Design and Analysis
of Experiments
Chapter 9
Response Surface Methods
(Phương pháp bề mặt đáp ứng)
Dr. Tran Thanh Hung
Department of Automation Technology,
College of Engineering, Can Tho University
Email:
[email protected] Chapter objectives
At the end of this chapter, students can
know:
• Response Surface Method and its sequential
nature for optimizing a process
• First order and second order response surface
models and how to find the direction of
steepest ascent to maximize the response
• How to deal with several responses
simultaneously (Multiple Response
Optimization)
Response surface Method:
Introduction
• Consider a chemical process: An engineer wishes to find
the levels of temperature (x1) and pressure (x2) that
maximize the yield (y) of the process:
y f x1 , x2
response surface
Response Surface Method:
Introduction
y f x1 , x2
• Problem: f is unknown estimate the system model
k
First-order model: y 0 i xi
i 1
k k k k
Second-order model: y 0 i xi ij xi x j ii xi2
i 1 i 1 j i i 1
• Response surface method, or RSM, is a collection of
mathematical and statistical techniques useful for the
modeling and analysis of problems in which a response
of interest is influenced by several variables and the
objective is to optimize this response.
RSM is a Sequential Procedure
• Factor
screening
• Finding the
region of the
optimum
• Modeling &
Optimization
of the response
5
Response Surface Models
Models that we will use in each stage:
• Screening
y 0 1 x1 2 x2 12 x1 x2
• Steepest ascent (Find optimum region)
y 0 1 x1 2 x2
• Optimization
y 0 1 x1 2 x2 12 x1 x2 x x
2
11 1
2
22 2
6
The Method of Steepest Ascent
A procedure for moving
sequentially from an
initial “guess” towards
to region of the
optimum
Based on the fitted
first-order model
ŷ ˆ0 ˆ1 x1 ˆ2 x2
Remember, in the first place we don't know if the 'hill' even exists so7
we will start somewhere where we think the optimum exists.
Example 11.1: a chemical process
Determining the operating conditions that maximize the yield of
the process. 2 factors: reaction time and reaction temperature.
Currently operating:
- reaction time: 35’
- temperature: 1550F
Yield ~40%: may
not be the best.
Use DOE to find the
optimum.
Start around the
current point.
yˆ 40.44 0.775 x1 0.325 x2
8
Example 11.1: determine the path
2.000
yield An approximate step size
and path can be
determined graphically
1.250
Formal methods can also
be used
B : Te m p
0.500
40.5 41 41.5
40 5
yˆ 40.44 0.775 x1 0.325 x2
-0.250
39.5
slope of steepest ascent path
-1.000
-1.000 -0.250 0.500 1.250 2.000 = 0.325/0.775
A: Reaction Tim e 9
Example 11.1: steepest ascent
The step size is 5 minutes of reaction time and ~20F.
The engineer computes points along this path and observes
the yields at these points until a decrease in response is noted
10
Example 11.1: optimum region
• Increases in response
are observed through the
first tenth steps
• All steps beyond this point
result in a decrease in
yield.
Another first-order model
should be fit in the general
vicinity of the point
11
(85’,1750F).
Example 11.1: second first-order
model
yˆ 78.97 x1 0.5 x2
Example 11.1: check the second
first-order model
The first-order model is not an adequate approximation.
The curvature in the true surface may indicate that we are
near the optimum.
At this point, additional analysis must be done to locate
the optimum more precisely.
General algorithm for determining the
coordinates of a point on the path of steepest
ascent
k
Fitted first-order model: yˆ ˆ0 ˆi xi .
i 1
Assume that the point x1 = x2 =...= xk = 0 is the base or origin point.
1. Choose a step size in one of the process variables, x j .
Usually, we would select the variable we know the most
about, or we would select the variable that has the
largest absolute regression coefficient ˆi .
2. The step size in the other variables is
ˆi
xi , i 1, 2,3,..., k i j
ˆ / x j j
3. Convert the xi from coded variables to the natural
variables.
Analysis of a Second-Order
Response Surface
When the experimenter is relatively close to the optimum,
a model that incorporates curvature is usually required to
approximate the response. In most cases, the second-
order model k k k k
y 0 i xi ij xi x j ii xi2
i 1 i 1 j i i 1
is adequate.
• These models are used widely in practice
• Fitting the model is easy, some nice designs are available
• Optimization is easy
• There is a lot of empirical evidence that they work very well
15
Location of the Stationary Point
16
Location of the Stationary Point
17
Location of the Stationary Point
18
Example 11.2: chemical process in Example 11.1
19
This is a central composite design (CCD)
Example 11.2: chemical process in Example 11.1
20
Example 11.2
21
Example 11.2
yˆ 79.94 0.99 x1 0.52 x2 0.25 x1 x2 1.38 x12 x22 22
Contour Plots for Example 11.2
The contour plot is given in the natural variables
The optimum is at about 86.93 minutes and 176.54 degrees 23
Multiple Responses
• Example 11-2 illustrated three response
variables (yield, viscosity and molecular
weight)
• Multiple responses are common in practice
• Typically, we want to simultaneously optimize
all responses, or find a set of conditions
where certain product properties are
achieved (or at least keeps them in desired
ranges)
• A simple approach is to model all responses
and overlay the contour plots
• See Section 11-3.4, pp. 496. 24
Multiple Responses
Multiple Responses
Multiple Responses
Multiple Responses
Target:
Yield : y1 78.5
Viscosity: 62 y 68
2
Molecular weight Mn:
y3 3400
Thực hành chương 9
Tối ưu hóa quy trình sản xuất:
• Trong một nhà máy, quy trình sản xuất hiện tại được chạy
với yếu tố A = 2.2, yếu tố B =3, sản lượng trung bình 10.5
• Chọn điểm hoạt động hiện tại làm điểm trung tâm, thực hiện
thí nghiệm 22 với 4 điểm trung tâm, ước lượng quan hệ vào
ra
• Xác định hướng đi lên đỉnh, thực hiện các bước đi lên đỉnh
theo phương pháp steepest ascent cho đến khi phát hiện
vùng đỉnh
• Chọn vị trí có đáp ứng cao nhất làm điểm trung tâm, thực
hiện thí nghiệm CCD
• Ước lượng mô hình bậc 2 của hệ thống mô tả vùng đỉnh, tìm
điểm hoạt động tối ưu nhất. Tính hiệu quả của quy trình sản
xuất mới so với quy trình cũ.
Thực hành chương 9
Lưu ý:
Dùng file Simulation for optimizing process.xlsx để mô
phỏng hệ thống sản xuất (nhập giá trị cho 2 yếu tố A, B,
ghi nhận đáp ứng)
Hướng dẫn:
- Tạo thiết kế CCD: Dùng menu Stat\DOE\Response
Surface\Create Response Surface Desgin. Chọn Central
composite
- Phân tích kết quả: Dùng Analyze Response Surface
Design
30
