Is it possible to perform a parameter sensitivity analysis using python?

Question

Is this possible? I have a basic equation:

Q = (pi*(Ta-Ts))/(((1/ha*Do))+(1/(2*k))*math.log(Do/Di)) * L

where; 
ha = 8.14
k = 0.0026
Do = 0.2
Di = 0.003175
L = 0.25
F = 0.0704
Ta = 293
Ts = 113
pi = 3.14159265

I want to see how some of the variables affect the final output (and build a variable sensitivity table). I've already managed this in a graph format, but would like some descriptive statistics.

For example, I want to have Do (outer diameter) as a range np.arange(0.1,2,100) and keep the other variables constant.

I have the following code for creating some plots of this:

def enthalpy_mod1(ambient_temp, LNG_temp, Flow):

    ha = 8.14
    k = 0.0026
    Do = 0.2
    Di = 0.003175
    L = 0.25
    F = Flow
    Ta = ambient_temp
    Ts = LNG_temp
    pi = 3.14159265
    Q = (pi*(Ta-Ts))/(((1/ha*Do))+(1/(2*k))*math.log(Do/Di)) * L

    e = (Q*3600)/F

    results.append(e) # append the result to the empty list
    df['Enthalpy Result']= e
    plt.plot(Flow, e)
    plt.rcParams.update({'font.size': 12})
    plt.annotate('Flow rate effects', xy =(0.1,14000))
    plt.show()
    print df

print Flow_mod(df['Temp'], df['LNG'], df['Flow'])

ambient_temp = [293,293,293,293,293,293,293,293,293,293,293,293,293,293,293,293,293,293]
Flow = np.linspace(0.04, 0.2, 18)
LNG_range = [113,113,113,113,113,113,113,113,113,113,113,113,113,113,113,113,113,113]

results = []

and put the results in a dataframe.. and plotting that way.

Answer 1

Sensitivity analysis is a methodology in itself so it should be independent of the language (of course you know that, just making a point) so you could just implement algorithms in python yourself. BUT as you asked about python, yes, people have done that. Take a look at SALib, a Python library for performing global sensitivity analyses with a variety of different methods.

The method you described moves one parameter at a time. This is a local sensitivity analysis and will not give you insights into interaction effects between variables, nor will you be able to measure non-linear effects in context. Given that your equation is quite simple, this may not matter, but this is very important in more complex models.