pmvpdd differences in Py and R.

[number9527-12]

Describe the bug
Her are the differences we have identified

• calculation of xf
• calculation of tcla

xf

In py the function is link
xf = t_cla / 50
in R is link
xf <- xn

tcla

py t_cla = taa + (35.5 - tdb) / (3.5 * icl + 0.1)
r tclA <- taa + (35.5-ta) / (3.5 * (6.45 * icl + .1))

@FedericoTartarini and marcel to review

[FedericoTartarini]

In the ISO 7730 the equations are

but in the ASHRAE 55

Consequently the equation in R for tcla seems to be incorrect.
Marcel what do you think about the xf?

[marcelschweiker]

Interesting to see the differences for XF between ISO 7730 and ASHRAE 55. I have no explanation for these.
As for tcla, in the ISO 7730 version, I have (2005 version + unpublished 2023 draft) it is

so in this case R would be correct. In any case, I would not expect too big differences in outcome as this line of code only affects the first guess for the surface temperature of the clothing, which will then be adjusted in iterations in the following loop and should result in the same value after the loop?

[FedericoTartarini]

@marcelschweiker great now I can tag you in the issues.

[FedericoTartarini]

Review

@marcelschweiker I looked into this issue again and I can say that changing the equation for t_cla does not affect the results and I could use either of the following:

    t_cla = taa + (35.5 - tdb) / (3.5 * icl + 0.1)
    t_cla = taa + (35.5 - tdb) / (3.5 * (6.45 * icl + 0.1))

on the other hand I need to use xf = t_cla / 50 otherwise the code does not enter the while loop since my iteration loop is checking that xn and xf are different

    while np.abs(xn - xf) > eps:
        xf = (xf + xn) / 2
        ....

changing the t_cla equation also does not affect the number of iterations needed to reach convergence.

Solution

Since all my pmv_ppd test are passing, if also the tests in comf are passing we can keep the code as it is and this issue. Since we have this issue that describe the problem and document that these code differences do not affect the results. ON the other hand, if the tests in R are not passing we will need to decide which formulation we should use and inform the relative standards.

Are you okay with this solution?