en español

Improving maintenance and safety performance with IoT data tools

Equipment, machines, roads, software and anything else that’s subject to wear and tear must have some type of maintenance regime to remain in reliable and safe operation. Within industrial manufacturing and production settings, traditional maintenance strategies have typically comprised two components.

One is preventive (or scheduled) maintenance, where equipment or facilities are inspected, serviced and repaired at regular intervals. These intervals may be based simply on time – once every six months, for example – or on another variable, such as a 12,000-mile service for a car. This is complemented by corrective maintenance, where equipment is repaired or replaced after wear, malfunction or breakdown.

Shutdown maintenance is another possibility, but most organisations seek to avoid this due to the costs involved with shutting down a plant. Sometimes, however, it is the only viable maintenance solution available.

In any case, the traditional approach is at best inefficient. Some equipment that’s replaced simply due to a schedule may actually have plenty of remaining operational life – and this represents a waste. Additionally, servicing at too-frequent intervals incurs unnecessary costs, yet failure rates may increase if these intervals become too long.

The strategy is also increasingly ineffective in today’s industry, where equipment has become more complex, and is run closer to its permitted operational limits to meet modern demands for productivity and efficiency. More timely and detailed insight is needed to ensure maintenance remains effective under these conditions.

The need for better maintenance tools is highlighted by the cost to businesses incurred by equipment failure. Back in 2014, the Aberdeen Group estimated downtime costs across all businesses at $164,000 an hour; by 2016, this figure had risen sharply to $260,000 an hour – a 60% increase in just over a year . Even an apparently minor plant failure can be surprisingly costly when all the consequences are allowed for. In addition to lost production, there may be spillages or equipment with spoiled product to clean, time spent by many staff in incident review meetings, efficiency losses on start-up and shutdown, unwanted redeployment of staff, and extra administration work related to reports, purchasing and accounting.

However, the advent of Industry 4.0 and the Internet of Things (IoT) is allowing equipment builders and users to benefit from more flexible, insightful and effective maintenance solutions. This article reviews these more advanced maintenance strategies, and how the IoT and its various components are enabling them. A real system example is given. There is also some discussion of safety, as maintenance and safety issues are often closely interrelated.

We then look at developments that are building the future of maintenance; for example, how the potentially overwhelming volumes of IoT data now being generated can be managed by suitable use of artificial intelligence techniques. We also touch on the concept of machines that can repair themselves, and see how virtual reality can be used as a maintenance training tool. Finally, we look at how these maintenance improvements can give machine manufacturers the confidence to offer users a ‘Maintenance as a service’ business model in which they offer not just a machine, but a guaranteed level of productivity for an agreed operating period. This can be a financially rewarding approach for the machine vendor and user alike.

Condition-based monitoring and predictive maintenance

Intervals for preventive maintenance, as mentioned above, are usually set by the machine’s manufacturer, and based on a statistical analysis of when parts are likely to degrade or fail, and need attention accordingly. However, a statistical likelihood of failure does not mean that a part will actually fail, either immediately or for some time to come. Alternatively, a part can fail long before it’s statistically supposed to!

By contrast, efficiency and reliability would be greatly improved if we could monitor the machine and its parts in real time, spot when a parameter strays outside acceptable limits, and take remedial action accordingly. We get the most out of every part, without risking a fault becoming a failure.

This process is realised as condition-based monitoring (CBM), where the condition of a machine is continuously monitored by looking at pre-defined parameters of the equipment. This reveals patterns that might indicate equipment failure.

A CBM strategy would start by identifying the key parameters to monitor; those that would define whether or not the machine is healthy. Some common parameters for monitoring include:

Vibration: Monitoring the vibration of equipment, usually bearing vibration
Temperature: Monitoring the temperature variation.
Oil Levels: Measuring the variation in oil levels of equipment.
Acoustics: Using ultrasound to detect changes in sound made by the equipment.
Motor voltage and current: Monitoring for nuisance corona, destructive corona, arcing and tracking.

Fig.1: Cantilever type vibration sensor

Defining these failure parameters includes determining their limits of acceptable operation. Historical data for each parameter must be gathered to establish a baseline of what’s normal.

Once these parameters are available for analysis, a failure model can be built, to spot deviations from these baselines. This is reasonably simple to set up if the combinations of parameter values that indicate failure are known. A set of rules for failure conditions can be defined, and classical data analytics and mathematics used to build the right model. However, if the causes of failure are not well understood, data science and machine learning will be necessary to develop algorithms that can spot significant patterns in the data.

Machine learning algorithms are classified into two broad categories:

Unsupervised learning. Algorithms that run on a data set with no human intervention. The result is a set of automatically identified patterns from measured data that can be mapped to equipment failure.
Supervised analysis. These are algorithms that are trained to detect the failure. They are given a subset of the data, which is already classified as a failure/not a failure. The algorithm learns from that and can then be run on the complete data to pick out equipment failure.

In practice, machine learning models must be constantly monitored for their effectiveness. Models often degrade with time and need reconfiguration or retraining.

Predictive maintenance (PdM) is related to CBM, but not quite the same . PdM is an equipment strategy based on measuring the condition of the equipment to assess whether it will fail at some future period, then taking appropriate action to avoid the consequences of that failure. The equipment condition could be measured using CBM, but statistical process control, equipment performance, or human senses can also be used.

Condition-based monitoring and the IoT

Although CBM offers great potential, it also has challenges. It can be expensive, due to the cost of adding instrumentation to machinery and connecting to it. This applies especially to already-installed plant. In the oil and gas industry, for example, the cost consideration meant that first-generation CBM systems focused only on vibration in heavy rotating equipment.

Extracting useful data can also be difficult. Even if some types of equipment can easily be observed by measuring simple values such as vibration (displacement or acceleration), temperature or pressure, it is not trivial to turn this measured data into actionable knowledge about the equipment’s health.

However, the situation is changing. Production and manufacturing systems are becoming costlier, as are the consequences of downtime. Meanwhile, IoT technology is allowing manufacturers and users to address the cost and technology issues more easily and at lower cost. Instrumentation sensors are cheaper, more rugged and reliable, and offer wider functionality, sometimes with multiple functions. There are also robust wireless protocols that allow data to be effectively gathered from the sensors into a local gateway for immediate analysis and filtering. It can then be transferred over the Internet to a cloud-based computing resource that can provide software as a service to users of all sizes. This software can store the data, and perform the analysis necessary to spot trends and identify potential trouble spots.

Fig.2: Predictive maintenance using a Datonis IoT platform – Image via Altizon

Example condition monitoring system – Siemens SIPLUS CMS and the Digital Factory

Siemens’ SIPLUS CMS is a CBM package that uses IoT technology as part of its system solution. The system is a step toward the digital factory, where all the players, including machines, products and people along the value chain become networked. CMS works with MindSphere, Siemens’ Cloud-based, open IoT operating system platform designed for analysing large quantities of data gathered from machine monitoring sensors to reduce downtime.

SIPLUS CMS records and analyses mechanical variables from machines, integrates them into the automation world, and provides decision-making aids to maintenance staff, operators, and management. The open system architecture and the efficient interaction between all automation components enable plant-wide condition monitoring of mechanical components across all levels. With this approach, control centres can closely monitor up-to-date status information. In the event of an anomaly, it is possible to quickly estimate how much longer secure operation is possible. Also, anomalies in a plant can be compared directly to the condition of surrounding components to determine whether an increase in temperature is an indication of a bearing overheating.

The SIPLUS system is implemented in a choice of three different packages, summarised below:

SIPLUS CMS 1200

SIPLUS CMS 1200 allows permanent status monitoring of critical mechanical components. The recorded data is analysed by the internal software of the CMS1200 and stored on the SM 1281 condition monitoring module. SIPLUS CMS1200 is fully integrated into the automation system via Siemens’ Totally Integrated Automation (TIA) Portal.

Functionality:

Parameter-based or frequency-selective data analysis
On-board analysis algorithms
Trend analysis
Communication of system and status messages
Time synchronization via LAN
Online data streaming of vibration data to the CMS X-Tools analysis software

Benefits include improved plant availability, early detection of mechanical damage to optimise assembly service life, and lower maintenance costs combined with minimum investment costs.

SIPLUS CMS 2000

SIPLUS CMS2000 is available as a stand-alone solution independent of the automation system. It can analyse, diagnose, and visualize signals without additional software. The modular design makes it easy to perfectly tailor the system to specific requirements.

Functionality:

Parameter-based or frequency-selective data analysis
On-board analysis algorithms
Trend analysis
Communication of system and status messages
Time synchronization via LAN
Online data streaming of vibration data to the CMS X-Tools analysis software

The CMS 2000 offers the same benefits as CMS 1200

SIPLUS CMS 4000

For monitoring both individual machine components and complex systems. SIPLUS CMS4000 is easy to integrate into any automation environment. The powerful CMS X-Tools diagnostic software also let you create and protect your own analysis models and integrate process data from the control system.

Functionality:

Recording of mechanical signals via as many as 180 sensors – synchronously and in real time with a sampling rate of up to 192 kHz
Process data acquisition via software IONs directly from SIMATIC S7, SIMATIC TDC, and SIMOTION
Data transfer to CMS X-Tools via TCP/IP communication

Benefits include simple integration with existing and new automation systems, quality assurance of production processes through flight recorder functions, and detailed analysis, diagnosis, monitoring, visualization, and archiving

MindSphere is the cloud-based, open IoT operating system from Siemens that connects your products, plants, systems, and machines, enabling you to harness the wealth of data generated by the Internet of Things (IoT) with advanced analytics.

Specialist predictive maintenance sensors

The article so far has discussed condition monitoring systems based on machine-mounted instrumentation such as vibration, rotation and temperature sensors. Other specialist sensors, however, are also available to service technicians to provide further insight into their plant’s condition and any developing problems.

Thermal imaging cameras:

Fluke offers three ranges of thermal imaging cameras for different monitoring applications:

Performance Series, as shown in Fig. 5, is for routine maintenance. It provides precisely focused images from as close as 15 cm with manual focus, while faster images can be provided using fixed focus. The cameras can generate reports and email them from the jobsite, using Fluke Connect remote diagnostic software.

Professional Series is for advanced inspection and troubleshooting. It captures clear, accurate images of high-temperature components up to 1200°C. The products monitor processes with video recording, live video streaming, remote control or auto capture.

Expert Series provides premium image quality and full 180-degree articulating lens for easy navigation round touch shots. The image shows small details and allows discovery of anomalies on a 5.7” tablet-sized touch screen.

Lux meter: Increasingly important in the workplace, not for machines, but for safety concerns . The meters are used for measuring brightness in lux, fc or cd/m². Some lux meters are equipped with an internal memory or data logger to record and save measurements. Many lux meters include software for detailed analysis and offer different interfaces for transferring measured data to a computer.

Chemical monitoring: Water used, for example, in cooling systems must have the right chemistry to ensure corrosion, deposit and microbiological control. SUEZ’s TrueSense for cooling water treatment continuously measures and applies the amount of chemistry needed for these factors. The system’s chemistry measurements include:

Orthophosphate for steel corrosion control
Polymers for the prevention of deposits from mineral scales and dispersion of suspended solids
Free halogen for the cost-effective control of microbiological growth

The system interfaces with Insight, SUEZ’s cloud-based knowledge management solution. This allows visualisation of current conditions and trends, diagnosis of problems and identifying opportunities for improvement, and generating alarms for events or trends before they threaten production or assets. The system reports on key performance indicators and their impact on business objectives.

Intelligent AC current clamp – CT clamp:

The Pressac Sensing version 3 Wireless 3 Channel CT Clamp is designed to measure and report the AC current flowing in 3 separate channels. Powered from any of the measured conductors, the measured current in all 3 channel is reported every 30 seconds using wireless communications. The Pressac Sensing version 3 Wireless 3 Channel CT Clamp is easily installed with no disturbance to the measured conductors.

Fig.3: Pressac Three channel CT clamp – Image via Pressac Communications Limited

How the IoT can contribute to plant safety

Improving maintenance strategies is ultimately about boosting plant productivity. This productivity can be further increased as plant managers fulfil their legal and moral obligations to optimise on-site safety. And plant safety and security can be enabled by IoT technology combined with big data analysis . KPIs such as employee absences, vehicle mishaps, property damage, near misses, injuries, and any loss or damage that happens during the course of normal daily operations can all be monitored.

Often, if left to human reporting alone, many of these metrics can slip through the cracks as they are either unreported or under-reported. IoT enables better safety overall by ensuring real-time insights into these key areas. Any issues that arise can be addressed immediately, assuring compliance with health and safety regulations and environmental concerns.

Workplace injury is a good example as minor injuries are often unreported. Sometimes, they go on to become bigger issues over time, but the conundrum is how to be able to connect a larger problem back to a previous incident.

IoT wearables can provide a solution to this problem as employees will be monitored constantly for various health metrics, including heart rate, movement, activity, fatigue, stress, and so on. They will also provide a means to deliver important safety information, thereby reducing insurance costs for liability and improving compliance throughout the workforce.

Digital tagging can also help keep tabs on the workforce. Specifically geared towards high-risk industries such as mining, tagging technology lets management know exactly who is on the job site, how long they have been there for, and ensure that nobody is forgotten or left behind in case of an emergency.

The future of maintenance

The maintenance function is evolving from being reactive, to planned, to predictive, to finally self-fixing. The trend is to ensure production managers achieve ever-better plant utilisation while needing less maintenance effort to do so. Against the general uptake of IoT technologies, some particularly interesting strands as described below stand out:

Self-healing machine tools: Researchers at the Technical University of Denmark (DTU) have developed robust methods for automatic compensation of wear in industrial machine tools. This essential step toward self-healing machinery is being developed in collaboration with Siemens in Germany.

Present computer control of machine tools (CNC) obtain high precision in conditions when the machine is new. Position control algorithms ensure that any single piece of a production, e.g. 10,000 die-cut metal components, has precisely the same dimensions and finish as the rest.

However, precision is no longer the only necessary feature for an industrial machine tool. Today’s expectations also include high reliability to foresee equipment degradation and prevent unplanned production stops.

Accordingly, DTU has now formulated algorithms that make the machine tool itself capable of indicating when maintenance of various mechanical parts is due. Their research also allows the machine tool to perceive changes in the workspace and adjust its operating parameters to adapt to new conditions. This means that the production need not come to a total stop for re-tuning its parameters, but can continue to provide the required accuracy even under equipment degradation. DTU’s vision is to develop the automatic control for the machine tool to a stage where it is both able to compensate for gradual mechanical wear of the machine and also to advise on suitable time for maintenance.

AI’s growing contribution to predictive maintenance:

Another term related to condition-based maintenance is predictive quality and maintenance or PQM . PQM solutions harness data gathered both from the IoT and traditional legacy systems, and focus on detecting and addressing quality and maintenance issues before they become serious problems, possibly causing downtime.

Traditionally, PQM solutions used algorithms and came up with average statistics to predict when quality corrections or maintenance were required. However, new approaches are being made possible with the availability of much bigger data sets and recent developments in AI. These allow analysis of the actual condition of a product rather than just using average or expected statistics; applying AI technology to the concept of condition-based maintenance.

AI-based PQM solutions use several technologies in concert, including machine learning, deep learning, and cognitive computing:

Machine learning: focuses on real-world problems by processing—and learning from—large amounts of data
Deep learning: uses neural networks to be able to sort through nearly unimaginable volumes of data to come to conclusions
Cognitive computing: a subset of AI that attempts to mimic the way humans think. And a very important subset of cognitive computing is associative-memory learning and reasoning, which mimics the way humans learn, remember, and reason by making associations.
Complementary learning: Because each type of AI is good at solving different problems, applying them simultaneously is the key to success. Complementary learning in the context of PQM applications involves combining all these types of AI—machine learning, deep learning, and cognitive computing—to get insight into quality and maintenance issues.

In effect, a PQM solution that embraces complementary learning first uses machine learning and deep learning to answer the question, “What is the problem?” Then, cognitive computing answers such questions as: "Have I ever seen this before? What type of a problem is it? Who knows how to fix this? What caused this problem? And will it happen again?”

Using VR for maintenance in the factories of the future:

ARVRTech is an Augmented and Virtual Reality company that sees two key roles for VR within smart factories; to stimulate training of the workforce in maintenance activities, and to streamline factory processes.

Their content incorporates users’ interaction with their environment, and allows rapid, cost-effective and interactive training. It allows staff to learn about their workplace environment and machines without jeopardizing their life and safety, while also helping them to master expensive machines before they try them out directly.

One tool, their Interactive 360 photo, operates through a head-mounted VR display to guide users through different scenarios on the factory floor. Users are presented with two options – maintenance and training. Within the maintenance option, an array of short lessons opens up with prompts such as ‘check quality of steel’ or ‘replace bearings’. More detailed prompts, such as ‘use the induction probe to test thickness’ then appear, leading users step by step through the maintenance procedures.

More content, provided in a VR factory model, is very realistic, and helps visitors understand how different production lines inside factories are interconnected and which glitches in day-to-day operations should be avoided at all costs.

Fig.4: VR can be used in maintenance training for factories

Maintenance-as-a-Service:

A white paper, ‘The future of Maintenance’ published by Infosys suggests that a data-based approach will take maintenance to the next level and ‘Maintenance as a Service’ will soon become the norm. This approach will allow equipment to be monitored and fixed remotely and potentially even heal itself.

New business models will emerge wherein a product will be sold not at a fixed price, but instead on the basis of the throughput it can deliver in a given time. The manufacturers will take responsibility for the equipment and ensure that the users enjoys the best possible productivity from it while they receive an annuity and additional business based on product and service quality. Total cost of ownership will take centre stage as the key measure of success.

Conclusion

IoT technologies, with their large numbers of field centres connected to systems that can collect their data and perform sophisticated analysis, are giving new insights into plant and equipment conditions in real time. This is facilitating far more efficient maintenance strategies, based on what’s actually happening rather than statistics-based guesses as to what conditions may be.

These newer, more efficient maintenance possibilities are welcomed by factories seeking to remain competitive through ever-improving uptime.