Defective batches in a blend of animal feed due to malfunctioning of the dose system

RESUMEN

Una empresa especializada en nutrición animal quería confirmar si todos los lotes de ingredientes para pienso para animales eran homogéneos y similares a los lotes anteriores. A pesar de que solo había unas pocas muestras disponibles para este producto, el análisis cualitativo de Chemometric Brain pudo detectar falta de conformidad en dos lotes. La empresa revisó el proceso de fabricación y descubrió un problema en el sistema de dosificación. Retiró los lotes defectuosos, evitando posibles reclamaciones, arregló el sistema de dosis y reanudó la producción.

SUMMARY

A company specialized in animal nutrition needed to know if all manufactured batches of ingredients for animal feed were homogeneous and compliant with previous batches. Despite there being only a few samples available for this product, Chemometric Brain’s qualitative analysis could detect non-conformities in two batches. The company reviewed the manufacturing process and found an issue in the dose system. They removed the defective batches, avoiding possible claims, fixed the system and resumed production.

CONTEXT

A company specialized in animal nutrition deployed Chemometric Brain to analyze raw materials and final products, thus improving quality control. Chemometric Brain applied a qualitative analysis to detect changes in the ingredients’ composition that can come out after the manufacturing process.

A qualitative analysis uses the full spectrum available from a NIR device and treats it as a digital fingerprint. Chemometric Brain uses a number of mathematical models, such as Principal Component Analysis amongst others, to compare the fingerprint from the current batch to fingerprints from a library of ‘good’ fingerprints from previous batches or reference libraries. Chemometric Brain can then give the user a simple Tick or Cross advising if the batch is in line with the library for the product being tested.

TESTED PRODUCT

Blend of ingredients for animal feed.

ISSUE

Two batches (batch 7 and batch 8 on image 1; orange and red circle, respectively) showed non-conformities with respect to the rest of the batches manufactured with the same raw materials and on the same day.

APPROACH

Using a model with 12 conforming samples, the analysis certified that batches 1 to 6 were compliant. Image 1 shows batches 1 to 6 inside the confidence area, thus considered like the previous ones. However, batch 7 and batch 8 appear outside the confidence area, although they had been manufactured the same day, in the same facilities and using the same raw materials.

They found the problem in the batches and after revision of the process, they found an issue with the dose system which wasn’t adding the ingredients consistently so one batch had a low level of one ingredient while the next one had too much. The issue with the dosing system was fixed and production resumed.

CONCLUSION

The customer was able to identify the non-conforming batches by implementing Chemometric Brain. This example illustrates how Chemometric Brain is highly effective at identifying potentially defective batches despite the small number of samples available (12).

A simple and accurate analysis was enough to detect errors in their production line and avoid future claims, as well as extra costs in external laboratories.

‍

‍