The Industrial Internet of Things (IIoT) has been shaping Industry 4.0. A primary goal of the IIoT is for industrial applications to stay operational, highly effective, and efficient. Predictive maintenance is a popular topic for IIoT, and for good reason. Knowing when about a potential issue ahead of time has many advantages. This article will look at how predictive maintenance (PdM) helps industry maximize uptime, industrial agility, and profitability, giving its users a competitive edge.
One way to achieve optimal performance and profitability is to keep machines running as much and as long as possible by accurately monitoring and assessing their health – a process known as Machine Health Monitoring (MHM). The concept of MHM is simple: By applying various sensors to machines and collecting and processing information, designers hope to determine what, if any, corrective actions they may need to take before significant and costly issues arise.
Maintenance is a process engineers use to preserve or extend the life of equipment and to avoid costly breakdowns. It is cheaper to maintain machines than to wait until they break to fix them — not just in relation to the machine itself, but also in lost productivity. Maintenance, however, requires engineers to monitor the machines to help detect problems before they occur.
The critical technology engineers use to collect information regarding the status of their equipment comes via electronic sensors that have their data processed by microcontrollers (MCUs) and microprocessors (MPUs). These two types of equipment maintenance support are distinguished mainly by approach and methodology. Neither is better; they complement each other. In both categories, conditions are monitored. Monitoring can happen in three methods: Continuous real-time sensing, periodic interval sensing, or remote sensing.
When sensors are employed with real-time sensor measurements and analysis, this is known as condition-based maintenance (CbM). CbM might be broadly considered to be reactive. That is, things are adjusted after they happen. CbM may require many ongoing real-time sensors, possibly requiring more components, time, and data. As a result, real-time sensing and processing are often not available or practical.
In the absence of real-time sensing and processing, another technique – predictive maintenance (PdM) – might be employed. PdM attempts to utilize decision criteria gleaned from processor sensory data to make a ‘crystal ball’ prediction of the best future maintenance events that might be employed to maximize uptime and profitability. It might be viewed as attempting to be proactive or preemptive.
The primary benefits of PdM approaches include increases in an asset’s availability, life, and utilization. Simultaneously, it provides the complementary benefits of reducing an asset’s capital replacement decision time, failure costs, failure frequency, maintenance costs, operational and safety risks, and unplanned downtime.
Industrial users generally wish to spot any potential problems sooner rather than later. The value or benefit of this is obvious: Earlier detection helps reduce system downtime, improve operational runtimes and throughput efficiencies, and minimize costly or significant repairs and safety issues. PdM is valuable because it saves both money and time. After all, maintenance only gets performed when necessary.
Over time, the potential for errors of omission will shrink. One should never expect perfection or a total elimination of any failure issues. They will, however, work to reduce and mitigate losses by a significant margin.
As noted, PdM employs predictions about the future. Predictions are projections based upon empirical observations of the past. There must be a high degree of correlation among past information collected in order to attempt to make a strong prediction of the future. If there is not a high correlation, any prediction may turn out to be uncorrelated speculation.
Predictive Analytic Models
The desired goal of PdM is to successfully achieve a predictive analytics model (PAM) that provides a set of probabilities to inform the maintenance process. A successful PAM model achieves its goal by preventing failure by scheduling appropriate corrective maintenance to remedy the situation and restore it to standard conditions
Successful PdM relies upon IIoT. IIoT collects data on the actions of the machines and places it into a digital format that can be transported, processed, aggregated, and analyzed. Wired and wireless connectivity solutions provide sufficient bandwidth to handle large amounts of data. This allows a complete model to be built at the edge or in the cloud.
Because the goal is to avoid asset breakdowns without warning, the ability to predict when maintenance should be scheduled is imperative. To help make these predictions with higher confidence levels, artificial intelligence (AI) is employed to assist because tiny changes can be analyzed very quickly. Reliability engineers and maintenance managers will utilize AI to support PdM efforts to mimic human behavior, including decision-making.
Sensory information from the system can be analyzed through machine-learning (ML) algorithms using advanced analytics software. ML employs statistical methods (mathematics) to assist machines in learning by experience (empirical observation). The machine develops concepts from these patterns in the data, meaning ML gives computers the ability to learn without being explicitly programmed. Sensors, AI/ML, and specialized software enable automated preemptive corrective maintenance actions to be performed.
Industrial items, including engines, motors, gears, compressors, turbines, and the like, can be sensed for information, such as temperature, vibration, humidity, sound and noise levels, and/or rotational or linear velocity to detect wear or apparent anomalies. Thermal imagers may be utilized to detect radiation from an object using Infrared (IR) thermography. Such cameras may be able to detect abnormalities in temperature. Temperature measurements outside of the ordinary can indicate a potential malfunction. IR cameras must be pointed at the correct location and are generally expensive. Due consideration should be given to where they are most helpful.
Acoustic sensing and monitoring may help detect leaks. Ultrasonic technology can be employed to detect mechanical issues such as friction or stress. Vibration sensors pick up vibration patterns indicating wear on internal parts such as bearings and gears. It can also generate information related to rotational or linear misalignments.
In industries, homes, and cities, PdM employs collected sensor data to help engineers correct problems before they arise. It is apparent that PdM will continue to provide companies with the edge they need to address their systems before breaking, failing, or running errant. PdM will help provide the competitive advantage needed to meet tomorrow’s challenges.
- Artificial Intelligence
- Industrial Internet of Things
- Machine Learning
- Predictive Analytics
- Predictive Maintenance