Process Knowledge Learning in Precision Dairy Farming



The project agriProKnow develops a novel methodology for process related Information management, which aims at significantly improving the milk production efficiency in precision dairy farming. In a particularly complex cyber-physical production system that combines people, animals and technology, the focus is on animal health and welfare modelling, monitoring, and control. The focus of the innovation is a procedure for process knowledge generation, which combines methods of stochastic analysis of sensor data, and semantic situation modeling and semantic data-warehousing. Furthermore, the use of semantic Web service technology enables creation of an open system that helps different actors in the value chain, to contribute to, and get access to new process knowledge which is continuously created and integrated. The system and the procedure will be implemented and verified using real data from several experimental farms.


Initial situation and problem to solve: Dairy farming today is characterized by the use of precision technology, such as robots, sensors and location systems, which integrated within a complex system, should help guarantee healthy herds, high quality of milk, optimized use of resources (feeding, insemination), and reduced impact on the environment. Precision farming equipment generates a lot of new process data, however to achieve efficient data-driven control of milk production three problems need to be addressed. Firstly the problem of interoperability between equipment and data of a single farm, secondly the problem of incomplete process knowledge, and thirdly the problem of collecting data from many different farms – which is essential for generation of new process knowledge. The first problem has already been addressed by several research initiatives in the Internet of Things area, which use semantics for integration of process knowledge for process monitoring and control on one single farm, however the translation of the research ideas into mature solutions is still missing. The complexity of the second problem – the incomplete process knowledge – has been pointed out by many scientific studies that show that the efficiency of milk production depends on various different factors (feed, animal health, weather) in a way which is still not being completely understood. Only recently some process factors become quantifiable as new sensors become available. Accordingly, new process hypotheses can now be tested and new knowledge acquired. For this purpose, data from many different farms need to be accounted for. This leads us to the third problem which needs to be solved taking into account business and security requirements and concerns of all involved stakeholders.  

Planned goals and approaches: The project agriProKnowaims at developing a methodology for dairy farming process knowledge creation, as well as its integration and actuation in a multi-farm multi-actor environment. In realisation of the system we will use information management methods for high volume data, stochastic analysis for pattern recognition, semantic modelling, semantic Data-Warehouse technology, rule-based reasoning, semantic Web-Services, ontology-based access role modelling und interoperability-management, instrumentation of experimental farms und veterinary evaluation of animals’ health in scientific sensor-supported studies.

Expected results and findings: Expected results include (1) data and service platform with flexible ontology-based data access control for all relevant actors and interoperability at interfaces between different data stores (2) the process and monitoring model (identification of all relevant states) (3) stochastic algorithms that translate sensor time series into monitored-state time series (4) semantic data warehouse for process model exploration (5) a methodology for structured process model learning as an interaction between data analysis and process exploration within the semantic data warehouse analysis (6) a concept for creation of logical rules for process control and realization of control interfaces via semantic Web- services (7) methodology and tools tested in the prototype that integrates data from several experimental farms.  


Mrs Dana Tomic PhD, Smartbow GmbH


Josephinum Research (

Institut für Stochastik, Johannes Kepler Universität, Linz (

Institut für Wirtschaftsinformatik - Data & Knowledge Engineering, Johannes Kepler Universität, Linz (

Department für Nutztiermedizin und öffentliches Gesundheitswesen in der Veterinärmedizin, Veterinärmedizinische Universität Wien (

Smartbow GmbH (

Wasserbauer GmbH, Fütterungssysteme (