CCE machine learning process diagram starting with identifying business goals, collecting and preparing data, handling recurring errors, training and tuning models, evaluating engineer features, monitoring the process, and deploying the model.

Machine Learning and Edge Computing

We combine 30+ years of process engineering with edge computing and ML — so plants get real-time insight and predictive maintenance without ripping out existing equipment or sending sensitive data to the cloud.

Built to work alongside existing PLCs/SCADA systems, not replace them.

Color Change Engineering puts small computers right at the machine or production line, not in the cloud, so data gets analyzed instantly, right where it's generated.

When something goes wrong, the edge system can detect it and even trigger a fix automatically, faster than an operator can respond. It also adjusts parameters such as fill levels and recipe settings on the fly to keep production within spec.

The edge systems monitor equipment in real time and catch problems before they cause downtime — such as weight/fill drift, package seal defects, or process variability — using machine learning models trained on how that specific line actually behaves.

Recipe Control per SKU - Icon of a clipboard with checkmarks and lines of text

Weight and Fill Optimization

ML algorithms continuously optimize filling mechanisms in response to variations in nut size, shape, and density across batches. Edge systems account for settling during packaging and transportation, adjusting fill levels to ensure compliance while minimizing giveaway.

Package Integrity

Vision systems inspect every seal for defects, detecting weak spots, incomplete closures, or contamination in the seal area that could compromise freshness. Thermal imaging verifies seal temperature profiles, catching issues before packages leave the line. ML algorithms learn to identify subtle seal defects that traditional systems miss.

Common Fault Correction - Balance scale or weighing scale

Recipe Control per SKU

ML and Edge programming can maintain and quickly implement each recipe input per specification. Elimination of rogue operator input reduces variability and increases predictability and consistency.

Millisecond Common Fault Correction

ML algorithms learn and execute common fault correction routines. Often, several faults occur during production runs that require only a reset to restart. ML can recognize an issue, and Edge can send the correction far faster than an operator can even recognize the problem, approach the HMI, then execute the correction.