March 17, 2026

 

US researchers explore ways to improve feed milling with low-cost sensors and AI

 

At North Carolina State University in the United States, researchers have embarked on a new project designed to help feed mills lower costs while optimising nutrition.

 

With seed funding from the NC Plant Sciences Initiative, an interdisciplinary team is studying ways to use low-cost, real-time sensors paired with artificial intelligence to make feed milling more precise.

 

Producing high-quality animal feed relies on effectively sourcing, processing and mixing of plant-based ingredients to meet animals' nutritional needs for amino acids, energy, protein, lipids and minerals. Currently, many of North Carolina's 130-plus feed mills rely on historical data about the ingredients, adjusting mixtures from week to week or every two weeks.

 

As the project's team leader, optical sensing expert Mike Kudenov, explained, the researchers are pursuing a low-cost optical sensing system that could detect protein, energy and moisture levels in real time.

 

The data captured by the sensors would be analysed instantaneously by an AI-based solution, allowing feed mills to adjust ground ingredients from batch to batch to meet precise animal nutritional needs.

 

Such a dynamic approach to feed mixing could offer several advantages, avoiding both an economically wasteful and an environmentally taxing tendency to over-fortify the feed to ensure that it meets minimal nutritional standards, Kudenov said.

 

The project draws on diverse expertise from the colleges of Engineering and Agricultural and Life Sciences, access to sensor fabrication capabilities of the Plant Sciences Building Makerspace and direct access to a unique testing ground NC State's Feed Mill Education Unit, one of the country's few university-owned mills.

 

The team consists of:

 

    - Kudenov, a professor in the Department of Electrical and Computer Engineering;

 

    - Yuchen Liu, a Department of Computer Science assistant professor who works on AI, machine learning, digital twins and sensing;

 

    - Adam Fahrenholz, the Feed Mill's director and professor in the Prestage Department of Poultry Science;

 

    - Zach Raff, a Department of Agricultural and Resource Economics assistant professor whose specialties include the economics of technology adoption in the livestock industry;

 

    - Miguel Castillo, a Department of Crop and Soil Sciences associate professor who leads NC State's Forage and Grassland Management Program.

 

Kudenov will be designing a sensing system rugged enough to survive the dust and vibration of a working mill, while Liu will develop an AI pipeline to process the data captured by the sensors. Castillo will validate the models against lab standards, and Fahrenholz will ensure the technology integrates seamlessly into existing mill workflows. Meanwhile, Raff will delve into economic implications for real-world profitability and regulatory compliance.

 

NC State Project Launch Director Lauren Maynard is supporting the team, providing guidance and resources for project development.

 

The project was born from a collaborative effort of two forward-focused NC State efforts aimed at advancing agriculture in North Carolina: the NC PSI and the NC Food Animal Initiative.

 

The team came together at a November 2025 workshop co-hosted by the initiatives. The event – the fourth in the NC PSI's Connecting2Grow series – brought agricultural industry representatives together with 41 NC State researchers from colleges and 17 departments at the Plant Sciences Building.

 

The goal was to spark collaborations to solve state agricultural challenges linked to both crop and animal production. Five teams of researchers subsequently submitted research proposals to tackle issues raised at the workshop, and Kudenov's team was selected for US$20,000 in funding.

 

While the funding is modest, Kudenov said it could be key to getting the team set up to succeed in securing the resources needed to support the team's ambitious goals: correlating optical signatures with protein and moisture levels, using light-scattering models to refine sensor sensitivity, developing AI workflow to detect nutrient drifts and automatically adjust feed ratios and linking feed variance directly to livestock yield and operational efficiency.

 

"Our near-term goal is to put sensors in the Feed Mill," Kudenov said. "And my highest hope would be getting it into commercial feed mills in North Carolina to determine if precision formulation actually impacts the animals in a significant way, either by increasing yield or reducing the time it takes for an animal to get to a target weight."

 

- North Carolina State University