The impacts of Artificial Intelligence on Technical Maintenance

If we go back a few years in time, it is safe to say that the term Artificial Intelligence, or AI as it is commonly called, was for most people something distant from their daily lives and brought to mind some of the most iconic science fiction films or books. However, the current scenario is completely different. The importance in the media, the amount of solutions and tools with AI resources, with which we are inundated on a daily basis, and the adjacent potential guarantee our attention.

Although the concept emerged several decades ago (John McCarthy, 1956), it was only in recent years that the necessary technological conditions were met to make its mass use possible. First of all, the development of the internet and high-speed communications, computing advances, the emergence of smartphones, the growing supply of cloud computing, allowing widespread access to supercomputers at an affordable cost, the development of programming software and the algorithms applied in AI and, no less relevant, the availability and access to large volumes of data have made this possible, through digitalization.

We are in the era of big data and the amount of information generated daily is extraordinary (≈ 328.77 million terabytes/day). However, the truth is that not all data is of interest or of good quality. It is necessary to process, standardize, catalog, create organizational structures, debug errors and prepare it for “consumption”.

Technological capacity aside, the raw material for the proper functioning of AI algorithms lies in the data. No matter how good and efficient the algorithm is, if the data sample is of low quality, inaccurate or limited, the results produced will perpetuate these deficiencies, and may even discredit what would be an interesting project to develop with AI resources. If we think about the field of Technical Maintenance, we can divide the basic service data of assets into two large groups: those that come from monitoring systems, whether via GTC – Centralized Technical Management, IoT – Internet of Things equipment, or directly from asset controllers, and those that originate from human action, which describe historical interventions on assets and are generally recorded in maintenance management software or, more recently, in BIM – Building Information Modeling models.

Each group of data has its own particularities so that it can be used conveniently in AI models. Monitoring, for example, allows access to a large volume of data, but generally what interests us in the analysis are the anomalies or trends that led to anomalies in the assets. This situation may depend on a particular variable or on the combination of several variables, directly or indirectly related to the assets (e.g.: load in service, number of working hours or even environmental operating conditions). In certain cases, the frequency of these occurrences may be so low that it is not possible to obtain useful data in an acceptable time window. And, although it is possible to expand the sample, evaluating the same or equivalent assets at the same time, it is necessary to ensure that this analysis of the operating variables (direct and indirect) is carried out under the same conditions, which can increase the complexity of the models and the necessary extrapolations.

In terms of historical interventions on assets, data is highly dependent on the human component and, regardless of the maturity and rigor of the technical teams that monitor and record data, the individuality and interpretation of each person will always affect the uniformity of the records. There are, therefore, major challenges in data management, so that AI can contribute in a widespread and positive way to improving the sector’s processes, both in the strategic management component and in supporting operations. Even so, we have gradually been seeing a growing number of companies and entities in the scientific and technological system launching commercial and pilot products that already make it possible to take advantage of the potential of AI in the area of Technical Maintenance.

One of the most pressing issues today is the development of dynamic predictive maintenance models that take into account the operating conditions supported by monitoring systems and the history of interventions, reducing the time spent on each asset to what is strictly necessary, in a balance between cost, efficiency and useful life. Another field of application is the autonomous operation of centralized monitoring and technical management systems, where alarm management, equipment control, and regulation of setpoints and schedules are now carried out autonomously and dynamically by AI algorithms, based on space usage, user preferences, environmental conditions, energy efficiency, among other variables (Living Buildings).

The next steps will involve the development of personalized virtual assistants that will enable each employee to improve their performance, expanding their analytical, perceptual, execution and decision-making capabilities, regardless of the roles they are required to perform in the organization. A practical example can be seen in augmented reality applications, which can integrate virtual assistants to support the resolution of technical failures. This and other similar tools will mitigate the effect of the reduction in qualified people in the job market, because they will allow new employees with less experience to be included more quickly.

In the medium term, there is still significant room for AI to evolve, with the maturation of applications and databases that serve as the basis for the technology.
We are moving towards AI tools being increasingly prepared to take control of the mechanical skeletons that robotics has been perfecting over the years, thus increasing its sphere of influence in the field.