#	Pagina
attuale pagina	/open-h2020/projects/221147/results.html

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - VALKA (ARTIFICIAL INTELLIGENCE SUPPORTED CONSULTANCY SYSTEM FOR FISH PROCESSING OPTIMIZATION)

Teaser

Summary

The fish processing industry is progressively adopting innovations to increase efficiency and profitability through automation. In the last 20 years, efforts have focused on the automation of primary, secondary and further processing. However, few efforts have been done to improve the overall value chain, from the catch to the sales of processed fish.
Currently, fish processing plants receive partial information regarding the catch they will process, based on a partial analysis on board. Precise details are only known once at the plant, shortly before the processing starts. This leaves fish processors in an uncertain and risky position, where they need to negotiate orders with sellers that lead them to reach suboptimal deals. Any missing raw materials will be bought in the market, while any fish surpluses will be frozen or sold at a lower price. In addition, production and logistics need to be planned in a short time range, since the fish needs to be processed, packaged and delivered as fast as possible to preserve quality.
To control and monitor production and trade, fish processors rely on production management software developed by processing equipment manufacturers, which lacks catch reporting or forecast, market forecast and value and production optimization. Valka FishPro aims to make a breakthrough development towards the anticipation of the catch categorization and the optimization of production and logistics. FishPro is designed as an AI unit that continuously takes in information from every stage of the operation, such as the catch brought in by the fishing vessels, the market demand combined with information from the processing plant. It then uses this information flow to make the optimal decision and make recommendations for adjustments to the processing system accordingly at every juncture. Importantly, this innovation will help utilize precious fish stocks in the most efficient way possible to feed a growing population in a sustainable way.

Work performed

To ensure Valka FishPro technical feasibility, we have designed the data structure for collection on-board vessels as well as the overall architecture for the Flow Control Synchronization System to control the processing plant, the databases to document and store catch information, the Sales Decision Support Module, the Simulator System Design for Cutting covering the whole plant process and the core AI supported Consultancy Software Design to guide plant managers to make the best possible decisions. We have also defined a technical plan extended over 24 months and organized into 7 work packages and estimated the needs resources to develop the technical activities.
To ensure FishPro commercial feasibility, we have identified the potential market for FishPro, the fish processing companies that need to be addressed and their needs. A commercial plan has been carefully established, selecting key countries. The pricing strategy has been defined. Valka has built a dissemination and communication plan to raise the awareness amongst every potential customer, which includes the organization of workshops, attending tradeshows and an increased presence in specialized media. According to our preliminary IP assessment, we have freedom to operate and we have plans to explore the possibility of patenting our FishPro software solution by linking it to the processing equipment methods and systems involved in realizing the recommendations generated by FishPro.
Finally, to ensure FishPro financial feasibility, we have analyzed and forecasted our overall costs related to the production and commercialization of FishPro, and built a 4-year financial forecast.

Final results

We will offer the very first cloud-based consultancy software for fish processing that utilizes the latest developments in deep reinforced learning and neural networkâ€™s decision theory. This will allow our software to predict potential scenarios in the future given a current one to attain a complex objective (most profitable production plan) or maximize this goal along a dimension over many steps.
The impact of RapidFish includes:
- Deep analysis (e.g. weight and quality distribution) of the load before it lands and before it is processed
- Optimization of the fish processing in advance and matching of the raw material with the sales orders based on deep reinforced learning and neural networkâ€™s decision theory. This will allow our software to make predictions about potential scenarios in the future given a current one to attain a complex objective (most profitable production plan) or maximize this goal along a dimension over many steps.
- Recommendations for optimal processing of the catch and most desirable product mix in sales orders to maximize value (market, type of fillets, conditioning).
- Optimal matching of the products with the demand, thereby minimizing waste. This feedback is presented via a powerful simulation model running different real-time scenarios from which the responsible person of the production plan can check how each scenario is performing considering different features of the raw material and select the best one.
- Profitability increased by 2.5% with respect to current procedures.
In addition, we can also include the value proposition of our current RapidFish software: Full traceability, Production control, Order management, Quality inspections, Label design & printing, Process monitoring and Planning.