April 2021

The Artificial Intelligence for maintenance 4.0: The Unsupervised machine learning applied to Maintenance schedule optimization based on equipment Remaining useful life prediction.

Abstract: The Unsupervised Machine Learning aims to define a pattern in the set of data without previous knowledge of data features. Therefore, the first understanding of your data set can start by applying the Unsupervised Machine Learning methods to understand the how your dataset can be organized and if there´s a pattern of such dataset based on their independent variables. The concepts behind Unsupervised Machine Learning cluster a set of data without knowing previous classification or any information about the data. In order to cluster the data, the Unsupervised Machine Learning models the dataset and try to organize it in a cluster. The further step verifies the result based on error and finally, If the result is satisfactory, the new dataset can be used the model defined based on the previous dataset and then the model is validated.

Concerning the maintenance engineering, the type of data related to equipment encompasses physical characteristics as well as performance, cost of operations, cost of preventive maintenance, corrective maintenance, spare parts cost and RUL. Therefore, by defining some of such variables, it´s possible to group equipment with similar characteristics. In fact, the data organization based on its common features is the concept behind the clustering data that is the result of such unsupervised machine learning methods. This paper aims to demostrate the USML applied to maintenance planning optimization based on the RUL input from different equipment. 

Read the paper here and download
The Artificial Intelligence for maintena
Adobe Acrobat Document 515.1 KB

June 2021

Infrastructures Physical Assets high performance achievement: Reliability and Maintenance Program, A.I and Asset Integrity Management.

Abstract: Nowadays the world invests around $2.5 trillion a year in infrastructure physical assets such as transportation, power, water, and telecom systems on which businesses and populations depend. Yet this amount continues to fall short of the world’s ever-expanding needs, which results in lower economic growth and (MGI’s 2013 report Infrastructure productivity) In addition to being a key enabler of investment, infrastructure can also be a significant and lucrative recipient of investment inflows. Private investment in infrastructure networks, alongside or in place of state-owned operators, has been on the rise worldwide for several decades. However, since the global financial crisis, this momentum has faltered somewhat: in 2010-2012 (OECD, 2013a)..

Indeed, the economic crisis in 2008 already led to enormous spending cuts across the globe. In Europe, the post-war infrastructure, especially bridges, is ageing. Despite that, the maintenance backlog, i.e. the amount of maintenance and rehabilitation that should have been completed in order to maintain roads in a good condition but has been deferred, is growing considerably. This problem could is being amplified because of the COVID-19 Pandemic, that cause further service cancellation, delays and consequently spending cuts.

The essential part of the investment in infrastructure physical assets is on maintenance activities. Therefore, the delay or lack of maintenance results in wider costs: it impedes mobility in the internal market, increases the risk of accidents and leads to higher CO2 emissions through the transport sector. Furthermore, “savings” from delaying or lack of maintenance will be a false economy as the infrastructure will degrade to the point where it must be replaced, which is more costly than regular maintenance and increase the risk of serious and major accident. (European Construction Industry Federation (FIEC) – 2021).  

Download the paper here
Infrastrucutre Maintenance paper 3.pdf
Adobe Acrobat Document 846.6 KB

September 2021

The Artificial Intelligence for maintenance 4.0: The Reinforcement Learning applied for Maintenance program