Monitoring the cleaning of food fouling in pipes using ultrasonic measurements and machine learning
Food and drink production equipment is routinely cleaned to ensure it remains hygienic and operating under optimal conditions. A limitation of existing cleaning systems is that they do not know when the fouling material has been removed so nearly always over-clean, incurring significant economic and...
| Autores: | , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2020 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:2072/530710 |
| Acceso en línea: | http://hdl.handle.net/2072/530710 |
| Access Level: | acceso abierto |
| Palabra clave: | Distributed Artificial Intelligence Artificial Intelligence & Big Data Industry |
| Sumario: | Food and drink production equipment is routinely cleaned to ensure it remains hygienic and operating under optimal conditions. A limitation of existing cleaning systems is that they do not know when the fouling material has been removed so nearly always over-clean, incurring significant economic and environmental costs. This work has studied the use of ultrasonic measurements and a range of different machine learning classification methods to monitor the fouling removal of food materials in plastic and metal cylindrical pipes. The experimental results showed that the developed techniques could predict the presence of fouling with prediction confidence as high as 100% for both plastic and metal pipes. The sensor technique performed marginally better in the plastic pipe and similar performance was found for the all of the machine learning methods studied. This work has demonstrated the potential of low-cost ultrasonic sensors to monitor and therefore optimise cleaning processes within pipes. It is discussed how new data set labelling strategies will be required for the techniques to be used effectively within production environments. |
|---|