In a world of increasingly advanced, connected and intelligent machines and equipment, the issue of maintenance is central to any organization that strives for continual improvement in productivity and quality. In this article, we will discuss the issue of predictive maintenance. From how it is defined and what approaches can be used to implement it, to the benefits that it can bring to organizations that decide to adopt it.
What is predictive maintenance and how is it different to other maintenance strategies?
There are many types of maintenance strategies, and not all of them are alternatives. Before we get into what really interests us in this article, let’s take a quick look at the possible alternatives.
This type of maintenance is the product of urgency. Unsurprisingly, reactive maintenance is what happens only after a fault has occurred, often when the component concerned has come to the end of its life cycle. In such cases, production is interrupted and the fault may have negative repercussions on other components, security, and even the environment.
In this case, the machinery undergoes maintenance checks at regular intervals, before any fault occurs. The parts replaced to avoid faults may, however, be fully functional. One example of preventive maintenance is condition-based maintenance, which determines the need for intervention by monitoring and assessing how well a component or machine is working (e.g. by checking that the key parameters of a process or machine stay within the necessary threshold or limit values).
Predictive maintenance is a proactive strategy that analyzes data and suggests changes to production, planning and maintenance in order to prevent faults and avoid machine failure. This strategy transforms the machinery’s data and processing history into information that then serves as a base for making analyses and improving maintenance processes.
Predictive maintenance is similar in some ways to condition-based maintenance, but it uses logics and models that analyze data history and assess the component’s deterioration over time, using this to estimate when to check for faults.
7 Advantages of Choosing Predictive Maintenance
In the next part, we will describe the major factors that make predictive maintenance the most modern paradigm for applicable maintenance.
Introducing proactive maintenance routines reduces the risk of unplanned downtime and unplanned slowdowns due to unpredicted failures. This means productivity rates are maintained, which then has a beneficial effect on revenues.
Reduced labor costs
Reducing and optimizing the number of maintenance operations makes it possible to manage costs more efficiently and leads to a significant reduction in operating time spent on maintenance. Additionally, by giving an indication of where and when to carry out maintenance work, interventions can be scheduled, the number of failures and, as a result, the duration of repair times can be reduced, given that there will be fewer components to replace.
Less production time lost to maintenance work
The results of predictive analyses mean that faults can be predicted days or weeks in advance, and therefore allow plant downtime to be planned at the best time when it will have the least impact on productivity.
This exercise is helped by defining the so-called P-F curves, where time intervals between potential failures, P (representing the point where the machine starts to deteriorate and fail), and a functional fault, F (representing the point at which the machine has reached its use limit and is no longer functioning) are evident. The greater the P-F interval, the greater the available margin for planning maintenance.
The fact that unplanned downtime has a direct effect on OEE (Overall Equipment Effectiveness) and causes a cascade in which a series of inefficiencies along the production chain, which may include both internal (shipping services, warehouses, etc.) and external (transporters, external depots, end customers, etc.) clients, makes planning all the more important.
Reduced machinery costs
Being able to perform maintenance work on a component before it causes a critical failure reduces costs down to that of the faulty part and the labor required to repair that part, instead of having to replace the entire machine. Reducing the number of failures also increases the machinery’s average lifespan.
Reduced environmental and security risks
One direct result of resolving potential problems before they appear is having safer and more environmentally-friendly working conditions, by reducing the risk of critical incidents and leakages into the environment.
Increased efficiency of maintenance personnel
By carefully planning maintenance work, it is possible to improve the efficiency of maintenance personnel.
One other advantage of predictive maintenance is the ability to create a historical database of machine performance and behavior, which can be used to increase prediction accuracy in the future.
The 3 fundamental elements for drawing up a predictive maintenance strategy
In order to create the best predictive maintenance plan, the key elements you need at your disposal are:
- Machine data collected by historian, IoT or other systems relative to a significant period which includes failure events
- The history of faults recorded by the machinery, including as much data as possible
- Personnel with experience operating the machines and who are able to diagnose fault causes both by reading both the collected data and other indicators that may not be managed by historian, IoT or other systems
Speaking of approaches: 3 Strategies for Planning and Implementing Predictive Maintenance
At Techedge, we have selected three alternative approaches, each distinguished either by how they are implemented or how they are managed.
The custom approach involves designing and implementing predictive maintenance systems by defining custom algorithms on open-source AI platforms.
While on the one hand this solution allows you to tailor the algorithm to the specific needs of the process and the machines to be maintained, on the other, it demands the availability of processors and mechanics with strong technical skills, as well as data scientists capable of developing complex predictive maintenance algorithms.
This approach also means that the management of the entire system is completely in the hands of the mechanics/processors and data scientists who designed and implemented it.
Main features of a custom approach:
- The investment cost of designing and implementing a customized system is high.
- Development and implementation periods are on average longer.
- The time it takes to see results is inversely proportional to the time spent on development. In fact, the more similar the model and the process to be monitored are, the faster the results can be obtained.
- Committing personnel to developing and managing the system increases internal know-how.
The hybrid approach involves adopting a product available on the market, which comes with prefabricated predictive maintenance models which have to be trained and calibrated.
This solution requires available mechanics and processors for the model-configuration phase; however, it is not necessary to understand AI algorithms to implement this solution, meaning that the typical data scientist skill set is not necessary. The same processors are also responsible for managing the system over time.
Main features of a hybrid approach:
- The investment cost of setting up the models is modest
- You will need to sustain software licensing costs
- How fast results are obtained depends on how easily the available models can be configured. In fact, the greater freedom afforded to model configuration, and the more the model adheres to reality, the faster the results will be obtained.
Off the shelf approach
In a similar way to the hybrid approach, the off the shelf approach involves implementing products available on the market which use configurable predictive maintenance models. However, in this case, the buyer gets not only software, but also a support service, which covers both model configuration and fault management and analysis over the years.
In this case then, no specific expertise is required on the part of the client company.
The off the shelf approach is typically offered to machinery vendors who have gained solid experience in the field over time with a range of clients, and are therefore able to provide comparative analyses on machine performance across all their clients’ plants.
Main features of an off the shelf strategy:
- A relative economic commitment in order to sustain licensing costs and service management
- The time it takes to see results is typically longer than with the other two approaches.
|Off the shelf approach||External||External|
To make a comparison now, we have provided a radar chart that visualizes the differences between the three approaches proposed based on the following elements:
- Licensing and management service costs
- capacity for customization
- development/implementation speed
- time to see results
- increase in company know-how
- comparison with third-party services
Generally speaking, it is not a case of best and worst approaches, but rather one of finding an approach that is well-suited to the specific needs of your company in terms of budget, internal know-how and the other factors that we have discussed.
It’s never too late to make an improvement: get in touch!
Our consultants are on hand to help you identify which predictive maintenance strategy is best for your plants.
Other articles of this series on Digital Advisory:
- People at the center: the key to a successful transformation
- The role of Technology Onboarding in innovation projects
- The pros and cons of remote Design Thinking
- Checking and monitoring devices remotely: IoT or Historian?
- Overall Equipment Effectiveness (OEE): the KPI for improving production processes
- Digital Twins, a pillar of digital transformation
- Heat and material balances: monitoring yield and consumption, identifying losses