MacKay case study
ENGG381-14A Engineering Statistics Due: Monday 9th June Assignment 5
2
Question 1
a) Display an and X s chart for MnThick established from the first 3 days but plotting all data so both establishment and monitoring periods may be displayed on one plot (use Stat > Control Charts > Variables Charts for Subgroups > Xbar‐S, and enter the appropriate details in Xbar‐S Options under the Estimate tab). You should also include a reference line to distinguish establishment and monitoring periods. [2 marks]
b) List what if any actions should have been taken after the establishment period and then during monitoring, given that the objective was to achieve and/or demonstrate process stability. [2 marks]
Question 2
a) Now repeat the control chart analysis of the process mean using the EWMA chart (see Stat > Control Charts > Time‐Weighted Charts > EWMA). You should as before establish on the first 24 subgroups. [2]
b) Provide plausible explanations where this chart suggests different courses of actions compared to the previous question. [2 marks]
Question 3 Use the Minitab Stat > Quality Tools > Capability Analysis > Normal menu to produce a capability analysis display for the mean thickness data which compares the data recorded with the specification. [3 marks] Question 4 Based on the analyses above, write a statement which summarises the quality of the rods in relation to the mean thickness specification, and comment on whether the analysis is likely to assist with the scrap reduction objective. [3 marks] Question 5 Now construct a new artificial mean thickness variable by subtracting 5.54 from each data value after subgroup 26.
a) Recreate the charts of questions 1 and 2 for the new variable. [2 marks] b) Comment on whether the results are consistent with the information on average run lengths
displayed in the extracts from Caulcutt’s book in the Topic 9 lecture notes. [2 marks]
Question 6 Does treating the specifications as being applicable to the mean over 4 positions increase or decrease the percentage of rods classed as out‐of‐specification? Explain without referring to the data provided. [2 marks] Question 7 Investigate whether it is important to consider position when considering reasons for scrap. [3 marks]
ENGG381-14A Engineering Statistics Due: Monday 9th June Assignment 5
3
TASK B: SPC for Pigment Manufacture Process dizogoblue2014.mtw [17 marks]
A company manufactures a range of pigments for use in the textile industry. One particular pigment, dizogo blue, is made by a well‐established plant which has recently been renovated. For example the capacity has been increased and the agitation system has been made fully automatic. The data file contains information collected from the first 50 batches after renovation. It is feared that the expected increase in yield has not yet been realised and that the level of particular impurity has increased. A data logger has also recorded incidents where the agitation speed has been automatically reduced because the agitator was overloaded. Question 1
a) Treating the 50 batches as an establishment period, set up individuals control charts to monitor the average levels of yield and impurity under the new conditions. [1 mark]
b) Comment on what you conclude from the control charts. [2 marks] Question 2
a) Explain why it is more important to check data normality for an individuals chart than for an X ‐ s chart. [2 marks]
b) Use probability plots to check data normality for the two variables being monitored and draw conclusions, being careful to recognise that non‐normality might sometimes be just an indicator of the presence of special causes. [3 marks]
Question 3
a) Repeat the setting up of an individuals chart for impurity but this time choose I Chart Options > Box‐Cox > Optimal Lambda. [2 marks]
b) Explain what you now conclude. [2 marks] Question 4
a) Set up a EWMA chart for impurity. [1 mark] b) Explain what you conclude, and whether this differs from the previous analysis. [2 marks] c) Describe the different purposes of an I chart and a EWMA chart. [2 marks]
ENGG381-14A Engineering Statistics Due: Monday 9th June Assignment 5
4
TASK C: Statistical Engineering Algorithm for truckpull2014.mtw Truck Pull [19 marks]
This data set presents another view of the baseline study of Truck Pull from Steiner and MacKay previously described in lectures. The full data set contains data on the 28,258 trucks manufactured over 44 days. The data file truckpull2014.mtw has the alignment information for every tenth truck off the line and we have restricted the data set to the first 40 production days, which we will treat as 8 five‐day weeks. (Check the column formulae & descriptions in the Minitab worksheets for useful information.)
Pull = 0.23 × (r‐caster – l‐caster) – 0.13 × (r‐camber – l‐camber)
crosscaster = (r‐caster – l‐caster) and crosscamber = (r‐camber – l‐camber)
Question 1
a) Display means and standard deviations for the four alignment angles and also pull. [1 mark]
b) Assume that the alignment angles are approximately independent of each other. Show how you would calculate estimates of the mean and standard deviation of pull from the sample means and standard deviations for the individual alignment angles. [2 marks]
c) Caculate the percentage error in your calculated estimate of the standard deviation of pull compared to the observed sample standard deviation of pull. State whether you would expect your calculated values of the mean and standard deviation of pull to match the observed sample mean and standard deviation. Explain your reasoning. [3 marks]
Question 2
a) Note: the following analysis is more easily done in Excel than Minitab. Assuming the estimate of the SD of pull you calculated in Question 1 was satisfactory, for each angle calculate the reduction in the standard deviation of pull which could be achieved if you halved the standard deviation of just that angle (i.e. calculate the reduction for r‐caster, l‐caster, r‐camber, l‐ camber one at a time, separately) . Which alignment angle does this analysis suggest should be worked on to reduce variation in pull? [2 marks]
b) Extend your “theory” from part a) to usefully identify the single component to be worked on to most reduce the standard deviation for an “assembly of independent components” of the
general form 21 1 2 3 3 4 4
Y a X a X a X a X . [2 marks]
c) What practical consideration implies that the strategy suggested in part b) will not always be
helpful, even if the component angles are independent? [1 mark]
Part C, Slide 231
Part A, Slide 59
ENGG381-14A Engineering Statistics Due: Monday 9th June Assignment 5
5
[OPTIONAL] Question 3
In slide 237 of topic 10, it is suggested that variation in pull might be reduced if assembly was “selective” instead of “random”. In other words we might consider the theoretical benefit of being able to match the alignment angles of an assembly to reduce variation in pull. In the data file truckpull2014.mtw I have created two new columns sortxcaster and sortxcamber by sorting the crosscaster and crosscamber columns (separately) from lowest to highest values.
a) Plot crosscaster against crosscamber, and sortxcaster against sortxcamber. Display the linear correlation coefficients for these two plots.
b) What can you deduce about the theoretical potential for a reduction in pull variation from selective assembly? (Be quantitative. Hint: be the manufacturer!)
Question 4
Staff applying the “statistical engineering algorithm” for improving crosscaster have decided that the dominant cause of variability lies in the “between trucks” family not the “between weeks” “between days” or “between shifts” families.
a) Produce a fully nested analysis of variance for the crosscaster variable (see Stat > ANOVA). Recall that for a nested ANOVA, you need to specify the predictor variables in order from the highest level of nesting (i.e. weeks in this case) to the lowest. Use the output to evaluate the decision to focus on the variation between trucks. You may assume the baseline variation covers the interval [‐0.1, 2.1]. [2 marks]
b) Calculate the percentage reduction in the overall standard deviation for crosscaster if we could completely remove the truck‐to‐truck variation. [2 marks]
c) Based on the estimate of overall standard deviation of crosscaster from the nested ANOVA, calculate the Capability Ratio (Cp) for crosscaster when it has specification limits of 0.975 ± 0.9. What is Cp if we manage to completely eliminate truck‐to‐truck variation? [2 marks]
d) What is the implication of the choice of dominant cause on how the reduction project should proceed? [2 marks]
Part C, Slide 159