Despite criticism of being overly hyped, IoT technology remains top of mind in many organizations and will continue to dominate conversations and drive new investments, and for many good reasons. There are numerous areas where Industrial IoT (IIoT) offers clear and significant business value potential. One of the better articulated use cases is service lifecycle management (SLM) and, in particular, remote monitoring and diagnostics.
The annals of equipment service reveal that remote diagnostics and telematics in general are anything but a new concept. In the 80s and 90s, IBM, Digital Equipment, Xerox and other product companies build equipment capable of “phoning home” using commercial telephone lines to report a malfunction, and remote access to enable equipment troubleshooting.
Why IoT, Why Now?
Modern IoT technology offers multiple advantages and capabilities not easily available to remote diagnostic devices of past decades. Wireless TCP/IP communication is available virtually anywhere around the globe at very low cost, and cloud services offer globally distributed storage and computing resources. Essentially, the entire communication infrastructure is owned by third parties that manage access, devices and data security, allowing service organizations to focus on the content rather than on setting up and manage the conduits.
Not only is the cost of setting up and managing the communication dropping, but also the cost of sensors, data acquisition and communication hardware continues to drop, making instrumentation and communication affordable.
Practically every piece of equipment is becoming a smart data-collecting node in an always-connected network. Secure connectivity and data exchange are no longer a challenge; they are a commodity.
But is it Enough?
Since access to a wealth of operational information from fielded assets is no longer a significant hurdle, is the problem of remote diagnostics solved?
In so many ways, the problem only begins, as companies need to know how to convert the deluge of data flowing from multiples sensors installed on fielded equipment into actionable and prescriptive diagnostic knowledge.
Detailed troubleshooting information such as symptoms and diagnostics tests—whether obtained remotely or not—has to be analyzed and applied in the broader context of the machine’s configuration, failure mode, maintenance history, and other types of external heuristics.
Granted, the ability to receive alerts and perform diagnostic tests remotely offers many savings. Sometime a calibration or software update can be performed remotely, remedying a malfunction without having to dispatch service personnel.
But the task of providing service technicians with accurate, detailed and task-specific information requires more than mere remote access and remains a critical cornerstone of effective and efficient equipment service operation.
It seems that no discussion of industrial Internet of Things is complete without describing the magic of predictive diagnostics.
Predictive diagnostics models, machine learning and other techniques that attempt to extract knowledge from complex machine data and provide proactive service advice are difficult to build and maintain. One of the more interesting and complex challenges stems from the broad variability in the installed equipment, even among similar pieces of equipment. A couple of examples will illustrate this point.
Consider a fleet of trucks rolling off the assembly line and delivered to different operations. Some of these trucks are used for long distance cargo hauling, covering great distances cruising long hours at highway speed. Other trucks make short trips, some in urban areas and frequently in start-stop traffic. Over time, the different traffic conditions, cargo loads and even the operator’s driving patterns cause these trucks to wear differently. Add to those the inconsistent service and maintenance practices that often do not follow the manufacturer’s recommendations, and the trucks are no longer close facsimiles of the original truck that was used as the model for the predictive data analysis.
Of course, the same challenge exists in stationary equipment. The diagnostic “signature” of a diesel engine used to drive a generator infrequently and at a constant speed will look very different than that of an identical engine that operates long hours under variable load. The as-maintained configuration of like-products and field-installed options are as critical to accurate analysis and diagnostic recommendations as the raw telemetry data.
Building reliable failure prediction models for highly engineered assets has proven difficult. These models require large data sets that are continually updated to reflect that ongoing changes caused by built-in variability, wear and tear, and configuration changes over the life of these machine.
Diagnostic Information and Product Lifecycle
Many product companies do not have reliable and detailed information linking early product lifecycle phases except for the slow flow of field service and warranty information. A pervasive network of IoT-connected devices can offer greater visibility and understanding of operation, utilization and failures not easily available otherwise. Manufacturing companies should integrate this rich information in the product lifecycle management flow:
- Provide access—real-time if necessary—to the entire installed base to improve the precision of the analytic and predictive models.
- Couple IoT generated information with detailed view of as-maintained, duty cycle and operating patterns, and maintenance history
- Continually update all service information, including predictive models to reflect changes such as new failure modes and frequencies, and failure modes eliminated through ECOs
Image: Section perpendiculaire du moulin des Verdiers (Jean-Jacque Lequeu, 1778)