The fourth industrial revolution, or Industry 4.0, uses technologies such as AI to optimise manufacturing processes. The implementation of AI follows the CRISP-DM standard, which emphasises key steps such as data collection. This paper improves interoperability within CRISP-DM by transforming heterogeneous time series data into a unified JSON format. This enables consistent data processing and effective AI model training. A case study demonstrates the successful application of this method to aggregate data from 27 machines into an InfluxDB.