At the end of June, the Fraunhofer Institute for Process Engineering and Packaging IVV and the Industrievereinigung für Lebensmitteltechnologie und Verpackung e. V. (Industrial Association for Food Technology and Packaging) organised the Future Days Chocolate Technology in Freising. Around 90 participants attended the conference, which focused heavily on artificial intelligence and digitalisation, to learn more and exchange ideas.
By Dr. Jörg Häseler
Oliver Stricker, August Storck KG, Chairman of the IVLV Chocolate Technology Working Group, and Dr Marc Lutz, Migros Industrie AG, Deputy Chairman of the Working Group, welcomed and introduced the conference from 25 to 26 June 2024, which focused intensively on the topics of digitalisation, compliance and legislation as well as raw materials. On the second day, particular attention was paid to the topic of bisphenol A, its use and the legal consequences.
Dr Lukas Oehm, Fraunhofer Institute for Process Engineering and Packaging IVV Dresden, gave "our answer to the shortage of skilled workers in the confectionery industry: digital operator assistance systems for quality monitoring and increasing efficiency". In his presentation, he presented figures showing that some employees leave the company after just one year. This is compounded by a lack of qualifications and experience, which leads to low process efficiency. He also presented a system based on image data analysis using deep learning. "imageSAM" is an assistance system for quality monitoring and uses various camera systems to recognise different classes of defects. Depending on the area of application, simple industrial cameras through to specialised cameras such as LumiHD can be used. One of the advantages is that there are instructions for the operating personnel, e.g. on how to rectify the detected deviation. These can then be called up via tablet, smartphone or PC in the control centre.
Product end control was also a concern for Ronald Krzywinski, Bi-Ber GmbH & Co. Engineering KG, was also a concern. Edge learning tools offer opportunities to prioritise customer benefits. The development of the technology is also approaching the desired goal for high throughput in front side inspection. But sometimes compromises have to be made: Production lines should be extended so that the ejection of defective products can be ensured, as the detection of defects is time-limited.
Dr Julian Senoner, EthonAI AG, addressed the topic of "Process Improvement with Causal AI". According to him, causal AI will become the standard in data-driven root cause analysis.
He added that machine learning algorithms, e.g. random forests, can deliver misleading conclusions as they do not take the line structure into account. Virtual Design of Experiments can help evaluate what-if scenarios before making changes in production.
PD Dr Tilman Sauerwald, Fraunhofer IVV, demonstrated that it is not only possible to detect aroma components in chocolate using traditional analytics such as gas chromatography and mass spectroscopy. He uses multi-sensor systems with digital evaluation to measure odour compounds, as smelling and tasting play a central role, especially for emotional decisions. The fact that the odorous substances are present in the product in an extremely wide range of concentrations is particularly challenging, meaning that sensors have to measure over several powers of ten. He explained: "The system costs for multi-gas measurement systems with AI evaluation using fingerprinting have decreased significantly. The systems are usually equipped with options for temperature-cyclical operation, so-called virtual sensor arrays."
Dr Sebastian Reimann, Delica AG, and Marc Zollinger, SQTS Swiss Quality Testing, reported on an analytical problem in the detection of foreign vegetable fat in dark chocolate, which was submitted to them by a commercial laboratory. According to an ISO method, a reference material is used that is no longer available today and does not take into account changes that should be reflected in this method. This analytical evidence should be placed on a more reliable basis, as there is a risk of further evidence of an additive. A number of questions should be clarified, e.g: Should the ISO methods be supplemented with the evaluation of the fatty acid distribution (fingerprints) and should this be included in the assessment of the sample? Should a separate, up-to-date cocoa butter line be created for each country of origin?
The topic of banning bisphenol A, which was presented by Dr Frank Heckel, LCI, was added to the programme at short notice. The EU has presented a draft delegated act that could lead to a ban on this monomer in plastic food contact materials. This would have dramatic consequences, as this substance is used in the manufacture of polycarbonate (PC) moulds. He finds it particularly incomprehensible that the EFSA and the BfR have come to assessments in which the value for the tolerable daily intake differs by a factor of 1000. In addition, two exceptions have been provided for in the draft, the justification for which could also be used for casting moulds. There is therefore an urgent need for action. One of the aims of the current efforts would be to invite all stakeholders (PC producers, manufacturers of PC casting moulds and their users) to the table in order to jointly explore possible next steps.
