Table of Contents:
Condition Monitoring (CM) is a process of monitoring the parameters of a device’s condition (e.g. vibration, temperature) to identify changes that may indicate developing damage. Thanks to Condition Monitoring, maintenance specialists are able to monitor an entire plant or production line, as well as a single machine or its component.
Constant access to information and the ability to actively monitor the condition of machines provided by CM is the next step in the development of companies that have given up on the unreliable maintenance strategies they had implemented so far. A preventive approach (execution of repairs according to the schedule) or a reactive approach (taking action when events occur) does not use the enormous potential of the data that is generated by machines, and therefore exposes companies to losses and limits knowledge about their assets.
The definitions describing Condition Monitoring clearly emphasize the relationship between CM and Predictive Maintenance (or PdM) – CM is one of the main components of Predictive Maintenance (PdM).
Predictive Maintenance is recognized as one of the most innovative solutions for predicting machine failure and is used in a wide variety of industry sectors. Predictive Maintenance (PdM) is a technique developed to assist in establishing the condition of equipment to determine the correct time for maintenance work. PdM uses various tools to fulfill the promises made in its definition. In the case of rotating machines, CM tools (e.g. vibro-diagnostics) will be used, while for static devices or structures, these will be non-destructive testing (NDT) and structural health monitoring (SHM).
Certainly, an important step in the development of PdM systems is the use of machine learning (ML) tools and artificial intelligence (AI) – in short, predictive analytics. These tools give the possibility to perform multidimensional analyses of large data sets, which allow us to take the next step in relation to classical diagnostic systems (such as classical vibration level analysis) – we can not only detect a failure at its earliest stage but in some cases also alert before it appears. For this purpose, they use all available data, combining information from systems such as CM, NDT, SHM, with information from measuring systems, environmental measurements, etc.
However, a “full” PdM system is not only about early detection of damage – extensive and properly selected measurement systems, storage and transmission of large amounts of data (including cyber security), analysis of failure causes, appropriate training of operators, cooperation with ERP or MES systems are also important.
CM or SHM techniques are usually based on known physical relationships (e.g. increase of vibrations in a given frequency band). Maintenance supported by predictive analysis also uses analysis of historical service and maintenance data, such as: large sets of sensor data (several years/hundreds or thousands of parameters), failure and repair data, data describing technical parameters and machine operation process, environmental data, etc. Furthermore, analytical techniques, as opposed to classical condition monitoring techniques (such as vibro-diagnostics) that are widely used in the industry, give the possibility of multi-level data analysis. This allows not only to detect failures early, but also to try to anticipate failures before they start. If vibration analysis informs us of an already existing problem at an early stage, the predictive analysis already predicts that vibrations may increase and will indicate their cause.
Advanced analysis methods also allow for simultaneous analysis of data from an entire installation or process, not limited to data from a single machine. Often, a failure occurs as a result of an abnormal event in the process, the consequences of which are propagated through the subsequent production steps. Even the most intelligent sensors will not be able to perform advanced analyses and generate appropriate conclusions.
Predictive analytics is about searching for relationships between data that explain the occurrence of failures and thus make it possible to avoid them in the future. According to research conducted by McKinsey & Company, implementation of a Predictive Maintenance strategy in the company can save up to 40% of maintenance costs in the long term, as well as reducing the expenditure for machinery and equipment by up to 5%.
In the world where we cannot control so many things we offer you our best practices and proven solutions.
What can you control with AI-driven tools for Predictive Maintenance? How can you reduce energy consumption using them?
Let’s meet at the 13th edition of SYMAS in Krakow! The fair covers a wide spectrum of topics related to the optimization of production processes.