*4.1. Data Generation*

The average indentation modulus and indentation hardness values are presented in Figure 2 Obviously, elasticity and hardness differed significantly between the various coated groups. A considerable influence of the sputtering power on the achieved *E*IT and *H*IT values was revealed. For example, C2, C4, C6 and C9, which were produced with a sputtering power of 2.0 kW, had indentation modulus between 13.3 and 16.4 GPa and indentation hardness between 3.7 and 5.1 GPa. In contrast, specimens Ref, C1, C5, C7 and C8 exhibited significantly lower *E*IT and *H*IT values, ranging from 3.6 to 4.9 GPa and 1.2 to 1.5 GPa, respectively. Compared to the latter, the central point represented by C3 did not indicate significantly higher elastic–plastic values. The variation of the bias voltage or the combined gas flow did not allow us to derive a distinct trend, especially concerning the standard deviation. In general, increased sputtering power could increase *E*IT and *H*IT by more than a factor of three. Accordingly, the higher coating hardness is expected to shield the substrates from adhesive and abrasive wear and also to shift the cracking towards higher stresses [28,35]. At the same time, the relatively lower indentation modulus leads to an increased ability of the coatings to sag without flowing [33]. As a result, the pressures induced by tribological loading may be reduced by increasing the contact dimensions [49]. Thus, it can be considered that the developed a-C:H coatings enable a very advantageous wear behavior [28,50].

**Figure 2.** Averaged values of indentation modulus *E*IT and indentation hardness *H*IT and standard deviation of the different a-C:H coatings (*n* = 10).

#### *4.2. Data Processing*

#### 4.2.1. Reading in and Preparing Data

After the coating characterization, the measured values were available in a standardized Excel dataset, which contains the plant parameters and the resulting coating characteristics for each sample. It could also be possible that the relevant measurements are already in a machine-readable format, for example the tribAIn ontology [51], but for our case we focused on the data handling via Excel and Python. To facilitate the import of the data into Python, the dataset had to be modified in such a way that a column-by-column import of the data was possible. Afterwards, the dataset needed to be imported into our Python program via the pandas library [52]. To facilitate further data processing, the plant parameters sputtering power, bias voltage and combined Ar and C2H2 were combined in an array of features and the coating characteristic such as the indentation hardness as a target for prediction.
