Kode develops for SCAM an AI-based solution and supports eco-sustainability

SCAM‘s Organo-Mineral Fertilizers have been developed over the years with meticulous calibrations, aiming to create solutions suitable for sustainable and green agriculture. These fertilizers are designed to protect the environment for future generations while ensuring high final crop yields, safeguarding industry workers, and end-users.

For this reason, careful monitoring of every product characteristic is essential to ensure product quality and adherence to the features that make them unique in the industry.

For this purpose, SCAM has decided to implement a system for detecting the moisture content of the product, a key factor in its proper granule agglomeration, which is essential for its effectiveness. The system, based on NIR probes, collects spectroscopic readings at two critical points in the production plant to monitor this fundamental aspect both during the process and in the final stage before commercialization (just before packaging). This ensures the quality of every single bag released to the market.

For the development of the entire system, Kode created two predictive models based on a thorough analysis and historization phase of the data, along with a significant spectrum calibration process:

• A Predictive Chemometric model, capable of capturing the spectral fingerprint and converting it into an accurate quantification and qualification of the sample’s moisture level (with extremely high precision, as the moisture level cannot deviate by more than 0.3% to maintain the expected quality standards).

• An Applicability Domain model for each product type, which, through the identification of outliers, enabled the extraction of a validity score for each acquired spectrum. Based on this model, an alerting system was developed to distinguish end-of-production from drifts and production issues.

Since the goal of this project is not only to acquire accurate and well-calibrated data from the probes but also to enable operators to take concrete action on the production line, the system for analyzing the spectral signals acquired by the probes has been integrated into custom software. This software was designed with all the operational features necessary for SCAM’s daily fertilizer production activities.

Based on the core version of SpectralizeR, this is a web application that can be accessed from any computer-equipped workstation. It includes features specifically designed for line operators (such as Historical data analysis, Real-time in-line monitoring of moisture content and a drift prevention system.)

Additionally, for Laboratory Chemists, the system provides a dedicated section that allows them to manage already developed calibration models (by selecting the specific model for each product type), launch new training phases on existing models and even create new calibration models.

Marco Calderisi, CEO of Kode, comments: “We are truly satisfied to have integrated our solutions in the field of NIR spectroscopy into SCAM’s production process because every improvement made to a production process brings sustainability benefits, both economic and environmental. Artificial Intelligence, nowadays, when used to create real tools, is by no means an investment for its own sake.”

Federico Tonelli, SCAM’s Production Director, adds: “The collaboration with Kode is part of a path undertaken over the years by the R&D department aimed at improving the quality of granular products, in line with the specific requirements of innovative soil spreading equipment and the aspects of agronomic and environmental sustainability.”