All Positions

Research
Mechanics

AI-Assisted Machine Condition Monitoring

DC-50
INSA-L and UNSW
Lyon (FR) and Sydney (AU)

Proposed Projects

Option 1

Digital-twins and artificial intelligence for robust machine condition monitoring

Artificial intelligence (AI) has attracted immense interest in machine condition monitoring (MCM). The enthusiasm of researchers in this relatively new approach stems from its proven value in other image and signal processing applications, like computer-vision and speech recognition. Recent works have however highlighted the key difference of MCM from these more traditional AI applications: the scarcity of fault data. In MCM-intensive engineering applications, failure is often very expensive and prevented by strict maintenance procedures, resulting in just a handful of failure observations. Another complementary criticism that has been made to AI approaches to MCM is that they neglect decades of accumulated knowledge in degradation dynamics and machine reliability. A solution to these problems is offered by the combination of AI methods with digital twins. The latter can produce large amount of data in any condition and codify knowledge about the machine behaviour. Yet, the portability of AI solutions developed in simulated environments to the real-world is still to be proven. A clear picture of the digital-twin characteristics which ensure this approach is successful is yet to be studied, and methods to combine scarce, yet valuable, experimental data with simulations have to be developed for MCM.

Option 2

Industry 4.0 sensing for machine condition monitoring

Machine condition monitoring (MCM) is still largely based on the installation of a few expensive accelerometers on few critical machines, acquired by means of expensive and centralised electronics. In a world moving towards fleets of assets (e.g., wind farms, drones for delivery) it is paramount that this approach is replaced by more affordable, scalable and self-sufficient senor technologies. This thesis aims at exploring self-powered, inexpensive sensor network technologies for MCM. Alternatives to traditional piezoelectric accelerometers (e.g., MEMS) will be investigated, both in terms of diagnostic capabilities but also in their suitability for integration with non-invasive, easy-to-install and self-powered data-acquisition systems, able to communicate a sufficient amount of diagnostic information wirelessly in a network of monitored machines. This thesis aims at revolutionising the way diagnostic data is collected, thus enabling the collection of big-data necessary for popular data-driven approaches (artificial intelligence) and the Industry 4.0 transformation.

Option 3

AI-Assisted Condition Monitoring (AIA-CM) based on data-driven optimal signal processing

Condition monitoring critically relies on signal processing for transforming the raw data (vibration, angular speed, strain, etc.) into interpretable features (scalar indicators, spectra, histogram, etc.). One everlasting challenge is to select the signal processing algorithms among a huge number of candidates, while the best choice is obviously case-dependent. Another challenge is to properly use signal processing algorithms, while they often rely on several critical hyperparameters whose optimal setting is again data dependent. The aim of this research project is to propose a solution to these issues, by making signal processing transparent to the user. It consists in developing a machine learning approach, where each algorithm together with its set of hyperparameters is seen as a probabilistic object in a Bayesian hierarchical framework. The idea is to simultaneously test several candidate algorithms and select the best ones, or a combination of them, according to a given dataset, and to jointly update the values of the hyperparameters that most likely explain the data.

The approach will be demonstrated for bid data processing, on open datasets consisting of large numbers of signals recorded on different machines.

Supervisors

Jerome Antoni
Zhongxiao Peng
Pietro Borghesani

Research Areas

Mechanical, industrial and aerospace engineering – mechanical system dynamics and modelling, measurements and signal processing, and machine condition monitoring