Fedrigoni: far-sighted digitalisation

When Big Data becomes a concrete means for monitoring and optimising the production process

Authors: Marco Calderisi, CEO of Kode and Francesca Sbolgi, Data Scientist at Kode

Introduction: The challenge of digitalising process data

In recent years, many researches and studies highlighted how digitalisation and data-driven innovation are the main means for supporting organizations’ growth and guaranteeing a competitive advantage in the market. The consequent great attention to investments in digitalisation, reserved by the European Union in the national PNRRs, had, indeed, a singificant impact in the whole Europe, as clarified by the results of the Digital Economy and Society Index (DESI) 2022 of the European Commission.

The study shows, indeed, how far behind the level of adoption of digital technologies such as Artificial Intelligence tools and Big Data analysis still is, which are fundamental to concretely optimize processes and operations.

In particular, manufacturing industries and small and medium-sized companies show the first major inconsistency. Although 72% of manufacturing companies collect process data, 15% of these companies in Europe do not analyse them in any way and almost 40% carry out ex-post analysis only. While a lack of internal expertise is put forward as an explanation for this delay, it becomes clear that a lack of vision compounds this.

The best industry 4.0 projects that bring concrete added value to a company’s development and growth, are in fact those born from a vision. This is the case of Fedrigoni, guided by the awareness that a coherent and organised management of all the factories, through data analysis, would allow the management to truly optimise production. This thought led the company to carry out a titanic project of digital transformation of the process control system.

Through slight customizations of the FactorAI Framework by Kode, our team created for Fedrigoni a tool allowing to monitor the efficiency of the continuous machines of all Fedrigoni plants. The objective was to develop, for the company management, an operational and strategic tool to optimise the production of each specific plant and, at the same time, to compare the overall performance of each plant with the others.

As a matter of fact, the project was born from the need to digitise, historicize and standardise all data coming from different plants to increase production performance control. This project was carried out through 2 main phases:

• Data Management Module: composed of an exploratory phase of data analysis and understanding on all data sources (supported by an important pre-processing work)
• OEE Module: the implementation of algorithms capable of measuring all aspects of the key indicators required for the construction of homogeneous dashboards for all plants.

Optimisation and control: the peculiarities of Fedrigoni’s production

Fedrigoni has over 70 plants for production, cutting and distribution. In these centres, thousands of types of special papers, labels and self-adhesive materials are created. Some of these plants often vary their production having in charge many SKUs, while other plants are focused on a few specific products. As a consequence, these two types of plant show a very different ratio between production and waste, making it difficult to compare their production data.

Besides, in paper manufacturing, each machine has dimensions comparable to an entire production plant, the construction of which consequently reveals specific complexities. Each of them is therefore equipped with very different integrated process data recording systems, both in terms of scale and extraction frequency. Therefore, since Fedrigoni needed to understand the performance of the entire production at a glance by comparing some key indicators (such as the quantity produced, production times orthe waste generated), data harmonization was critical and its challenge was indeed amplified when the comparison involved machines that do not fully extract one of the required indicators.

Making data from multiple plants comparable was, therefore, a fundamental requirement, which led to the choice of using machine tags. Starting from the basic production signals (punctual and irrefutable), allowed to calculate consistent KPIs. The project, born with a few production sites as a test, quickly evolved to involve 7 plants throughout Italy and 13 different machines in just 3 years.

“Over the last 4 years, Fedrigoni undertook a significant transformation path which led the group to double its turnover, triple its Ebitda and acquire 15 companies around the world (from China to the United States). Today we produce 25,000 products distributed to over 30,000 customers in 132 countries. Such a strong acceleration in growth brings with it an enormous complexity, which we hardly would have managed without the help of Big Data platforms. The collaboration with our partner Kode allowed us to optimise the production of individual plants and, at the same time, compare the performance of different production sites with each other, favouring our process of continuous improvement”. Gionata Berna, Chief Information Officer, Fedrigoni Group

The data management module: the integration of Big Data

Each plant displays thousands of machine signals; for each one, we identified together with Fedrigoni the tags to analys in detail and verified the reception, frequency and meaning of each of them in the production process. Each tag was subjected to a process of historicization and comparison with the data exposed by the integrated analysis system of each machine.

Thanks to all these stages, we verified the availability of the necessary data to apply the algorithms calculating the required indicators for all involved plants’ performance monitoring.

The subsequent phases of nomenclature, cleaning and normalisation of these Big Data, allowed to have a single common database architecture for all the machines, offering a homogeneous vision of the entire production.

Since Fedrigoni didn’t seek for the overall vision only, the system was structured to allow in-depth analysis, based on various time aggregations and type of product, such as distinguishing between the different types of downtimes.

Furthermore the platform gives the possibility of entering specific production data, such as the SKU, the weight of the specific paper for each production, the reason for each downtime and many other aspects. These functionalities allow to add details to the insights provided by the machine empowering the operators’ understanding of reality and a wider interpretation supporting strategic choices.

The OEE module: performance control

For the effective monitoring of production process performances, we implemented and customized the FactorAI OEE module, that calculates monitoring key indicators:

• time efficiency: the percentage of productive time compared to total time (availability)
• material efficiency: the percentage of good product compared to total product (quality)
• speed efficiency: the ratio between the production speed and the theoretical speed (performance)
• OEE: summary indicator of the three previous ones, synthsizing in a single KPI the performance of the machinery under monitoring.

Besides, for Fedrigoni, we realized an ad hoc customization of this module, including a phase of aggregation and segmentation of continuous data to split them per paper roll, allowing to promptly attribute the OEE performance to each single roll produced by the machinery.

KPIs and data thus calculated were made available to end users, such as plant managers and company management, via an interactive dashboard. It provides a significant level of interaction with data, allowing plant managers to carry out analyses at different levels of temporal aggregation (with annual, monthly and daily views) as well as at product level. In addition to that, the system automatically sends to company management performance reports and monthly summaries with noteworthy information, thus providing an overall vision of the up-to-date performances.

The effects of OEE monitoring as production increases

Overall Equipment Effectiveness is an indicator that requires continuous monitoring, especially in growing companies. Production increase can lead to significant fluctuations in waste and time efficiency.

Thanks to this real-time process monitoring platform, Fedrigoni was able to monitor the production yield of its machinery and identify the optimizations to focus on, in order to support production efficiency and guarantee year-on-year production improvement.

Out of the 7 plants and 13 machines involved in the project, the performance trend showed a +1.88% year-on-year exceedance of the OEE target in the last three years and the reduction of downtimes by more than 1 percentage point with more than doubled production time.

Fedrigoni is a clear example of how productions (even already performing ones) can increasingly reduce waste and improve their results through constant analysis of process data.

When we look at big companies like Fedrigoni, we tend to consider their case as non-replicable. But Data Science products and frameworks like FactorAI have been designed and structured exactly for that: to standardise tools and good practices, that will allow businesses of the Italian manufacturing of all sizes to enter the digital transformation at affordable costs.

Today, in order to truly optimise our investments, costs and emissions, it is essential to know how to make data-driven choices in company management, by taking advantage of reliable analyses.