Customized prediction model
One of the biggest challenges within supply chains: the availability of the right parts and/or raw materials at the right time. Not just an important factor in the manufacture of soda cans, but even crucial when it comes to aircraft maintenance and emergency repairs. In the latter case, in particular the forecasting around that need for spare parts is complex. Especially when it comes to new machines where little historical data is available. Moreover, many suppliers use their own predictive models around parts life cycles that do not necessarily align with the customer's practice. An aircraft or drone that frequently flies in extreme weather conditions, for example, wears out faster than predicted and that affects its deployability and safety. Accurate, flexible prediction models that take into account environmental factors and user patterns are needed. In short, a challenge for which Artificial Intelligence is perfectly suited to make a difference.
Opportunities for optimization
There are plenty of opportunities to optimize spare parts management with AI and data analytics. Consider, for example:
- cost savings through better inventory management and reduced waste.
- proactive maintenance (predictive maintenance), making systems, materials, products, machines and/or equipment last longer.
- combining sensor and maintenance data to recognize patterns in wear and tear.
- comparing models and methods for the most accurate inventory management predictions.
An additional opportunity is to improve data quality, since often that data is fragmented or incomplete. This requires a clear data policy (identifying which data points are relevant) and AI that fills in missing information and enriches insights.
Different models for a specific solution
At KPMG we recommend not to choose just one model, but a combination of a number of different AI models. Such as a combination of logistic regression (does a part break down within a certain time, yes or no), Poisson (how often does a part break down within a certain time) and survival analysis (how long does it take until a part breaks down within a certain time). Together, these three give a complete picture of which parts are likely to be needed and when they should be replaced. It also reveals what factors affect the life span. Data quality and management play an important role here. Therefore, it is important to ensure clear governance and processes to improve data quality. This mainly involves collecting the right data, such as part numbers, maintenance history and delivery time. The result is a more accurate data model that is more secure and reliable. The combined AI models provide better predictions around spare parts and inventories. This reduces machine downtime, making the overall supply chain faster and more efficient.
Galyna Ignatenko, Senior Manager Supply Chain & Procurement, is specialized in large-scale supply chain transformations powered by data and technology. She is passionate about driving lasting change and making a significant impact for clients. Her mission is to enhance efficiency in both business and the public sector.