We are immersed in an historic technological revolution in which data analysis has taken centre stage and sooner rather than later will lead all business organisations, or at least those that want to remain competitive and profitable, to become fully Data Driven organisations.
This technological revolution has given rise within the industrial sector to the term Industry 4.0, or fourth industrial revolution, a new scenario that leverages both the automation of processes and the interconnection of data based on IIoT technology (Internet of Things applied to Industry).
This is a set of tools, devices and, of course, sensors, which are responsible for both data collection and analysis for subsequent decision-making at the operational and management level within the organisation itself.
Sensors, an essential component
Sensors, therefore, have become a basic component, since, through the detection, measurement and analysis of factors, they enable greater automation of industrial processes. Their measurements are subsequently translated into commands, which are then executed by the actuating/executing components within a well-defined action/response plan.
However, the functionality offered by sensors is not limited exclusively to increasing process automation; their use becomes essential for industrial maintenance, as these assets can enable significant savings in maintenance or repair costs caused by unplanned production stoppages, improvements in profitability thanks to constant monitoring throughout the manufacturing process, thus generating higher performance rates on production lines, as well as improvements in the safety of industrial workers themselves.
So, what exactly do the sensors measure?
As can be expected, the answer to this question is a wide range of variables, which will depend on the specific characteristics of what is manufactured, but we can group them into environmental variables (temperature, humidity, light, vibrations…), mechanical variables, derived from the machinery itself (position, proximity, speed,…), electrical variables from energy consumption (voltage, current, resistance, power,…) and process variables about physical or chemical conditions generated during manufacture (fluid level, temperature increase in machines and cooling times, waste level, densities,…).
Sensors have become one of the most sensitive parts in the process of capturing early information to provide an adequate response in time and form during manufacturing.
Given such heterogeneity of available variables, it follows that sensors have become one of the most sensitive parts in the process of capturing early information to provide an adequate response in time and form during manufacturing.
This is why both the identification of the type of sensor to be installed, its location within the chain and the maintenance of these sensors are crucial to ensure that the measurements are reliable and significant, as incorrect measurements due to a defect or fault in the sensor can end up leading to imbalances in the composition of the manufactured goods or can even mean a total shutdown due to a critical error, either because of having used too many or too few components or ingredients that are necessary in the right proportion to maintain the quality expected and approved in the standards, protocols and certifications.
So, what type of maintenance should be carried out?
There are different approaches to address this question, which can be summarised in 4 different types of maintenance, depending on the implementation strategy.
- Corrective, where the sensor can work until it fails, at which point it is repaired or replaced.
- Preventive, which is carried out systematically through inspections, whether or not the asset has failed, and which, together with corrective maintenance, are the most widespread strategies to date.
- Predictive maintenance, which makes use of predictive algorithms to estimate in advance the moment of sensor failure, so that maintenance will only be carried out when necessary, anticipating the incident.
- Prescriptive strategy, which is based on predictive maintenance and incorporates elements of maintenance management, costs, etc.
A proactive strategy focused on anticipating and correcting and will determine with greater precision the useful life of equipment, risks of failure and potential impact on the system.
As sensors become cheaper, their implementation continues to be promoted throughout the production chain, and this interconnection of data generated during manufacturing, in combination with Artificial Intelligence techniques within the Big Data technological environment, is causing a shift from prevention to forecasting in maintenance processes.
It will be less and less necessary to stop processes to analyse errors and/or solve problems when deploying constant predictive maintenance, since predictive models executed in real time using historical, inventory and process data will be used to model the failure pattern by learning patterns that precede failures in a machine, sensor, asset, etc. and, consequently, predicting when maintenance or replacement of the sensor or part will be necessary before the functional failure occurs.
In other words, we will be implementing a proactive strategy focused on anticipating and correcting and will determine with greater precision the useful life of equipment, risks of failure and potential impact on the system.
This proactive strategy based on the ‘sensorisation’ of the plant and the adoption of automatic learning techniques contributes to the fact that, in the long term, predictive maintenance offers lower recurrent costs than other maintenance strategies, since the higher initial investment is returned in increased ROI, as the number of incidents detected in advance increases, thus reducing the rate of critical failures in the chain.
A clear example of this change in maintenance strategy can be seen in those industries that have an intensive use of electric batteries, both in controlled static environments (industrial facilities, telephone systems, etc.) and in dynamic mobility environments (railway environment, electrified transport, etc.), where it is vital to estimate acceptance-rejection values for batteries with a projected useful life of several years to ensure that they will not be operating in the near future within critical ranges that compromise their integrity.
In the automotive sector, more and more car manufacturers are relying on predictive maintenance to continuously monitor the performance of electric vehicle batteries. Sensors installed in the car constantly feed data to a virtual model of the battery, known as a digital twin, which enables large-scale modelling of the service performance and estimation of optimal battery life under different usage conditions in a laboratory environment.
This approach to creating digital batteries leads to significant time savings, as physical testing of different conditions is a handicap due to the long lifetime of batteries, while allowing multiple simulations in parallel without the need to deploy complex and costly physical test environments.
Knowing how long it will take to reach critical values that compromise performance will allow specific actions to be deployed to extend battery life by replacing parts and improving the design of new cells and batteries.
Moreover, this optimisation of performance has the additional positive effect of reducing the environmental impact, as less and less waste will be discarded and reused more frequently, extending vehicle lifetime and allowing batteries to be a real key lever for change in the decarbonisation of transport and part of the industrial processes.
🔵 More content on IoT and Artificial Intelligence can be found in other articles in our series – the first article of which can be found here,
