Janet Kaplan

Cooling Water Function

__________________________________________________________________________________________________________________

In this investigation we are given data involving the temperature of boiling water as it cools, ultimately, to room temperature. Starting at 212 degrees Fahrenheit, the temperature of the water is taken every minute for 30 minutes.

The data is graphed below, with the independent variable as time in minutes and the dependent variable as temperature.

Given this data, how can we predict what the water's temperature will be in 45 minutes, or 60 minutes, or even 300 minutes?

____________________________________________________________________________________________________

Our first task will be to determine specifically what the cooling function is. This is real life, however, and chances are the data will not fit neatly into a perfectly smooth line or curve. We must, therefore, determine a best-fit curve for our data.

How do we begin to do this?

Take a look at the shape of the curve. What type of function best models our data? It doesn't look linear, as it is not a straight line. It's not a quadratic function, for it has no resemblance to a parabola. Nor is it any higher level polynomial, again because it has no similarity to a cubic or higher degree equation.

It's clearly not a trigonometric function, for there is no periodicity. There are a few other families of functions that also do not apply. But what our data does resemble is an exponential function. More specifically, a decreasing exponential function, often known as a decay curve.

Notice that it starts off at 212 degrees and decreases fairly rapidly at first. As time moves on, the water temperature is still decreasing, but at a somewhat slower rate. Although it's hard to see on the graph, with each minute the reduction in temperature progresses at a slower rate. At 30 minutes it is down to 130 degrees. We can surmise that the water will continue to cool, and that ultimately it will come to rest at room temperature (let's say 70 degrees).

So we will begin by trying to find the best-fit exponential function for our data.

There's a catch, however. Exponential functions increase or decrease without bounds (to infinity). But experience tells us that the water will maintain the temperature of the air around it indefinitely, not decrease to absolute zero. So we will need to modify our model to incorporate this fact.

____________________________________________________________________________________________________

At this point we can approach the problem from one of two ways. We can use a spreadsheet application, like Excel, to show us the results of otherwise time-consuming calculations quickly and easily. Or we can graph a general exponential model using Graphing Calculator software. Often it is helpful to use both avenues simultaneously.

The general form of a decreasing exponential function is y = ae-bx, where a is the starting value, e is the natural base, (-b) is the rate of decrease, and x is the time in minutes. In our example, a will be the starting temperature, and (-b) will be the cooling rate.

Exponential functions increase or decrease by the same multiplicative factor each time period. That is, we multiply or divide by b for each unit change in x. My first model is saying that starting at 212 degrees, the temperature will decrease from the previous temperature by e -0.02 every minute.

Here I have used Excel to list the actual data along with temperature calculations using My First Model, y=212e^-0.02x.

 

Actual

 

 

y=212e^-0.02x

 

 

 

 

 

 

 

Actual

 

 

My First

 

Minute

Temp

diff

 

Model

sq of diff

0

212

0

 

212.00

0.00

1

205

7

 

207.80

7.85

2

201

4

 

203.69

7.22

3

193

8

 

199.65

44.28

4

189

4

 

195.70

44.90

5

184

5

 

191.83

61.24

6

181

3

 

188.03

49.38

7

178

3

 

184.30

39.74

8

172

6

 

180.65

74.90

9

170

2

 

177.08

50.09

10

167

3

 

173.57

43.18

11

163

4

 

170.13

50.89

12

161

2

 

166.77

33.24

13

159

2

 

163.46

19.92

14

155

4

 

160.23

27.31

15

153

2

 

157.05

16.43

16

152

1

 

153.94

3.78

17

150

2

 

150.90

0.80

18

149

1

 

147.91

1.19

19

147

2

 

144.98

4.09

20

145

2

 

142.11

8.36

21

143

2

 

139.29

13.73

22

141

2

 

136.54

19.93

23

140

1

 

133.83

38.04

24

139

1

 

131.18

61.12

25

137

2

 

128.58

70.82

26

135

2

 

126.04

80.31

27

133

2

 

123.54

89.44

28

132

1

 

121.10

118.89

29

131

1

 

118.70

151.33

30

130

1

 

116.35

186.37

 

 

 

 

 

 

 

avg diff

2.73

 

 

47.29

I have also calculated the deviation between my model and the actual data using the least squares method. 47.29 is higher than we would like; we want this value to be as close to zero as possible.

Here is a graph of the data, comparing the actual data to my first model. How close did I come?

 

 Notice that my model decreases too gradually at first, and then too quickly later on. I’ll need to do some refinement.

This is also a good time to think about including the limiting temperature in my model. How do I build this in?

Newton’s Law of Cooling states that the final temperature of an object that is warmer than the air around it can be determined by the formula Tf = Tr + (T0Tr)e-rt, where Tf is the final temperature after t minutes, Tr is the temperature of the surrounding air, T0 is the original temperature of the object, and r is the rate at which the object is cooling. Newton’s law of cooling is an example of a modified decay curve.

We can state the same thing this way: y = c + ae-bx , where c is the minimum temperature (70), a represents the range of temperatures in the function (212-70 or 142), (-b) is the cooling rate, and x is the time in minutes.

We’ll use Graphing Calculator now to graph this function. CLICK HERE to manipulate the function yourself.

y = 70 + 142e-0.325x

You may wonder how I knew what value of b to use. Nothing beats trial and error, and nothing beats technology’s capacity to help in this matter. Starting with b = -.1, I just kept using different values in Graphing Calculator until I came as close as comfortable to the curve I was looking for. I then entered my figures into Excel and calculated the deviation again using the least squares method.

 

Actual

 

y=70 + 142e^-.0325x

 

 

 

 

 

 

 

Actual

 

GCF

 

 

Minute

Temp

 

(b=-.0325)

sq of diff

 

0

212

 

212.00

0.00

 

1

205

 

 

 

 

2

201

 

 

 

 

3

193

 

 

 

 

4

189

 

 

 

 

5

184

 

190.70

44.89

 

6

181

 

 

 

 

7

178

 

 

 

 

8

172

 

 

 

 

9

170

 

 

 

 

10

167

 

172.56

30.91

 

11

163

 

 

 

 

12

161

 

 

 

 

13

159

 

 

 

 

14

155

 

 

 

 

15

153

 

157.21

17.72

 

16

152

 

 

 

 

17

150

 

 

 

 

18

149

 

 

 

 

19

147

 

 

 

 

20

145

 

144.13

0.76

 

21

143

 

 

 

 

22

141

 

 

 

 

23

140

 

 

 

 

24

139

 

 

 

 

25

137

 

133.01

15.92

 

26

135

 

 

 

 

27

133

 

 

 

 

28

132

 

 

 

 

29

131

 

 

 

 

30

130

 

123.56

41.47

 

 

 

 

 

 

 

 

avg diff

 

 

25.28

 

At a deviation value of 25.28, I came very close to a function that models the actual cooling data.

I then use this model, y = 70 + 142e-0.325x, to predict the water temperature at 45, 60 and 300 minutes. The values are, respectively,

45 minutes            102.90 degrees                                        

60 minutes              90.20 degrees              

300 minutes            70.01 degrees                      

RETURN to Janet’s Home page