Automation of dairy industry: practical experience of Andersen

May 10, 2017
Blog

Agriculture is a rather conservative branch, and dairy farming is not an exception. Andersen was lucky to undertake one of the first big dairy farming automation projects in the Ukraine. Finally with the help of analyzing and processing of large data volumes impressive results were achieved.

Problem

International company bought a large agricultural enterprise for its production capacities. This enterprise needed fundamental changes: the soviet in its essence enterprise didn’t provide the appropriate for the international holding quality level and cost price. Besides, the EU export shipments of milk, due to lack of data, had to stay at customs for many hours, what significantly reduced its storage life.

The potential for optimizing on the livestock company is usually quite serious, but due to lack of data on the project start we can only guess at. Our specialists together with these company experts established the directions that should be adopted:

  • Quality and quantity control of feeding.
  • Quality and quantity control of milk given.
  • Analysis, representing and transfer of data received.

Feeding

Despite the detailed feeding norms it is believed that in animal farming of CIS countries 20-30% more animal feed is consumed. It is connected with many factors, including averaging of feeding, not optimal contents and other factors. In addition, if the animal feed doesn’t correspond to quality and quantity animals’ needs, and the feeding time is not optimal, it significantly reduces the milk yields. The equally fatal impact has got not regular feedings, what is the reason of stress for the animals.

The quality and quantity feeding sensors allowed to control the needs and when required to change the time, quantity and ingredients of feeding. What is more, special sensors find the contamination of animal feed with particular fungal pathogens. The exclusion of contaminated feeds had a positive impact on animals’ health.

Milking

The quality of milking process is largely determined by content of somatic cells and pathogenic flora. All this significantly effects on milk term of storage, processing possibilities and quality standards. Thus, according to the State Standard functioning on the territory of the Ukraine, the quantity of somatic cells in the Extra class milk should not exceed 400 thsd/cm3. The same content is the limit for milk import to the EU territory. Such milk is only about 4% from the produced one on the territory of this country, similar Extra milk output was on this enterprise before the modernization.

Milk composition analysis in the process of milking allows to monitor cow’s health and correct the ingredients, quantity and time of feeding when needed. All this positively impacts on the milk yields, animals’ health and quality of milk. On the other side, the sensors and data analysis in real time allow to immediately define the problem and react to it, if the milk yield of any particular animal began to decrease.

Data collection and analysis

Collecting and analyzing data from dozens of sensors the adapted by our specialists hardware-software system allowed to provide the livestock keepers with the real picture of the situation with every particular cow.

Handling of such a huge amounts of information in real time is almost impossible without use of methods of Big Data and Artificial Intelligence. The main challenge is that our case involves all three main aspects of Big Data: the system had to require to analyze huge volumes of various data almost instantly, in other words on a rather big velocity.

Our specialists developed a solution combining two technologies, which proved their efficiency: framework Hadoop for the distributed data processing and R-Project for statistical analysis. Hadoop is a set of powerful tools that enable to build the massive parallel processing system to fast enough computing an array of different data flows simultaneously. As for R-Project it is a development environment based on the statistics-specific programming language R. The syntax of it contains a wide range of statistical and numerical methods so it is one of the most useful tools to solve such types of tasks.

These two tools allowed developing and implementing an AI that analyzes all parameters in real time and gets almost instant decisions regarding to each separate case. The capability of self-learning helps the system to improve the precision of decisions it makes, also as to quick connect, setup and later analyze any required new sensors.

Result

The possibility to analyze and present the data collected from the sensors revealed many bottlenecks in the enterprise work. Livestock keepers eliminated and minimized many of weaknesses found. Using the automated parameters control allowed reducing the number of laboratory staff members from 21 to 2 people.

Thanks to automatic processing of data got from the sensors the time of taking a decision when finding a problem was cut by half. As a result the animals were stressed less, their health became better, which was confirmed by yield growth on 28% and growth of Extra milk share from 4% to 37%.

Now the enterprise automatically sends to the customs the processed results of milk analysis before the shipment reaches the border. Due to this time of customs clearance reduced from 5.5 hours to 20 minutes.

Despite the relatively high prices of development and deployment the effect of implementation of the system exceeded the customer’s expectations. All investments spent on automation returned in 3 to 4 months.

Previous articleNext article