**3. Use of microclimatic measurements for numerical model validation**

The influence of mesh refinement on results deviation was investigated to assure mesh independent results and assess the model reliability. Simulations were carried out in condi‐ tions labelled as "at rest". The accuracy and independence of results from mesh refinement were estimated by evaluating result variation (magnitude of air velocity and temperature) at point PT01 (see Figure 1), by increasing the number of elements of the computational grid. Table 4 provides the absolute deviation and normalized variance of the values compared to those obtained using the finest mesh (M4). It can be gathered that over a certain value of mesh refinement, the relative variance is lower than 2.5% for the velocity magnitude, i.e. the most sensitive variable with respect to mesh size. Numerical results were also compared with experimental data to check and validate the numerical model. Comparisons were made in points where probes were located (PT01, PT02, PT03). An extract of comparison is given in Figure 4a-c, where experimental and numerical results obtained for operational "correct use conditions" are compared with each other. In this figure the average value of the acquired data (*EX PAV* ) and the numerical result obtained in steady state conditions (*NUM* ), were plotted together. The experimentally detected absolute deviation above (*EX PMAX* −*EX PAV* ) and below (*EX PAV* −*EX PMIN* ) the average value is presented by means of error bars. In the same diagrams, circled symbols to indicate the maximum difference between acquired data and simulations (max*<sup>i</sup>* { | *EX Pi* − *NUM* | }), and squared symbols to present the difference between experimental average data and numerical results (| *EX PAV* − *NUM* |) were used. Deviations were normal‐ ized using experimental average values as reference for temperature and velocity (Figure 4a and 4c), while the absolute values of deviation for RH (Figure 4b) was retained.


**Table 4.** Mesh accuracy tests.

For the remaining dependent variables impermeable/insulation conditions at walls were imposed. The occupants contribution for particle emission rate was applied as a boundary flux [particle/(m2 s)] at the occupants/surrounding air interfaces. The procedure adopted to assess the particulate flux depending on particle dimension is explained in the next section of the present chapter. In thermal analysis, a convective thermal flux was applied to the walls, considering a heat exchange coefficient of 7.7 (W/(m2 K)) and a constant temperature of 20 °C in the adjacent rooms. Insulation conditions were applied to the solid/fluid interfaces for all other dependent variables that were solved. For the "at rest" conditions, the lamps were considered the only internal sources of sensible heat. Otherwise, in operational "correct use" conditions, internal heat sources were related to medical staff and patient presence, as sensible and latent heat loads. Other boundary conditions were not considered to change from "correct use" to "incorrect use" operational conditions: heat and vapour source values were un‐ changed, except for the additional load due to the walking healthcare assistant, whose location was variable in accordance with the moving object position during time. The governing equations together with their boundary conditions were spatially discretized on non-struc‐ tured grids, made of second order tetrahedral elements. Steady state solutions of discrete equations were carried-out by applying an iterative dumped Newton-Raphson scheme [45] based on the discretized PDE linearization by a first-order Taylor expansion. Algebraic systems of equations coming from differential operator discretization were solved by a PARDISO package, a direct solver particularly efficient to solve unsymmetrical sparse matrixes by a LU decomposition method. The convergence criterion was set to 1E-5. Time integration of governing equations for transient simulations was performed applying an Implicit Differen‐ tial-Algebraic (IDA) solver [46], which uses variable-order and variable-step-size Backward Differentiation Formulas (BDF). Because the time-matching scheme is implicit, a nonlinear system of equations was solved at each time step. All computations were carried out on a workstation with two 64-bit 6-core/12-thread processors speeding up to 2.3 GHz of frequency

**3. Use of microclimatic measurements for numerical model validation**

The influence of mesh refinement on results deviation was investigated to assure mesh independent results and assess the model reliability. Simulations were carried out in condi‐ tions labelled as "at rest". The accuracy and independence of results from mesh refinement were estimated by evaluating result variation (magnitude of air velocity and temperature) at point PT01 (see Figure 1), by increasing the number of elements of the computational grid. Table 4 provides the absolute deviation and normalized variance of the values compared to those obtained using the finest mesh (M4). It can be gathered that over a certain value of mesh refinement, the relative variance is lower than 2.5% for the velocity magnitude, i.e. the most sensitive variable with respect to mesh size. Numerical results were also compared with experimental data to check and validate the numerical model. Comparisons were made in points where probes were located (PT01, PT02, PT03). An extract of comparison is given in Figure 4a-c, where experimental and numerical results obtained for operational "correct use conditions" are compared with each other. In this figure the average value of the acquired data

and handling 128 GB of RAM.

524 Current Air Quality Issues

For temperature (Figure 4a), the maximum normalized difference between numerical and experimental values is 12.3% in PT01, 4.6% in PT02 and 5.6% in PT03. Referring to the average experimental values, the relative gaps decrease to 6.3% (PT01), 1.7% (PT02) and 1.8% (PT03). Comparison of numerical temperature in PT01 with the experimental Tmr highlights a maxi‐ mum normalized difference of 5.6% and a normalized deviation to the average value of 2.3%. Referring to RH (Figure 4b), a maximum absolute deviation of 6.3 (PT01), 4.8 (PT02) and 4.5 (PT03) percentage points, and an absolute deviation to the average value of 1.9, 2.9 and 3.4 percentage points, respectively can be found. Analysing the air velocity value, obtained under the plenum (Figure 4c), the detected range of variation is quite high with respect to the mean value (0.23 m/s), and the simulated velocity magnitude (0.27 m/s) highlights a consistent gap. However, it should be noticed that numerical values globally stand within the range of variation of the measurements, foregrounding a satisfactory agreement with the experimental data. A good agreement can be noted between microclimatic experimental with numerical results with relative differences lower than 10%. Comparison between microclimate experi‐ mental data with average parameter values, suggested by Italian and International standards, show that [47, 48, 49, 50,51, 32, 52] the average temperature values at probe locations are within the suggested limits. This is always confirmed for the "at rest" condition, but for the "opera‐ tional" one, the acquired values exceed the recommended thresholds due to medical staff presence and movements that produce local temperature and airflow modifications. The air velocity values globally fit the standard limits, suggested for unidirectional flow.

**Figure 4.** Comparison between numerical results (NUM, black grey) and mean-average experimental data (EXP, light grey). Error bars indicate the detected experimental deviation (minimum and maximum value). Circled and squared symbols (referring to the second y-axis) indicate the maximum deviation of numerical values from experimental data, and deviation of numerical results from experimental time averaged data, respectively. Temperature (a), RH (b) and velocity (c) values are reported at locations PT01, PT02 and PT03 (as shown in Figure 1).
