While connected machines share both production and process information, it is the equipment reliability inspection sensor technology that may prove to be the most interesting and challenging element of the IIoT. The amount of inspection data that can be generated by inline monitoring sensing technologies is impressive. So, the challenge is managing the volume and variety of data, while deriving value from it. Deriving value is where it gets interesting.
Practitioners in vibration, motor current, thermography, ultrasound, and oil analysis (the five core PdM technologies) build up their knowledge, best practices, and ability to recognize tell-tale signs of common failure modes within various classes of equipment. When all the technologies are used together, and across multiple instances of common classes of equipment, patterns emerge in the larger population of data.
Further, combining the data of all technologies along with years of PdM domain expertise promises to yield amazing predictive assistance to those of us charged with improving production capacity while lowering maintenance costs. Deriving this value with the emerging data analytics technologies of IIoT is what this author finds extremely interesting.
Let’s take our best practices of inspecting our equipment with PdM technologies at a frequency greater than twice the expected arrival rate of equipment defects causing failure. In many cases, it is possible to use route-based inspection technologies, connecting our instruments to the internet after completing a day’s route. In other cases, due to accessibility, it is desirable to employ permanent sensing, including wireless or wired sensors. In even more complex cases, where operating conditions change frequently, it is desirable to employ intelligent data acquisition systems that are able to detect each operating condition, collect appropriate inspection data, and label the inspection with the operating condition.
Now that the inspection data is collected, the typical approach is to employ one or more trained analysts to review the data, using traditional analytical tools, to create an assessment report of the equipment’s health. Some analysts are so good at reviewing their domain’s data, the pattern matching analytics between their ears helps to speed the process. However, not all analysts have the experience and training to make rapid interpretation, and as such, analysis becomes the bottleneck to deriving value from our highly-instrumented equipment. What is needed is an analytical engine that recognizes the tell-tale patterns in inspection data, coupled with documented domain expertise, including possible and likely failure modes of the equipment.
When data is assembled in a large computing platform (a cloud computing platform), inspection data from common classes of equipment, operating in similar operating conditions, tends to cluster into patterns of normal or healthy patterns and patterns of common failure modes. Our human analyst then is greatly assisted by cloud computing technology. The bottleneck separating big inspection data from value begins to diminish.
Adding to the data science analytics portfolio, we add our documented body of knowledge from our maintenance libraries. These libraries contain known failure modes, failure codes, and associated inspection technologies that detect the defects leading to equipment failure. Building on our inspection technology training (the training that produced the analytical power between our ears), we create expected pattern templates to help our data science analytics learn more quickly. This is known as “supervised machine learning.” The human analyst now has an apprentice, learning from real-world data as well as from our historical domain knowledge.
The combination is powerful. As owners/operators of equipment who are challenged to staff each facility with the appropriate manpower for PdM inspection and analysis activities, we are relieved of the associated stress knowing that proven data and analysis technologies are automating PdM on our behalf. Further, there is a human in the machine who we can talk to, who appears at our monthly meetings, and who is in charge of the work his apprentice is conducting.