Over the years, the basic challenge for production managers does not change – how to minimise the number and duration of machine breakdowns. It is impossible to avoid the progressive use of components and the failure of machines and equipment. But with the right maintenance strategy and the right tools, we can minimise the risk of unplanned shutdowns and unexpected failures, which bring serious financial losses.

Breakdowns and unplanned stops in the production or work of machines lead to:

• decrease of production efficiency (quantity of produced items),
• lower quality of products,
• increase of production costs.

The way of servicing or maintaining machinery depends on the specific character of manufactured products and the organisation of the production process. Therefore, it is crucial for each manufacturer to develop an optimal maintenance system.

There are several maintenance concepts. Before Predictive Maintenance (PdM) was developed, the most popular schemes were Reactive Maintenance (also called Run-to-Failure) and Preventive Maintenance (PM). They are based on reacting only after failures have occurred (Run-to-Failure) or following predefined schedules for maintenance, component replacement and repair (PM). Reacting only after a failure has occurred, causes a longer duration of the failure and involves the necessary cost of repairing the failure. Preventive maintenance, on the other hand, can generate unnecessary costs for maintenance activities when performed as planned but not actually necessary.

Predictive maintenance is an approach based on observation of processes and the condition of machines in order to prevent failures and predict the time of necessary preventive repairs. These activities help to increase the efficiency of machine use and minimise unplanned shutdowns. This concept is one of the main pillars of Industry 4.0, using the latest technologies and innovative methods and solutions. Modern failure prediction systems are based, for example, on Artificial Intelligence, Machine Learning or Cloud Solutions and applications. This allows for faster data collection and lower information gathering costs.

To make predictive action effective, data acquisition is essential. Therefore, it is necessary to collect and analyse a number of parameters, which sometimes, at first glance, may seem unrelated to the operation of the devices, e.g. the temperature of the machine and the environment, the efficiency of heating systems, air vents and air conditioners, machine vibrations, the pressure of water or other utilities in installations, the time taken by the machines to perform particular activities, etc. Machines are designed to alarm about out of norm states – they are equipped with necessary sensors. The predictive maintenance approach is based on using them or installing new ones. On the basis of the data collected from them, it is possible to analyse the work of the machine and to plan in advance and prepare for focused renovations.

By adding IoT sensors and cameras to devices, we can gain data that can be precisely analysed with Computer Vision and the use of Artificial Intelligence (AI) solutions. State-of-the-art technology makes it possible to process huge amounts of data (Big Data) in the cloud, so that the results can be checked anywhere and anytime. In everyday work, this enables remote control of multiple machines at the same time. This results in more efficient work for those involved in the maintenance of production lines.

In order to predict potential failures or damages in advance, it is necessary to:

• collect as much data as possible from machinery and equipment and create virtual models of their operation (patterns),
• automatically and continuously analyse and compare the current status with the developed model,
• automatically and continuously analyse the values of current and historical states and predict their future values.

The methodology and algorithms used in predictive maintenance are ideal for:

• manufacturers of machinery and equipment used in mass production,
• factories and production plants using machines and equipment for production,
• system integrators and installers providing solutions for mass production.

The question is whether predictive maintenance is better than preventive maintenance and which solutions generate less costs. Certainly, the two approaches differ in implementation cost. Predictive maintenance requires spending more time and resources on implementation activities. The preventive approach, on the other hand, generates costs of regular inspections and repairs of machines, regardless of whether they are necessary or not. In the end, it is therefore a more expensive and resource-intensive solution. Successful Predictive Maintenance solutions lead to optimised maintenance planning and to avoiding unplanned breakdowns and downtimes. This will ultimately lead to savings and minimisation of production costs.

It is important to remember, that successful Predictive Maintenance solutions are specific and even unique to a single machine, piece of equipment or process line. They replace preventive actions, although a combination of these two approaches may be the most effective. The methodology, algorithms and software need to be adapted to the concrete implementation. Predictive maintenance is the future of industry. Innovative solutions in this area will help to gain competitive advantage, which will have an impact on the final product prices. Companies aware of the need of responding to market demands will look for ways to maintain their machinery stock in optimal shape. The functioning of a machine can be compared to the functioning of a human organism. So, the maxim that prevention is better than healing fits perfectly here.

Andrew Szajna, Chief Digital Officer

We are the specialists in Predictive Maintenance. Our in-house developed solutions allow our customers to significantly improve the maintenance process in production plants. Under an EU project, we are currently developing innovative analytical and decision-making methods to support predictive maintenance processes. The result of the project will be a product allowing to control the condition of machinery in a production plant, indicating the most probable causes of future failures. Thanks to our experience with technologies such as: Cloud Computing, Artificial Intelligence, Computer Vision and Big Data, we can provide support at every stage of the prediction process.