The Industry 4.0 revolution is changing the way companies manufacture their products and operate factories, and predictive maintenance is amongst the hottest buzzwords of the new digital revolution.
Predictive Maintenance, also referred to as Maintenance 4.0, leverages new technologies and frameworks such as industrial internet of things (IIoT), artificial intelligence (AI) or machine learning (ML) and usually involves IT experts and data scientists in order to understand patterns in machine behavior. Analysis of these patterns with the help of advanced algorithms can identify deviations from the default pattern that may indicate or predict potential failures or malfunctions. Prevention or correction of these malfunctions in advance may prevent longer and more expensive downtimes.
Predictive maintenance in manufacturing is an integral part of the wider concept of factories becoming smart factories, is based on real-time production data to find the abnormal patterns in machine behavior, and is crucial for factories that are losing billions of dollars every year due to unplanned downtime.
In this blog post, we cover the four common maintenance models in manufacturing, and discuss the advantages and disadvantages of each model. We will show why real-time data and visibility into machine performance is the most important building block of PdM and every industry 4.0 related technology, necessary for factories that want to become smart factories, make data-based decisions, and identify and resolve bottlenecks and inefficiencies.
Let’s review some of the main maintenance methods in manufacturing, their advantages and disadvantages in operating machines on the shopfloor.
Reactive Maintenance is probably the oldest and cheapest form of maintenance, in which machines and equipment are repaired only when they fail or become defective. However, this often results in supply bottlenecks or financial losses.
In small businesses with one-off productions, reactive maintenance is an easy way to fix problems, but it’s also an ineffective one because the process does not involve any real progress.
Relying on reactive maintenance and not having a more effective, proactive solution integrated means that production must scramble to find available parts, technicians and experts as quickly as possible to repair machinery and equipment after a breakdown. If the experts are not available, this quickly leads to delivery delays and disgruntled customers. When the time for emergency maintenance and repair comes, the ongoing manufacturing process is stopped.
Reactive maintenance requires no upfront capital investment but often proves to be expensive later due to the negative impact on productivity, performance metrics, and the bottom line of production.
Another issue is that if the equipment already failed, fixing it could be very expensive.
If factories can prevent this, they can avoid the expensive fix by maintaining their machine properly and replacing parts that have aged without waiting for them to fail and cause bigger damage. This maintenance model is called preventative maintenance.
Preventative Maintenance (PM) is aimed at maximizing the lifespan of every machine in the factory and is all about preventing performance degradation, improving machine reliability and uptime, and replacing worn components before they fail.
Preventative Maintenance typically involves the educated and timely use of processes such as cleaning, adjustments, lubrications, repairs and replacements that help extend and maximize the lifespan of the involved equipment and machine on the shop floor, and essentially turns these procedures and actions into a systematic protocol.
PM is used mainly on prioritized assets as it has higher cost of management, utilizing both manpower and production time resources. Therefore, the costs of performing preventative maintenance, should always be lower than the cost of failure.
A more efficient way would be to plan maintenance according to the actual condition of the machine, by collecting performance data and other component characteristics to understand the status of machines in the factory.
Condition Based Maintenance (CBM) is performed based on the evaluation of the latest operation status and condition of an asset, and is being implemented by factories around the world today. The decisions are based on analysis, either through technicians inspecting the machine, or by leveraging machine data to monitor the characteristics of the machine to determine the most cost-efficient time to perform maintenance.
Unlike preventative maintenance, which is held according to a fixed schedule, Condition Based Maintenance emphasizes data collection from vibrations, temperature, pressure, voltage or speed to indicate when maintenance is needed, and to prevent unplanned downtime.
CBM will not only improve the lifespan of equipment, it is also performed while the asset is working, and is therefore a scalable solution.
Condition based monitoring leverages real-time information in order to prevent malfunction on the shop floor today, but does not indicate or alert decision-makers about failures and downtime tomorrow. Therefore, CBM does not predict the future in the production.
“Information is the oil of the 21st century, and analytics is the combustion engine.” – Peter Sondergaard, senior vice president, Gartner Research.
Predictive maintenance (PdM) is the next level in maintenance as it combines condition-based diagnostics with complex algorithm-based insights and artificial intelligence, to recognize machine abnormalities and predict future failures. It is the most technologically advanced form of preventative maintenance based on smart technology.
The technology is new and not necessarily financially manageable for many organizations, therefore PdM is mostly relevant for companies that have the potential of losing millions of dollars every year due to unplanned downtime.
Process Manufacturing focuses on creating a modular production process, something that allows outputting products in mass numbers.
Due to the fact that a single malfunction can bring the entire production process to a halt, Predictive Maintenance is crucial for this type of manufacturing. For instance, PdM could be crucial in a food production plant, where multiple ingredients are mixed systematically to produce thousands of packages, bottles, and cans per day, and when every hour of downtime results in losses of hundreds of thousands or even millions of dollars of products not being manufactured.
With the focus being on mass production, unexpected decrease in productivity has a huge impact on the manufacturer’s bottom line and even brand image.
Discrete manufacturing, on the other hand, focuses less on volume production due to the high number of processes involved in producing just one unit. In discrete manufacturing, every manufactured unit is a part of a more complex product being assembled. A downtime in a single machine in a discrete manufacturing plant is not likely to result in huge losses, and therefore predictive maintenance brings low value to this industry, which is struggling with downtimes caused by inefficient processes rather than machine failures.
The fact is that despite the obvious benefits and advantages of implementing the previously mentioned maintenance and monitoring procedures, most factories still don’t have the required infrastructure to do so.
A large number of factories are currently incapable of acquiring and utilizing data from their shop floors. Machine connectivity and digitalization are the key for real-time data that can eventually prevent unplanned idles, and increase machine productivity and efficiency.
The common initial challenge in most of the factories is to digitalize and retrofit brownfield machines and equipment in order to generate data and insights.
There are a few utilizable data channels in manufacturing. It can start with sensor data on or in the machines, all the way to Programmable Logic Controller (PLC) data. Unlike PLCs, sensors have the ability to harvest data from any type of machine, regardless of its make and model. External sensor data can measure temperature, vibration, acoustics and electricity, usually in a wider range as the internal machine controller itself.
Combining machine performance insights with additional data sources (i.e. – planning data), could help determining even a more accurate time to schedule maintenance.
IIoT (Industrial Internet of Things), as some of the trends in industry 4.0, can have a huge impact on digitalization. IIoT combines data harvested on the shop floor with internet technologies and tools to create a digital manufacturing platform, that is not only interconnected, but communicates, analyzes, and uses information to drive actionable insights and improve the efficiency and performance of the machines.
Any maintenance methodology is justifiable as long as it brings high value at a reasonable cost.
Predictive Maintenance is a relevant solution mainly in process manufacturing because it can prevent machine downtime which causes significant financial losses.
In discrete manufacturing however, the value of asset monitoring and improving processes based on real-time machine data analysis is significantly higher than the savings due to predictive maintenance, because downtime of a single machine will not have an immense impact on productivity as in process manufacturing.
Whereas real-time data from the shop floor should be utilized in process manufacturing to predict failures and prevent downtime, in discrete manufacturing real-time data should be used for asset performance monitoring, to identify bottlenecks and improve processes that will increase efficiency and reduce operating costs.
