Mathematics applied to poultry farm optimization

Research performed with funding from FAPESP uses computer models to improve broiler chicken production.

By Diego Freire

Agência FAPESP – A partnership between a poultry farm in the Jundiaí region (São Paulo State, Brazil) and the Center for Research in Mathematical Sciences Applied to Industry (CeMEAI), one of the Research, Innovation and Dissemination Centers (RIDCs) funded by FAPESP, remotely monitors the conditions in which the animals are raised for slaughter and uses mathematical models to improve breeding and optimize bird growth and production.

The farm’s more than 23,000 broiler chickens are uninterruptedly monitored using computer software developed by the Federal University of Rio Grande do Sul (UFGRS), Brazil, which collects data on feed and water levels, ventilation intensity and other factors and sends the data every five minutes to researchers at CeMEAI. Mathematical models are applied using a second program to support decision making on any adjustments to these factors that may be required.

“The feed-to-meat conversion ratio, which correlates with the animal’s weight gain, is the coefficient that measures the efficiency and cost of broiler chickens, and mathematics can help achieve the desired results using models designed to simulate what happens in reality,” said José Mário Martinez, a researcher affiliated with the University of Campinas’s Mathematics, Statistics & Scientific Computation Institute (IMECC-UNICAMP) and one of CeMEAI’s principal investigators. “The software learns to recognize behavioral patterns in the farm, a complex system within which environmental conditions must be optimized so that the chickens eat well, grow at the desired rate and produce a satisfactory yield.”

Management of the farm therefore aims to stimulate the chickens to feed properly and to assure optimal functioning of their metabolism so that their genetic potential is fully developed. This involves controlling the availability of feed and fresh water, temperature, humidity, ventilation, air CO2 content, ammonia and light, among other factors.

Some sheds are partly automated, with relays (electromechanical switches) and motors governed by programmable logic controllers (PLCs). All these devices are linked to a computer containing a microprocessor that stores instructions to implement specific functions in accordance with operating rules based on temperature and humidity sensor readings.

Other activities are manual, but only one person performs all the operations required. The model developed by the researchers at CeMEAI combines multiple factors with the aim of supporting fast, comprehensive decision making.

“Decisions based on isolated factors may have side-effects that are harmful to the animals in other ways,” said Daniel Conti, also a researcher at IMECC-UNICAMP. “For example, if there’s much ammonia in the environment and the temperature outside the shed is very low, the ventilation will reduce the amount of ammonia but make the temperature cooler. Ideally, the decision should be based on the impact of each factor on the animals’ performance.”

The work conducted by the researchers at CeMEAI consists of developing a mathematical decision-making model that considers all the ideal conditions for broiler chickens, cross-referencing different quality indicators and avoiding results with side-effects that are harmful to the animals and adversely affect production.

According to the Brazilian Animal Protein Association (ABPA), demand for chicken in Brazil is growing and will rise from 43 kg per person to 45 kg per person by the end of this year. Exports are also expected to rise by 2%-3%. Brazil is the world’s third-largest producer and exporter of chicken.

“Even if poultry farms produce the same lineages, which eat the same feed and are raised in identical environmental conditions, differences in management can lead the farms to display significant variations in feed-to-meat conversion rates,” Conti said. “Better management practices can reduce the amount of feed consumed per kg of live bird by 0.1 kg.”

Thermal comfort

The sensors that monitor the animals, which on average live in the aviary for 45 days until they are transported to the packing plant, are spread over an area measuring 15 m in width and 150 m in length.

In addition to temperature, humidity, air speed, ventilation and other data transmitted by the sensors, the mathematical model developed by CeMEAI analyzes videos transmitted in real time to evaluate the animals’ thermal comfort.

Using parameters fed into the system, the algorithms are capable of distinguishing between the animals and other elements in the environment such as flooring or human beings as well as identifying situations that point to anomalies in the animals’ behavior.

“If the animals are crowding together, for example, this may suggest they’re feeling cold and the temperature needs to be raised,” Conti said. “If they insist on staying very close to the drinking troughs, they may be too warm, so the temperature should be lowered.”

Every five minutes the system calculates thermal comfort and matches this with the temperature, humidity, air speed and other readings from the sensors to identify and classify behavioral patterns. By means of mathematical models called clustering algorithms, derived from a data mining technique that automatically groups certain values together, it combines all these values to determine whether the animals are too warm or too cool, whether they are eating or drinking too much, and so on.

“If the animals’ behavior is ideal, their biological response is to grow, gain weight significantly and thereby improve production with quality, which is the system’s final objective,” Conti said.

CeMEAI already has a collection of values that represent situations identifying thermal comfort in the aviary and disruptions to routine operations such as power outages or the presence of intruders. The system automatically detects these events that adversely influence the animals’ behavior. The next step is to finalize the development of an open-source online system that automatically controls thermal comfort in the aviary every five minutes.