Optimizing production planning through artificial intelligence
Case study: Artelys and Hitachi Energy
If you are a manufacturer in the zero-emission transportation sector, producing electric vehicles, charging stations or batteries, this case study may inspire you. The project carried out between Artelys and Hitachi Energy demonstrates how artificial intelligence can support planning processes in a complex industrial environment.
Reducing losses, lowering costs and improving operational efficiency is the result achieved by the Artelys and Hitachi Energy project.
Although the example focuses on the custom manufacturing of electrical transformers, the same principles remain applicable to any company wishing to better plan its production and reduce uncertainty.
Through this study, discover how artificial intelligence can optimize production planning in a complex industrial context.
Context and needs
Artelys is a company of French origin, established in Montreal since 2013, specializing in mathematical optimization and artificial intelligence applied to planning and decision-making.
Hitachi Energy operates a factory in Varennes, Quebec, that produces custom electrical transformers for power producers.
In a context of rapidly growing demand for transformers linked to the energy transition, order books are lengthening, placing increased pressure on production and delivery times. At Hitachi Energy, each product is unique and designs are evolving toward more complex transformers. Therefore, planning and estimating production effort are two major challenges for the company.
To meet these challenges, Artelys supported Hitachi Energy in implementing an innovative approach based on artificial intelligence to better estimate and plan production effort. In the long term, the project thus aims to develop a reliable predictive model to estimate production effort, improve planning and resource management, and reduce the gaps between estimates and reality. This project will help increase production capacity without requiring additional major investments beyond those already made.
Deployed artificial intelligence solution
The collaboration between Artelys and Hitachi Energy was built in several successive stages, each aimed at improving the accuracy and efficiency of production planning.
The first phase of the project focused on two critical steps: the manufacturing of coils and cores. Using historical production data, the teams designed a supervised machine learning model capable of accurately estimating the time required for each operation. The approach is based on an advanced statistical algorithm, published in 2016, chosen for its ability to establish reliable correlations between the technical parameters of a transformer and the time needed for its manufacture. Modeling in collaboration with the engineering teams made it possible to establish and validate the choice of variables to use for each step. This modeling approach, designed specifically for Hitachi Energy, relies on local and secure training of factory data, ensuring full sovereignty over the models and sensitive information, without depending on cloud giants or external language models such as ChatGPT.
Building on these initial results, the teams began the second phase, consisting of extending the model to the entire production chain. The estimates produced by the artificial intelligence were integrated into a software demonstrator, a concrete tool intended for the plant schedulers. This allowed visualization of the expected durations for each step, comparison of estimates with observed real times, and demonstration of the added value of AI to the operational teams.
This initiative was part of a collaborative approach supported by the IVADO consortium (Institute for Data Valorization), which brings together experts in research, training and knowledge transfer in artificial intelligence. This partnership, generated and stimulated by IVADO, made it possible to combine Artelys' expertise in applied mathematics and the field knowledge of Hitachi Energy's teams, thus ensuring the robustness and relevance of the developed model.
Observed results
The initial results proved particularly encouraging. The mean estimation error was reduced by nearly 40% on two key steps of the production process, reflecting a notable improvement in the model's accuracy. This methodology was then applied to the 17 production steps.
Beyond measurable gains, the project also had a structuring effect on Hitachi Energy's teams. The direct involvement of engineers from the early development phases promoted a shared understanding of how the model works and alleviated the apprehensions often associated with the use of artificial intelligence. The tool is now perceived not as a substitute, but as a true decision-support, credible, transparent and useful in daily operations.
Lessons and perspectives
The experience carried out by Artelys and Hitachi Energy highlights several key lessons. Artificial intelligence primarily establishes itself as a decision-support tool, complementing and enhancing human expertise. Its ease of use and the clarity of its results are essential to its adoption in an industrial environment. The success of the project also depends on constant collaboration between technical teams and domain experts, who helped shape a solution tailored to the real needs of the plant.
The next step of the project is to develop a model capable of planning production efforts. The objective is to integrate the prediction into a comprehensive scheduling and optimization system capable of anticipating production disruptions, whether related to resource availability, equipment maintenance or order variability. This approach aims to strengthen production resilience and flexibility, two essential levers in the context of an energy transition that requires rapidly producing solutions that support decarbonization.
A direct application for zero-emission transportation
The challenges encountered by Hitachi Energy strongly resonate with those of manufacturers working in zero-emission transportation. Like transformer production, the manufacture of electric vehicles, batteries or charging stations relies on complex processes that require rigorous planning, fine resource management and coordination among multiple stakeholders.
The approach developed by Artelys can be transposed to these contexts to optimize assembly lines, anticipate maintenance needs of charging infrastructure and strengthen the resilience of local industrial sectors in the face of global competition. In short, artificial intelligence positions itself here as a concrete lever for performance, sustainability and competitiveness for the entire zero-emission transportation sector.
AI in the service of ethics
However, artificial intelligence cannot be deployed without reflection on itsresponsible use.
Like any powerful tool, it must be used with discernment:
- respecting thedata protection ;
- ensuring thetransparency of decisions ;
- and keepingthe human at the heart of processes.
It is equally important to ensure that the use of AI is aimed at increasing collective well-being. Furthermore, considering its significant energy use, it should be used in ways that reduce its ecological impact as much as possible and only be used when it can have a real positive impact.
AI ethics is not a constraint, but aessential condition for its use.
There are resources to support transport stakeholders towards aethical, reliable and sustainable AI.
- TheMontreal Declarationis addressed to policymakers as well as to anyone, any civil society organization and any company wishing to participate in the responsible development of AI. It lists principles that serve as the directions of an ethical compass for guiding the development of artificial intelligence toward morally and socially desirable goals.
- Here is also aresource offered by the Ordre des ingénieurs du Québec. This document presents a summary of the 6 vigilance areas of the Professional Practice Guide, highlighting the recommendations of the Ordre des ingénieurs du Québec on the responsible use of artificial intelligence.
We encourage you to adopt a thoughtful and responsible approach in your use of artificial intelligence. The resources we offer below are not exhaustive; we therefore invite you to deepen your knowledge and remain attentive to best practices in ethical AI.
