What is predictive maintenance and how does it work?
From a service perspective, maintenance is something we do to a machine to maintain the performance and or prolong the useful life of it. Preventative maintenance adds a schedule to the maintenance activity. That schedule is typically based on the usage and theoretical failure modes and timeframes and timeframes are often determined by the average calculated or observed use rate. Since actual rates will vary with end users, preventative maintenance schedules usually result in maintenance being applied across the spectrum – early, late and occasionally, on time.
As sensorization and more detailed asset monitoring became pervasive, preventative maintenance made way for a condition-based program. We’ve seen this with automobiles as oil changes moved from time based, to milage based, to oil condition based. Predictive maintenance is a further evolution of maintenance. It is a technique that uses libraries of recorded machine operating parameters, detailed, historical failure information and data analysis tools to detect inconsistencies or defects in equipment and processes and inform service technicians and mechanics to address them before they produce a failure. Predictive maintenance goes beyond preventative and condition-based strategies that are rooted in statistical norms and failure mode analysis. It helps prevent unplanned reactive maintenance, a maintenance strategy that involves fixing equipment or machines only when they break down or malfunction.
Predictive maintenance can reduce the need for excessive preventative maintenance, a proactive approach to equipment maintenance involving time-based scheduled tasks to keep equipment in optimal working condition, as well as increase equipment longevity and company profits. And unlike condition-based maintenance, which relies on usage rates and wear indicators to identify when to perform maintenance on equipment, predictive maintenance uses sensor data and trends along with historical performance and failure data to predict future equipment failure.
Predictive maintenance uses data and analytics to identify the time when maintenance is needed. This involves monitoring equipment performance and analyzing data to isolate and identify patterns, interactions and potential issues before they become major problems. Preventative maintenance can help avoid unexpected downtime for a subset of the asset population, but predictive maintenance can be more efficient and cost-effective, as it allows for targeted repairs and avoids unnecessary maintenance across the asset population.
What is a predictive maintenance program?
A predictive maintenance program is a dedicated effort focused on reducing costly breakdowns and improving the longevity of machinery by directly monitoring the condition and performance of equipment during normal operation. Predictive maintenance initiatives leverage current and historical operating data, along with historical failures, to isolate patterns and trends that lead to failures. Once these events can be reliably identified, they can be integrated into day-to-day operations to help reduce the probability of unexpected breakdowns or malfunctions, saving companies money and maintaining an efficient operation.
Infrared testing, acoustic inspections, vibration analysis, sound level measurements, oil analysis, computerized maintenance management systems (CMMS), condition monitoring solutions, data integration tools, and various other sensors can be used to track equipment conditions and performance with predictive maintenance. Implementing predictive maintenance techniques allows for a deeper understanding of machinery, taking proactive, rather than reactive, repair measures.
Pillars of a predictive maintenance program
In the simplest terms, technology is a tool, it’s not a solution. If people are not informed on how to use the tool or what it tells you, it’s not going to help. If the tool is not integrated into daily processes, it’s not going to be adopted or used. Thus, the interfaces between People, Process and Technology are critical.
Operational keys to a Predictive Maintenance Program:
- Processes—Predictive maintenance requires well-established processes for optimal asset performance involving both people and equipment, including those responsible for maintenance tasks and how machines interact with data and systems. It's important to assign responsibilities, review tasks regularly, and provide clear instructions for each job or project.
Both people-driven and equipment-driven processes are involved in a predictive maintenance program. People processes are how the maintenance team interacts with machines, data, and each other. This can include understanding responsibilities, data review frequency, communication, planning, escalation, and task completion.
Equipment processes are knowing what processes the equipment completes, and how to capture asset data as well as highlight the asset failure modes.
- People—People are the key to maximize the benefits of predictive maintenance program. A solid plan on paper is not enough, as successful implementation requires support and participation from all involved. Without proper collaboration and adoption across the organization, even the best strategy will be ineffective.
Predictive maintenance requires people to interpret data and manage technology. It's important for everyone involved to understand predictive maintenance and contribute to its success. Getting people to adapt to change can be a challenge, but it's essential for a successful predictive maintenance program.
- Technology—Technology plays a crucial role in program efficiency. It helps monitor production lines and predict malfunctions, while also collecting data on preventive maintenance costs. Technology manages and monitors real-time data, tracks events, and identifies potential risks for informed predictive maintenance plans.
Autonomous systems and live monitoring dashboards improve accuracy throughout the predictive maintenance cycle to help operators detect patterns, prevent downtime losses, and anticipate future failures. Overall, technology is essential for optimizing operations.
- Policy—This can be overlooked, but in our service world today, we need to consider our policies, both internal and external. For example, do our service level agreements need to be changed, do we need to restrict employee access to data or systems and even basic questions like, who owns the data?
- Data—Predictive maintenance programs rely on data to connect the past and future. By gathering information on the current performance of assets, businesses can anticipate changes or issues down the line. Without quality data from plant floor or field based equipment, decision-making becomes difficult.
Even with excellent algorithms, without good input data, predictions are useless. Gathering accurate and current information from various sources is crucial for a successful program. Without a baseline for normal performance, it's difficult to identify or predict anomalies ahead of time. The key here is accurate. Make sure you understand the accuracy of the data you acquire as this will impact the reliability of the predictive model.
- Equipment—Critical assets with observable failure modes are suitable for predictive maintenance programs to identify potential equipment issues before they become problematic. It’s crucial to identify equipment that enables you to detect potential failures. Focusing on critical assets with observable failure modes and supporting data is recommended.
It should be noted that not all machines were designed with predictive maintenance capabilities. A balance between condition based, proactive and reactive maintenance is important for efficient management. Reactive maintenance still has a role to play for routine repairs and quick fixes on equipment not suited for predictive maintenance.
- Tools and Parts—Predictive maintenance relies on tools that measure asset condition and parts, which refer to the various components of equipment. Advancements in tools and understanding of parts have made predictive maintenance more efficient and cost-effective. High-quality tools and durable parts allow predictive maintenance to stay on schedule and avoid equipment failures or delays.
Using high-quality parts designed for predictive maintenance can prevent minor issues. Regular replacements can lead to better overall machine performance. Machines can run longer and more efficiently with tools and parts designed for predictive maintenance, resulting in significantly reduced downtime.
Prerequisites to a Predictive Maintenance Program:
Key steps for establishing a predictive maintenance program
Establish a strategy
To begin a predictive maintenance program, start with one or two critical assets to allow for a slow and careful approach with room to address mistakes without affecting a large number of assets. Ideally, you should Pareto the failure modes of a piece of equipment and start with a failure that is common, but also easily profiled with data. A successful pilot program can build momentum for expanding the program.
Choose the right asset to test
Determine which assets are suitable, starting with less significant equipment and scaling up gradually to improve processes. It’s also advisable to avoid piloting the program on expensive or critical equipment. Some assets may not need predictive maintenance, as they may be expendable or have low-cost solutions.
Develop a pilot program to prove success
Create a pilot program to test and refine predictive maintenance strategy and help to identify any obstacles that may arise during implementation. Clearly define goals and objectives to measure the program’s success and make any necessary adjustments before implementation. Remember, a small improvement that impacts most customers is often more valuable than a large improvement that helps a handful.
Set a response procedure
Establish a procedure to respond to flaws in a predictive maintenance plan. Set parameters for immediate shutdown. Install the sensor, acquire data, and document the project. Review lessons learned in a postmortem for the next phase of maintenance.
Build a data strategy
Create a data analysis strategy for continuous online monitoring. Storing and transmitting large amounts of data can be resource intensive. You can either train or hire in-house experts or use third-party service providers. For companies with limited resources, third-party field technicians can use online condition monitoring sensors to track asset health and troubleshoot problems.
Plan to Fail
Predictive models need to be evaluated for accuracy so we need to let assets fail in order to see the actual accuracy of the model. Of course, this is relative as some failures may be too costly. A statistical model that looks at frequency of “failure x” for the population before and after will provide insights as well.
Scale to more assets
After proving the success of the pilot program and identifying the resources needed for data monitoring, review the list of assets for the program. These may include critical assets, assets that are difficult to source or replace, and assets in remote locations or with a history of failure. Consider the cost of downtime and potential ROI for each asset to prioritize. Just remember, it costs a service organization roughly the same amount of money to fix a product on Monday as it does on Friday, whereas the downtime costs for the end user might be insurmountable. The more valuable uptime is to the customer, the more likely the opportunity to sell it as a service feature.
How to implement a predictive maintenance program
When implementing a predictive maintenance program, the first challenge is deciding which failure to address, and which data to use. Many industrial machines produce key data sets that can be collected and analyzed, such as machine current, torque, or pressure, which can help identify early signs of problems and various failure modes. Manufacturers often analyze this data for incident logging or historical analysis. Some analytics programs can identify key parameters from a set of data, others will rely on you to understand the physics of the system/problem to manually identify them.
To scale predictive maintenance, it's important to have algorithms that work under a variety of operating conditions. For service teams that provide multi-vended services (MVS), your algorithms should ideally work on multiple vendor’s products. Smart solutions learn and improve over time, finding the best algorithm for each monitored asset and optimizing their performance with more data.
AI advancements allow for the analysis of machine data on a massive scale. Computers can automatically identify signs of machine failure from data of tens of thousands of machines. Large-scale monitoring requires significant compute resources, which the cloud can provide as needed. Small initiatives can quickly be scaled up to include thousands more machines. Smart predictive maintenance algorithms provide empowering insights.
Analytics tools for large-scale predictive maintenance deliver substantial productivity. By monitoring all machinery, not just critical points of failure, manufacturers can reduce unplanned downtime by up to 50%, improve throughput, quality, and margin, and decrease maintenance costs by up to 40%.
Benefits of establishing a predictive maintenance program
Some service and maintenance organizations spend 80% of their time reacting to issues instead of preventing them. Establishing a predictive maintenance program can significantly reduce that percentage of reactive time by accurately predicting failures. Committing to a predictive maintenance program can lead to significant improvements in asset reliability and cost efficiency. It can also result in a tenfold return on investment (ROI), a 25–30% reduction in maintenance costs, and a 70–75% elimination of breakdowns.
Establishing a Predictive maintenance program is a proactive approach to equipment maintenance that can help to increase customer satisfaction. Businesses can provide better results for their customers by reducing downtime, preventing equipment failure, and improving equipment performance. Investing in a predictive maintenance program can lead to increased customer loyalty and a competitive advantage in the market. As noted earlier, the cost of downtime is significantly higher for end users than the cost of repair is for the service or maintenance team.
To stay competitive, it’s important to have all the right parts on hand to avoid delaying a job. Reacting to a failure often results in an assessment, a best guess at the failure mode and the parts required to fix the asset whereas predictive maintenance specifically identifies the pending failure mode and the required part or parts. Additionally, proper job planning is crucial to ensure technicians have enough time to complete a job and avoid follow-up visits. Clear and precise communication between office and field staff (or customer) is also necessary to accurately diagnose and resolve issues on the first visit for improved first-time fix rate.
Understanding the current state of predictive maintenance programs
Predictive maintenance is a strategy that uses the past to look ahead, in an effort to prevent problems before they occur. Newer technologies like IoT have made predictive maintenance more efficient, affordable, and achievable for a variety of industries. IoT strategies are commonly implemented for predictive maintenance and service strategies in industrial manufacturing, where manufacturers are adopting data usage and predictive strategies with analytics. Remote maintenance and e-maintenance research support predictive maintenance in unsafe, geographically challenging or scattered locations. Predictive maintenance is essential to the success of data-driven organizations in controlling variation (reactive maintenance) by creating a smart maintenance plan (predictive maintenance) for their specific assets.
Subscribe to the Speaking of Service Podcast
Visit Speaking of Service to listen to the latest episodes from our experts, review host information, and explore all topics covered in past episodes.
Listen Now