Open access peer-reviewed chapter

Response Surface Designs Robust against Nuisance Factors

Written By

Nam-Ky Nguyen, Mai Phuong Vuong and Tung-Dinh Pham

Submitted: 11 October 2020 Reviewed: 05 March 2021 Published: 10 November 2021

DOI: 10.5772/intechopen.97047

From the Edited Volume

Response Surface Methodology in Engineering Science

Edited by Palanikumar Kayaroganam

Chapter metrics overview

236 Chapter Downloads

View Full Metrics

Abstract

This paper discusses an algorithmic approach to constructing trend-free and orthogonally-blocked response surface designs. The constructed designs have the main effects, 2-factor interactions and second-order effects being orthogonal or near-orthogonal to the nuisance factors such as the time-trend or the blocking factors. The paper also provides a catalogue of (near-) trend-free Box–Behnkens designs and orthogonally blocked Box–Behnkens designs arranged in rows and columns.

Keywords

  • Box–Behnken designs
  • D-optimality
  • Interchange algorithm
  • Orthogonal blocking
  • Trend-free designs

1. Introduction

Consider an experiment to study the effect of processing time, temperature and shear stress caused by pumping on the quality of skim milk powder. The milk for this experiment is blended and stored at 4°C and is used over a week for a series of experimental runs. As the milk quality deteriorates over the week, it is desirable for the scientist to have a design whose runs are in a particular order such that the main effects (MEs), 2-factor interactions (2FIs) and second-order effects (SOEs) are orthogonal or near-orthogonal to the time trend.

Box et al. [1], p. 486 discussed an experiment using the Box–Behnken design or BBD [2] to study the influence of four factors on nylon flake blender efficiency. The four factors are (1) particle size, (2) screw lead length, (3) screw rotation and (4) blending time. The design has 27 runs in three blocks of nine runs each. Let us assume there is a need to add an additional factor, i.e. supplier in addition to the existing blocking factor, say operator and find a design which can accommodate the new blocking factor.

The two mentioned experiments emphasise the need for a special class of response surface designs (RSDs) and experiment designs in general which are robust against nuisance factors such as the time trend and the heterogeneous environment. Particular attention will be given to the BBDs since BBDs are the most widely used 3-level designs.

Advertisement

2. Measure of goodness of a (near-) trend-free design or orthogonally blocked design

Consider the following second-order model for an n-run design with v time trend columns (or blocking factor columns) z1,,zv and m factors x1,,xm, of which m3 factors are at 3-level and the rest at 2-level:

yu=δ1z1u++δvzvu+β0+i=1m3βiixiu2+i=1mβixiu+i=1m1j=i+1mβijxiuxju+ϵuE1

where yuu=1n is the response value of the uth run, zw‘s are the nuisance columns/variables and ϵu is a random error associated with the uth run. If the nuisance columns zw‘s are the time trend ones, there will be two columns, one for linear trend and one quadratic. The linear trend column (first column) is created by scaling a column of numbers 1n by subtracting each value from the column mean and then dividing the resulting number by the largest one. The quadratic trend column (second column) is created by scaling a column obtained by squaring each element of the linear trend column.

If the nuisance columns zw‘s are associated with the blocking factors, there will be v=i=1rbi1 columns where r is the number of blocking factors and bi is the number of blocks/categories for each blocking factor. Before standardisation by subtracting the values of each column by the column mean, these columns are dummy variables, which take value 1 if the uth run is in the wth block and zero otherwise (see [3], Section 8 and [4], Chapter 8).

Eq. (1) can also be written in matrix form as:

y=Zδ+Xβ+ϵE2

where y is an n×1 response vector, Z a matrix of size n×v containing vzw columns in (1), δ a v×1 column vector representing nuisance effects, X is the expanded design matrix of size n×p, β a p×1 column vector of parameters to be estimated, and ϵ an n×1 column vector of random errors. The least square solution for the unknown parameters δ and β in (2) is thus the solution of the following equation:

ZXy=ZZZXXZXXδ̂β̂E3

When the condition

ZX=0E4

is satisfied, it can be seen that the solution for β from (3) will be the same as the one from the equation Xy=XXβ̂, i.e. the equation for a model without the nuisance columns. The condition in (4) is called the time trend-free condition or orthogonal block condition.

To measure how good of a (near-) trend-free design or orthogonally blocked design is, we use the following fraction:

XX/ZZXX1/pE5

where X=ZX includes the nuisance factors and p is the number of parameters of the model to be estimated (i.e. the number of columns of X). A well known result of Fischer [5] states that given a positive definite hermitian matrix G of the form:

G=ABBDE6

then GAD, and the equality holds if and only if B=0. If follows from this inequality that:

XXZZXXE7

which implies the measure in (5) is less than or equal to 1, and it becomes 1 if and only if ZX=0.

Advertisement

3. An algorithm to attain the orthogonality condition ZX=0

Let xi and xu be two rows of X and zi and zu be the corresponding rows of Z. Swapping the ith and uth rows of X is the same as adding the following matrix to ZX:

zizuxixu.E8

We use this matrix result to develop an algorithm to achieve the condition ZX=0, i.e. for constructing (near-) trend-free designs and orthogonally blocked designs. This algorithm has two main steps:

  1. Construct the nuisance matrix Z and the expanded design matrix X. Randomly assign each of the n rows of X to each of the n rows of Z. Calculate f, the sum of squares of the elements of ZX.

  2. Repeat searching for a pair of rows of X such that the swap of the positions of these two rows results in the biggest reduction in f. If the search is successful, swap their positions, update f and ZX. This step is repeated until f=0 or until f cannot be reduced further.

Table 1(a) and 1(b) display the X=ZX matrices of a BBD for three factors in 15 runs: (a) ordered in the presence of both linear and quadratic trends; (b) arranged in three blocks of five runs each. This example shows how the Z matrices are constructed. The X matrix contains a column of 1’s representing the intercept, followed by three columns representing SOEs, three columns representing the MEs and three columns representing 2FIs.

z1z2x0x12x22x32x1x2x3x1x2x1x3x2x3
−111000000000
−0.860.5710110−1100−1
−0.710.21110110−10−10
−0.57−0.091101−10−1010
−0.43−0.321011011001
−0.29−0.481110−110−100
−0.14−0.581110110100
0−0.611000000000
0.14−0.581110−1−10100
0.29−0.4811101−10−100
0.43−0.3210110−1−1001
0.57−0.091101101010
0.710.211101−1010−10
0.860.57101101−100−1
111000000000
(a)
z1z2x0x12x22x32x1x2x3x1x2x1x3x2x3
0.67−0.331101−1010−10
0.67−0.331101101010
0.67−0.331000000000
0.67−0.33101101−100−1
0.67−0.3310110−1−1001
−0.330.671000000000
−0.330.671101−10−1010
−0.330.67110110−10−10
−0.330.6710110−1100−1
−0.330.671011011001
−0.33−0.3311101−10−100
−0.33−0.331110−110−100
−0.33−0.331110−1−10100
−0.33−0.331110110100
−0.33−0.331000000000
(b)

Table 1.

The X=ZX matrices of a BBD for three factors in 15 runs (a) ordered in the presence of both linear and quadratic trends; (b) arranged in three blocks of five runs each.

Remarks

  1. The above two steps make up one computer try. Thousands of tries are required for each design parameters and the one with the smallest f will be chosen. Among designs with the same f, the one with the smallest fraction calculated in (5) will be chosen.

  2. For a factorial or fractional factorial design (FFD), the orthogonality between MEs and nuisance variables is considered more important than the orthogonality between 2FIs and nuisance variables. For an RSD, the orthogonality between the MEs (and 2FIs) and nuisance variables are considered more important than the orthogonality between SOEs and nuisance variables. In these situations, partition X as X1X2 where X2 is associated with the more important effects. Similarly, partition ZX as ZX1ZX2. Let g be the sum of squares of the elements of ZX2 and f the sum of squares of the elements of ZX as defined previously. Step 2 of our algorithm now becomes repeating searching for a pair of runs such that swapping their run positions results in the biggest reduction in g (or f if g cannot be reduced further). If the search is successful, swap their positions, update f and ZX. This step is repeated until f=0 or until both g and f cannot be reduced further.

The designs whose X matrices are in Table 1(a) and 1(b) were constructed by this approach. In Table 1(a), the X2 matrix is associated with the MEs and in Table 1(b), the X2 matrix is associated with the MEs and 2FIs. Readers can verify that ZX2=0.

The algorithm we describe in this Section is closely aliased to the ones of Nguyen [6] and Nguyen [7]. As such, it does not require matrix inversions and therefore is much faster than the ones by other authors (See e.g. [8, 9, 10, 11]).

Advertisement

4. BBDs robust against time trend and blocking factors

The algorithm in the previous Section has been used to construct BBDs for 3–7 factors which are (near-) trend-free and orthogonally blocked with two blocking factors (i.e. rows and columns).

We use the correlation cell plots (CCPs) proposed by Jones & Nachtsheim [12] to display the magnitude of the correlation between the SOEs, MEs and 2FIs with the columns/variables representing nuisance factors. The colour of each cell in these plots ranges from white (no correlation) to dark (correlation of 1 or close to 1). Figures 1(a) and 1(b) show the CCPs of the BBDs whose X matrices are Table 1(a) and 1(b). In Figure 1(a), it can be seen that the MEs are orthogonal to both the linear and quadratic trends. In Figure 1(b), it can be seen that both the MEs and 2FIs are orthogonal to the block effects. For both Figures, the SOEs are slightly correlated with the nuisance effects.

Figure 1.

CCPs of the BBDs whose X matrices are Table 1 (a) and 1 (b).

Appendix 1 tabulates the BBDs which are near- trend-free. These BBDs are for three factors in 15 runs, four factors in 27 runs, five factors in 46 runs and six factors in 54 runs. The BBD for seven factors in 62 runs can be found in the link given in the next Section. These BBDs have the MEs being orthogonal to the time trends and have CCPs very similar to the one in Figure 1(a). Let TF (time factor) be the fraction computed in (5) for these designs. The TF values of these five designs are 0.91, 0.959, 0.986, 0.974 and 0.976 respectively.

Appendix 2 tabulates the BBDs which are near-orthogonally blocked. These BBDs are for three factors in 16 runs arranged in two rows and two columns, four factors in 28 runs arranged in two rows and two columns, five factors in 48 runs arranged in two rows and three columns, and six factors in 54 runs arranged in two rows in three columns. The BBD for seven factors in 60 runs arranged in two rows and three columns can be found in the link given in the next Section. With the exception of the first BBD, these BBDs have both the MEs and 2FIs being orthogonal to the block effects and have CCPs very similar to the one in Figure 1(b). Let BF (blocking factor) be the fraction computed in (5) for these designs. The BF values of these five designs are 0.944, 1, 0.992, 0.927 and 0.962 respectively. Note that for the 4-factor BBD, all effects are orthogonally blocked.

Advertisement

5. Conclusion

This paper describes an algorithmic approach to arrange the runs of an experimental design in general and an RSD in particular so that it is robust against nuisance factors such as time trend and blocking factors. Designs constructed by this approach can supplement the existing catalogue of designs in the literature. Although 3-level designs are used in this paper to illustrate our blocking approach, ours can also be used with 2-level designs (the factorial and fractional factorial designs) or mixed-level designs [13, 14, 15] or mixture designs [16].

The link of the supplemental material which contains the Java implementation of the algorithm in Section 3 and additional examples is: https://drive.google.com/drive/folders/14g7E3I4F8KIL2rcZMovlvJmsaM7iC7pJ?usp=sharing.

Advertisement

-11000-11-10-100-11000000
-0.8570.571-10-1-0.9560.86710-100-0.9620.887-11000-1
-0.7140.209011-0.9110.73900011-0.9250.778110001
-0.571-0.0880-11-0.8670.61800000-0.8870.6730001-11
-0.429-0.31910-1-0.8220.5030100-1-0.8490.5730-1001-1
-0.286-0.4841-10-0.7780.3940-10-10-0.8110.478100-110
-0.143-0.582110-0.7330.291-10100-0.7740.3860-1-1-100
0-0.615000-0.6890.1941000-1-0.7360.299-100-110
0.143-0.582-1-10-0.6440.10300000-0.6980.2160001-1-1
0.286-0.484-110-0.60.01800110-0.660.13810-10-10
0.429-0.319-101-0.556-0.06100101-0.6230.064-1010-10
0.571-0.08801-1-0.511-0.1331-1000-0.585-0.006000000
0.7140.2090-1-1-0.467-0.2000-11-0.547-0.0710-10011
0.8570.571101-0.422-0.26100000-0.509-0.1321-10100
11000-0.378-0.3150-100-1-0.472-0.18901-10-10
-0.333-0.36401100-0.434-0.2411-11000
-0.289-0.406-100-10-0.396-0.289-100-101
-110-1-10-0.244-0.44211000-0.358-0.332-100110
-0.9230.7691100-0.2-0.473-11000-0.321-0.372011010
-0.8460.557100-1-0.156-0.497001-10-0.283-0.4060-100-1-1
-0.7690.3630011-0.111-0.51500000-0.245-0.437-110001
-0.6920.188-1100-0.067-0.5270-1010-0.208-0.463011-100
-0.6150.031-1010-0.022-0.53300-110-0.17-0.48500110-1
-0.538-0.10800-110.022-0.53300-10-1-0.132-0.50211000-1
-0.462-0.228001-10.067-0.52701-100-0.094-0.515-1-10100
-0.385-0.32900000.111-0.51500000-0.057-0.524000000
-0.308-0.412-1-1000.156-0.49710001-0.019-0.52800-1011
-0.231-0.477-10-100.2-0.473-100010.019-0.528-10-10-10
-0.154-0.52300000.244-0.4420010-10.057-0.524001101
-0.077-0.5510-10-10.289-0.406-1-10000.094-0.515110100
0-0.560-1010.333-0.36400-1-100.132-0.502100-101
0.077-0.55110010.378-0.31500-1010.17-0.485000000
0.154-0.523010-10.422-0.261000-1-10.208-0.463000000
0.231-0.47701100.467-0.2010100.245-0.437100-10-1
0.308-0.4121-1000.511-0.133-100100.283-0.406-100-10-1
0.385-0.32910-100.556-0.0610-1-1000.321-0.3720110-10
0.462-0.22801-100.60.018100100.358-0.3320-11-100
0.538-0.10801010.6440.103101000.396-0.28900-101-1
0.6150.031-100-10.6890.1940001-10.434-0.2410-100-11
0.6920.18800000.7330.291100-100.472-0.18900101-1
0.7690.36310100.7780.3940-10010.509-0.13200-110-1
0.8460.5570-1100.8220.5030-11000.547-0.07100-1101
0.9230.76900-1-10.8670.618010-100.585-0.0061-1-1000
11-10010.9110.739010010.6230.064-1-1-1000
0.9560.867000000.660.138000-1-11
11-1000-10.6980.21601-1010
0.7360.29901-1-100
0.7740.386001011
0.8110.478000-1-1-1
0.8490.573000000
0.8870.673100110
0.9250.778-110100
0.9620.8871010-10
11-1-11000

11-1-1011-1001011001011
1110-11100101111-1-1000
11000110-100-111110001
1101111000001111000-1
12-101110010-11100101-1
1201-11100000111010-10
121-1011010-1011000000
120001110-100111-11000
210-11120100-111-1010-10
21-10-11200000120110-10
210001200-1-10120-100-1-1
21110121000-11201-1-100
22101120-10-10120-1-1-100
220001200-110120-100-11
220-1-1120000012-1-10100
22-110121000112000000
13-10-1001201-10-10
130-110012-110100
1100-1-1130001113-100-101
11-100-113-1-10001301-1010
11110013000-111300-1101
11-1001131001013-100-10-1
110-1101301100130-1001-1
111-10013-110001300-110-1
110101211010013011010
1201102101-10013000000
1210-10210-1001130-10011
120000210000021-1-1-1000
12001-1210000021000000
120-10121010102100-1011
120000210-1-1002100-101-1
12-10-1021-100-1021-10-10-10
211010220000021-110001
21-101022-1000-121-1-11000
2100-1122-1000121-11000-1
210-1-1022000002110-10-10
21010-1220-101022100110
2100002201001220-11-100
21000022001-1022-100110
220-10-1220011022110100
2201-102300-10-122000000
22100-12300-101221-10100
22-1100231-100022011-100
22100123-1010022-100-110
22-1-10023100-1022100-110
22001123000-1-123100-101
2311000230001-11
230001-123001101
2300110-1
23000-1-11
23000000
230001-1-1
23000-1-1-1
23100-10-1

References

  1. 1. Box, G.E.P., Hunter, J.S. & Hunter, W.G. (2005) Statistics for Experiments 2nd ed., New York: Wiley.
  2. 2. Box, G.E.P., & Behnken, D. (1960) Some new three level designs for the study of quantitative variables. Technometrics, 2, 455–475.
  3. 3. Box, G.E.P. & Hunter, J.S. (1957) Multi-factor experimental designs for exploring response surfaces, The Annals of Mathematical Statistics, 28, 159-241.
  4. 4. Khuri, A.I. & Cornell, J.A. (1996) Response Surfaces, Designs and Analyses. 2nd ed. (New York, Marcel Dekker).
  5. 5. Fischer, E. (1908) Über den Hadamardschen Determinantensatz. Archiv Der Mathematik und Physik, 13, 32-40.
  6. 6. Nguyen, N-K (2001) Cutting Experimental Designs into Blocks. Austral. & New Zealand J. of Statistics 43, 367-374.
  7. 7. Nguyen, N-K. (2014) Making Experimental Designs Robust Against Time Trend. Statistics & Applications, 11, 79-86.
  8. 8. Atkinson, A.C. & Donev, A.N. (1996) Experimental designs optimally balanced for trend. Technometrics 38, 333-341.
  9. 9. Cook, R.D. & Nachtsheim, C.J. (1989) Computer-aided blocking of factorial and response surface designs. Technometrics 31, 339-346.
  10. 10. Gilmour, S.G. & Trinca L.A. (2003) Row-Column Response Surface Designs. Journal of Quality Technology 35, 184-193.
  11. 11. Jones, B. & Nachtsheim, C. J. (2016) Blocking Schemes for Definitive Screening Designs, Technometrics 58, 74-83.
  12. 12. Jones, B. and Nachtsheim, C. J. (2011). A Class of Three Levels Designs for Definitive Screening in the Presence of Second-Order Effects. Journal of Quality Technology 43, 1-15.
  13. 13. Jones, B. & Nachtsheim, C. J. (2013) Definitive Screening Designs with Added Two-Level Categorical Factors. Journal of Quality Technology 45, 121-129.
  14. 14. Nguyen, N-K., Pham, D-T. and Vuong, M.P. (2020) Constructing D-Efficient Mixed-Level Foldover Designs Using Hadamard Matrices. Technometrics, 62, 48-56,
  15. 15. Nguyen, N-K., Kenett, R.S., Pham, D-T. and Vuong, M.P. (2021) D-efficient Mixed-Level Foldover Designs for Screening Experiments. To appear in Springer Handbook of Engineering Statistics, 2nd ed, Hoang Pham (Editor).
  16. 16. Cornell, J.A. (2002) Experiments with Mixtures: Designs, Models and the Analysis of Mixture data. 3rd ed. New York: John Wiley & Sons,Inc.

Written By

Nam-Ky Nguyen, Mai Phuong Vuong and Tung-Dinh Pham

Submitted: 11 October 2020 Reviewed: 05 March 2021 Published: 10 November 2021