Manufacturing has evolved. Organizations no longer need human intervention to manage every aspect of manufacturing. Today, an increasing volume of industrial activities are supported by robots, artificial intelligence (AI) and other modern technologies that enable organizations to get the most from every effort.
With this growth and change, comes an evolution in the tools that are designed to support manufacturing. Some of the biggest changes have been methods for keeping different manufacturing components functioning, and functioning well, for longer periods of time.
Predictive maintenance is one such improvement – which works by using the techniques further outlined in this reading to intercept possible failures and to fix them in order to avoid downtime that would otherwise be caused by failure. This then results in less time spent out of work, less money spent on repairs, and less human effort in managing all of these assorted processes – which together have been demonstrated to cost industrial manufacturers $50 million each year. For more on predictive maintenance, you can check out our predictive maintenance article.
3 classes of predictive maintenance tools
There are 1000+ tools enabling predictive maintenance in 3 categories. Tools in separate categories complement one another:
- Sensors generate data
- Analytics & monitoring tools help clean and analyze that data
- Scheduling tools coordinate teams in the field guiding them to carry out maintenance activities
Sensors have always been an important part of any maintenance plan because they allow us to monitor slight changes and make adjustments accordingly to prevent small issues from evolving into major problems. Having multiple different sensors monitoring different metrics can be key to getting a better understanding of your processes and preventing early failures and the resulting downtime they cause.
Regardless of what kind of sensors your organization requires in order to be successful, there are a few best practices to keep in mind during implementation.
- Ensure the accurate orientation and aim of any device by minimizing exterior conditions that could otherwise lead to incorrect readings
- Build a long-term imaging plan based on factors such as reliability demands, component-specific findings, budget considerations, manufacturer recommendations and similar
- Invest in training or consider bringing in outside assistance to ensure the correct usage of the tool by employees
- To get the most from your investment, be sure to take a baseline reading from which you can compare changes over time – this can ultimately help in justifying your initial cost to stakeholders
- Encourage full participation throughout your organization to get a balanced perspective from different levels of responsibility and expertise
- Remember that having multiple diagnostic tools working together can help to prevent a greater number of failures, and in the case of failure, to better pinpoint exactly what was the cause
- Ensure that your sensors automatically feed data into your analytics and monitoring systems
Over time, the data gained from sensors can be used together with other key analytics to help craft strategies that include seemingly disparate operations. Ultimately, this deep and detailed level of knowledge will have far reaching impacts on the business felt well beyond the manufacturing floor.
Sensors enabling vibration, sonic, and ultrasonic analysis
System components undergo normal wear, stress, and strain that are then indicated in its vibration and frequencies. Most components have a ‘normal’ frequency and deviation from this standard indicates conditions that may lead to failure if left untreated.
Unexpected vibrations can be fatal to a machine. In the highly technical sport of Formula One, for example, Honda’s engines faced unexpected vibration issues. These issues were so severe that the engines would literally shake themselves to death, failing (often spectacularly) in the middle of a competition. Even if it is not Formula 1, vibration related failures will still cause downtime.
Across practices in manufacturing and other industries, sensors feed information back to the systems connected to them. Fast Furrier Transform analyzers, for example, can detect minute vibrations that were previously undetectable. Once calibrated, a system will notice and record any unusual vibrations.
Vibrations can occur due to any number of factors. A machine’s bearings or brackets might start to lose their tactile strength. A component may be nearing the end of its lifespan. Upon analysis, technicians (if needed) or learning machines will determine the appropriate course of action, and when possible, take it as needed.
Some common uses for these types of analyses include:
- Ultrasonic leak detection that can be used further in advance than vibration or infrared. Infralogix has ultrasonic sensors that can detect sound waves beyond what the human ear can hear. This information can help technicians find vacuum seal failures, as well as air and gas leaks.
- Support condition monitoring by the early detection of friction between components
- In mechanical inspection, any changes in ‘normal’ sounds can be detected to prevent later failure
- Corona/electrical discharge
Thermal Imaging Sensors
Excessive heat is a death sentence for metals, composites and electronics. Excessive heat is a major threat to electric motors. Excessive heat is a primary maintenance concern for telecom companies. Dangerous working conditions and catastrophic delays can occur due to something as simple as a poorly lubricated set of bearings.
Thermal imagery utilizes infrared images to monitor temperatures of interacting machine parts – allowing any abnormalities to quickly become apparent. As with other change-sensitive monitors, they trigger scheduling systems which would then lead to the appropriate action being taken automatically in order to prevent component failure.
Simple thermal imagery equipment is easy to get and easy to operate. In its simplest form, technicians can take mobile readings with a handheld device. There is no downtime required for a simple handheld thermal image scan. The positives to this sort of predictive system are simplicity and ease. The downside is that constant observation is likely impossible with a handheld device.
A more sophisticated and accurate system would need diagnostic thermal tools with connectivity. Compared with baseline data, this equipment would show abnormal temperature ranges. These sensors would track the machines for potential deviations from acceptable temperatures. Once relayed, that information would alert technicians to any issues. This system would need greater capital investment and technologically competent staff.
Some common uses for thermal imagery that can benefit a predictive maintenance plan include:
- Process monitoring
- Electro-mechanical equipment
- Electrical power distribution systems
- Preventative facility maintenance in systems such as HVAC systems, buildings, roofs, insulation
Oil & lubricant sensors
Oil analysis can determine many factors of your machine performance. Actual oil viscosity versus expected viscosity can show how your machine is preventing oxidation, dilution, moisture, etc. Metal shards in the oil can alert technicians to parts grinding that might slow or break a machine. Sensors that calculate fluid dynamics might help expose a leak or faulty connector.
Oil analytics systems have been around for a while. Most modern cars have them integrated into the central computer system. Your car checking oil quality is a practical example of predictive maintenance.
These systems are not difficult to integrate into existing machines. You should have guidance from your lubricant provider on acceptable temperatures, viscosity, etc. You could cross-reference your actual results against the expected results. Analytic systems are commonly designed to detect impurities in oil. Metal, dirt and sludge will be easily found. Moisture is easily detected, even in trace amounts. Your system will calculate any aspect of the oil which could cause failure.
Monitoring and industrial analytics tools
Industrial analytics is often considered to be an integral part of the ‘fourth industrial revolution’, which is characterized by the convergence between traditional industrial practices and modern IT improvements. These advances include data analytics and their related interpretation via machine learning, and also advances in connectivity through IoT. What this means practically speaking is that a greater number of decisions and actions are starting to become based much more deeply on measurable data that can be acted upon quickly.
One important part of this field includes IoT sensors to monitor key changes in components. To meet the increasing demand for these technologies, a wide range of options are available to help industrial businesses find success – no matter the need or function. Monitoring tools work by utilizing advanced algorithms and machine learning in a way that enables them to take action in real-time.
Some examples of industrial analytics and monitoring in action include:
- Predictive maintenance on equipment, machinery, and assets
- Optimizing specific machine parameters
- Decision support systems
- Condition monitoring of assets
- Supply chain optimization
Once you understand the need you are trying to solve with monitoring tools, based on your pain points, such a solution generally functions in the following manner:
- Get data: sensors and data storage, programmable controllers, manufacturing execution systems, BMS, manual data, external data from APIs and similar
- Explore and clean data
- Enrich this data by connecting it with other meaningful and relevant data sets
- Visualization with the help of data science or data team tools that enable staff to understand and make use of the data
- Deployment of improved processes
These activities in practice can lead to larger and more consolidated data sets that can support deeper analytics and better decision making. Additionally, other overall benefits can be felt throughout the supply chain and order fulfillment processes owing to a better understanding of the individual components that create the most essential parts of any profitable organization.
A few best practices for ensuring the success of any IoT or similar analytical solutions include:
- Develop an effective IoT framework that is collaborative and enables usage of the right resources when needed
- Consider working in a cloud environment so that stakeholders across locations can get the most from the data
- Focus initially on connecting people, and then begin connecting things
- Adapt the tasks of functional groups and how they communicate between themselves so that they reflect the changes caused by IoT; use this info to find the right balance between external and internal resources
- Choose technology partners that understand the unique challenges related to industrial environments
The internet of things (IOT) and Industry 4.0 make predictive maintenance possible. The sensors and analytics are one part of the equation, another part is the actual maintenance work.
Software leaders like IBM, SAP and SAS create full range technology suites. These suites combine machine learning and the sensor data to compile maintenance plans.
Remember that predictive maintenance is about monitoring equipment and acting only when necessary. Technology programs designed for industry are honing in on when, precisely, action is required.
These available systems will automate much of the maintenance analysis. Your computer system will not be able to change parts, but it will be able to alert technicians of a pending issue. The programs will not create maintenance schedules, but rather, proactive behavior when a component faces the end of its life cycle. Even better, these systems can request maintenance long before a machine faces failure. When a machine starts to decrease in productivity or output, proactive maintenance can occur.
These modern versions of a traditional solution work by automating much of the maintenance analysis traditionally managed by a person. This person, who previously would have analyzed multiple inputs, ongoing processes, and other relevant factors needed for building an effective maintenance schedule, can then focus their energies on the results of any changes or adjustments that were made.
By having a scheduling solution take over this analysis, the time and resource requirement needed to take into account all factors drops exponentially. As this scheduling takes place without human intervention, it is important not to forget the ‘people’ aspect of any schedule in terms of general knowledge surrounding an operation, like in the case with a customer that is considered to be ‘high priority’ unexpectedly by a computer’s standards.
By utilizing the right scheduling tools for your organization, the following results can be realized:
- Assigning resources and schedule activities based on a wider range of external and internal factors
- Optimization of production schedules proactively, based on learned models in the past
- The ability to apply countermeasures much sooner when possible, increasing the changes to balance any issues that may arise
- Detection of bottlenecks in separate departments and practices that may be impacting other seemingly unrelated processes
Some best practices to help manufacturers achieve a successful implementation of any scheduling tool include:
- The adjustments of algorithms so that they are programmed to make adequate decisions quickly, rather than perfect decisions slowly – which may require increased use of approximations
- Deliver ideal production schedules that managers can choose from based on priority
- Determine ideal production and workflow velocity to achieve the right balance of quality and quantity
Working with predictive maintenance tools
For manufacturers and other industrial organizations, finding the best way to minimize waste and inefficiency can have a major impact on the bottom line of your business. When working together correctly, these tools act in a complementary manner – enabling the success of one to aid in the success of another. However, any decision made with regards to predictive maintenance should include inputs from employees and other relevant parties throughout the business to ensure that the functionalities required are the ones that are obtained.
What are leading predictive maintenance tools & vendors?
Sensors enabling vibration, sonic, and ultrasonic analysis:
#1 Texas Instruments
Texas Instruments offers multiple products with different capabilities. Their ultrasonic sensor solutions are able to measure flow, level, proximity, position, imaging and distance. The below image provides a product comparison of Texas Instrument’s ultrasonic sensor solutions:
#2 Cisco IoT solutions
Cisco IoT solutions enable businesses to digitize manufacturing so that they increase the visibility of machines to maximize productivity. With Cisco IoT solutions, data analysts can perform analytics at the edge to gain industrial insights.
#3 Intel IoT solution
Intel IoT solution provides real-time data analysis at the network edge, help manufacturers understand when maintenance is needed to minimize downtime and costs. Their industrial IoT sensors on the assembly line track key indicators of equipment failures such as vibration and temperature.
Thermal imaging sensor tools:
Flir and its subsidiary brands (Raymarine, Extech, Armasight) leverage thermal sensors for substation security and predictive maintenance. Businesses can use their integrated solution to frequently monitor assets and launch a predictive maintenance program that reduces the risk of asset failure and generates substantial savings by increasing uptime.
#5 Process Sensors
Process Sensors‘ solution supplies Infrared (IR) temperature sensors, thermal imaging camera systems, blackbody calibration sources and portable thermometers to use in industrial processes and research applications.
With Fluke infrared thermometers, manufacturers can get temperature readings from a safe distance.
Monitoring and industrial analytics tools:
#7 General Electrics (GE)
General Electrics’ Predix solution collects data from sensors and enables edge-based processing and analytics to reduce downtime.
Falkonry Analyzer is a portable self-contained engine that discovers and recognizes conditions from a live data stream to enable business monitor data streams at the edge. The solution can be deployed to monitor multiple similar assets or systems.
FANUC Zero Down Time (ZDT) solution provides predictive analytics to prevent unexpected downtime. ZDT identifies component failures and recommends proper intervals for routine equipment maintenance activities. With ZDT, businesses can monitor their manufacturing process remotely using a web portal that provides a clear picture of device health, equipment utilization, and energy consumption.
Hitachi’s solution, Lumada, enables businesses to calculate remaining useful life and extend asset life so that businesses can schedule regular production downtime for maintenance.
#11 Stottler Henke
Stottler Henke’s Aurora tool uses artificial intelligence to encode and apply extensive scheduling knowledge and rules to perform automatic resource scheduling. Its graphical user interface enables users to enter domain-specific scheduling constraints easily without coding knowledge.
#12 Honorable Mentions
Interested in learning more about how predictive maintenance, AI, and other advances could impact your industrial organization? We have a list of AI applications in operations, most of which are applicable to industrial organizations. If you are looking for AI problems or need to figure out your AI strategy:
How can we do better?
Your feedback is valuable. We will do our best to improve our work based on it.