The future of predictive maintenance Part 2: Advancing technologies
While the business case for predictive maintenance is compelling, widespread adoption faces significant technological hurdles. Current data transmission limitations, infrastructure costs, and processing constraints keep comprehensive PdM programs out of reach for many organizations.
However, two key technological innovations may drive the next wave of predictive maintenance adoption:
- Edge processing could significantly reduce data handling and transmission requirements
- Innovations in sensors can make sophisticated monitoring more accessible and cost-effective
Advances in these technologies can potentially evolve past incremental improvements. They can create a fundamental shift that makes predictive maintenance more effective and accessible to organizations of all sizes and across diverse industries.
This is the second installment of a three-part series exploring the future of predictive maintenance. Read Part 1 and Part 3.
Expanding capability with edge processing
Current data transfer methods lack the bandwidth to transmit the quantities of real-time data necessary to support predictive maintenance programs and it is cost-prohibitive to store such large amounts of data on servers. As predictive maintenance systems become more comprehensive, increasing resource requirements may continue to keep PdM unattainable for many organizations.
Edge processing holds the potential to reduce costs while implementing comprehensive physical asset monitoring. This could be achieved by handling more data processing and analysis before transmitting to a higher-level system.
Edge processing is a method of processing data closer to that data’s source instead of sending it to a centralized data center for analysis.
In this scenario, AI models on a sensor would perform analysis and determine what data is actually worth transmitting. The sequence may look like this:
- Sensors typically monitor and report "all normal" to the higher-level system.
- Sensors can switch into high-alert mode when they detect deviations from established patterns.
- Sensors would capture and transmit detailed raw data for further analysis.
For example:
A vibration sensor might only send full time waveform data a couple of times per day during normal operations. However, it will immediately begin streaming detailed information when it detects anomalies.
Dual applications for edge processing
As edge processing in predictive maintenance evolves, it could serve two potential functions:
- Equipment availability monitoring: give machines the ability to self-diagnose. A motor or pump would analyze local vibration patterns and communicate when a part begins to deteriorate mechanically. This would provide advance warning before failure occurs without paying to send a constant stream of dense vibration data.
- Real-time process quality optimization. Edge-enabled sensors would continuously monitor multiple parameters and immediately flag if conditions deviate from optimal ranges.
For example:
A dairy facility running a milk fermentation process discovers a required temperature wasn’t held long enough. The problem went unchecked for hours. The facility must discard hundreds of thousands of gallons of product. Edge processing can facilitate early intervention to correct the process and avoid costly waste.
Rather than streaming massive amounts of raw data to central systems continuously, the future may lie with intelligent sensors that process information locally and communicate only when necessary.
Overcoming trust barriers
In working with dozens of facilities to integrate predictive maintenance methodologies, a trend emerges: Data science teams and reliability engineers often resist processed data. They typically prefer access to complete raw datasets.
This occurs especially if they can’t verify that the edge processing is interpreting the data correctly.
Edge-processing sensors should demonstrate consistent, reliable results. As the technology matures, sensors can provide dozens of accurate predictions and work orders during a six-to-twelve-month proof-of-concept project.
If the systems can attain this level of trust, facilities with existing data analyst teams won’t have to stretch resources across new projects as their predictive maintenance programs scale. The goal should be to move from overwhelming data streams to actionable intelligence for faster, more informed decision-making.
Sensor evolution: Improvement and expansion of monitoring capabilities
Sensors provide comprehensive insights into equipment failures and help reduce downtime through early detection. Sensor monitoring is evolving rapidly due to:
- Technological advances
- Cost pressures
- Potential expansion beyond traditional industrial settings
Addressing limitations in these developing technologies should help PdM expand throughout the commercial sector and into consumer goods.
Multi-parameter sensors: Opportunities and challenges
Multi-parameter sensors integrate multiple sensing capabilities into a single device. This trend already produced:
- Vibration sensors that include temperature monitoring.
- Flow sensors that measure totalized flow, media temperature, and maximum flow rate.
Multi-parameter sensors become even more attractive if they include edge processing:
- Correlating multiple data streams locally would provide richer context for anomaly detection AI or data analysts.
- Less data transmission potentially reduces complexity and burdens on network infrastructure.
However, multi-parameter sensors typically require more sophisticated design and manufacturing, which may drive up hardware costs.
ROI will likely vary based on the sensing data and applications. Developing new products should hinge on whether the installation savings and operational benefits justify the higher sensor price point.
Energy harvesting sensors: Promises and limitations
Energy harvesting sensors capture ambient energy from their environment and convert it into electrical power to operate the sensor. For industrial settings, they can draw energy from:
- Kinetic energy from rotating equipment
- Thermal differentials
- Vibrations
Energy harvesting sensors can potentially reduce infrastructure and power requirements, especially on retrofit applications for:
- Machines spread across large areas
- Equipment in difficult-to-access locations
- Intermittent monitoring applications
For example:
Typically, a $100,000 predictive maintenance project can require another $100,000 to provide 24-volt power to sensors and run communication cables to sensor locations throughout a facility.
However, sensors will still need to communicate real-time data even if they don’t require power cables.
High data density and millisecond response times still require wired transmission. Bluetooth mesh networks or other wireless protocols would have to replace wiring, but they can become complex and expensive.
Wireless communication limitations will likely outweigh the power source advantages of energy harvesting.
The alternative would be battery-powered equipment. But, batteries typically only hold enough power to send isolated data points every few hours, not in a real-time stream.
Energy harvesting adoption in this space will most likely be driven by cost-effectiveness.
For example:
Will a sophisticated energy harvesting sensor costing several thousand dollars provide better value than a simpler battery-powered sensor costing a few hundred dollars and requiring periodic battery replacement?
Is the cost savings from not requiring power or communication wiring enough to offset the cost of a more expensive sensor and more robust wireless data transmission?
Competitive pressures driving sensor innovation
Over the next ten years, price pressure can force established sensor manufacturers to differentiate through intelligence. New manufacturers produce increasingly high-quality sensors at dramatically lower prices, particularly for basic measurements. To stay competitive, sensor manufacturers will likely incorporate more intelligence at the device level, such as:
- AI models
- Advanced analytics
- Edge processing
Predictive maintenance can potentially become standard in residential consumer goods and light commercial equipment.
This may occur as sensors become:
- Less expensive
- More intelligent
- Easier to install
For example:
Residential and commercial heating and air conditioning units contain rotating equipment. Vibration monitoring integrated with smart thermostats, could provide early warnings when equipment begins to deteriorate. A $5,000 emergency replacement can be reduced to a $500 service call addressing early warning signs.
Continue reading this series:
How to get started
The best predictive maintenance strategy is simple to implement and maintain. It also prepares a facility for the future. The best way to create a customized preventative maintenence program is to start with a proof-of-concept project. For a free consultation with a Solutions Architect, fill out the form below.
For immediate assistance, contact our service center:
info.us@ifm.com
800-441-8246