Predictive Maintenance, the evolution of fault prevention systems in Industry 4.0

With industry 4.0 the systems are increasingly connected and the machines more efficient. This also involves an evolution in the fault prevention method, which sees predictive maintenance as a valid alternative to traditional and manual methods, being able to provide precise information in real time on the health status of production systems.Predictive maintenance is carried out by acquiring and subsequently processing physical quantities relating to the production system. Subsequently, using mathematical models or algorithms, raw data are processed to obtain information on the health status of the system.The objective is to identify physical quantities of reference to obtain information on the production system in useful time, through their analysis, in order to prevent a problem. By analyzing magnitudes such as speed, acceleration, temperature, electric currents, vibrations and noises it is possible to monitor and estimate the time to failure.For the acquisition of physical quantities, the sensors used can be of different types, to be chosen taking into consideration factors such as working conditions, corporate technological maturity, costs, etc. Below is a list of some types of sensors divided by technologies:MEMS technologies (Micro ElectroMechanical System): it is considered one of the most promising XXI century technologies. The main features of these sensors are: compact dimensions, low costs and reduced power consumption. The fundamental advantage is the easy integration of these sensors in the system;Radio technologies: they are the privileged channel for data exchange and transmission, making the connection between devices independent of the laying of dedicated conductors. These technologies make very easy the application of sensors for predictive maintenance even in retrofit conditions;Piezoelectric technology: it exploits the correlation, typical of some materials, between tension and pressure on the faces of the crystal lattice. This technology allows the creation of sensors such as microphones or microactuators for indirect measurements;Resistive technology: it uses the modification of the resistance of the sensitive element, opposite to the passage of current. This technology is used for the creation of a very wide range of transducers such as: temperature sensors, strain gauges (for measuring sample deformations).There are many advantages in choosing this type of maintenance:Access to accurate and precise information about the state of health of the system, useful both for the prevention (or solution) of the fault and for the choice of spare parts, technologies and production plans;Reduction of production times and costs lost for plant shutdown and repair;Reduction of labor and machinery costs, thanks to the preventive identification of the components to be repaired/replaced;Creation of a database, also useful for future analysis;Increased safety and efficiency thanks to the continuous monitoring of the health status of the system.For the development of a system for predictive analysis, first of all we must study the diagnostic needs and define the project specifications with the client. The measured data are then subjected to an analytical study whose main objective is the identification of two or three physical quantities of reference for system diagnostics. The cost-effective specifications are then defined so that the system can be managed and maintained by the end customer. Finally, it is possible to customize the hardware sensors with the aim of controlling the physical quantities of interest, by first performing a pre-processing of the data at the local level near the system and then a pre-treatment of the firmware level data on the microcontroller.An interesting case to show the effectiveness of the application of predictive maintenance, can be that related to the bearings of asynchronous motors. The main faults can be traced back to these components, whose defects are detected above all by analyzing the vibrations present in the mechanical quantities. Generalized roughness is one of the main consequences of the mechanical defects present in the bearings, difficult to identify with the classic methods of estimating defects (such as the spectral analysis of the vibrations or phase currents). The proposed predictive maintenance method was therefore aimed at identifying a generalized roughness failure index, and divided into two parts. In the first part of the activity, statistical analysis techniques of mechanical vibrations and stator currents were used to identify the frequency bandwidth(s) in which the phenomenon occurs. A fault index was then defined based on the energy contained in the previously identified frequency band(s). This method was finally validated by experimental tests on different levels of roughness and rotation speed, resulting in good reliability.Predictive maintenance therefore represents the evolution of traditional maintenance methods, which can be more complex and inaccurate.