Lately, my daughter and I have been binge watching doctor shows. In most episodes, patients are wheeled in presenting symptoms that need immediate attention. The doctors and surgeons work the obvious symptoms but sometimes prescribe treatments that don't address or, even worse, even mask more severe, underlying problems. Our heroes, Dr. Sean Murphy in The Good Doctor and Nurse Nic in The Resident, save the day by picking up bits of patient context from a conversation with a family member, an overlooked x-ray, a buried report or some other stray clue that leads to a brilliant moment of clarity and a prescriptive recommendation that saves the patient's life.
Every time I finish one of these episodes, I can’t help but think about the analogy between industrial equipment health and patient health.
Doctors can diagnose patients using sensor-based readings -- temperature, blood pressure, pulse and oxygen saturation -- but it is difficult to predict outcomes or prescribe a course of action without patient context. Patient context is the information that describes how someone lives and the traits they have inherited. How often do they exercise? Is there a history of diabetes in their family? Do they smoke? If so, how much? Do they work as a software developer or a carpenter? For Nurse Nic to treat the “whole patient,” she needs a 360° view of the human; snapshot and historical records of vitals along with correlating information describing what has happened to the patient over time.
In heavy industry, it can be a struggle to create a complete picture of a machine’s health. Mostly, we take action based on “snapshot information” and wind up treating symptoms after they have already affected machine and system performance. This snapshot data is usually time series in nature and derived from equipment control systems. This machinery snapshot data is analogous to the vital signs that a nurse takes when a patient is admitted to her ER. It’s important information, but it is not enough context to make decisions about care.
Similarly, to treat the “whole machine” you need to know who built it, how it performed at the factory before it shipped to you, how it has been cared for, what kind of parts and fluids has it consumed and what systems it is a part of. This equipment context must be merged with sensor and control system based performance data to derive a 360 view of the machine before prescriptive analyses can reliably be performed.
Similarly, to treat the “whole machine” you need to know who built it, how it performed at the factory before it shipped to you, how it has been cared for, what kind of parts and fluids has it consumed and what systems it is a part of.
At a recent OSIsoft Regional Seminar, I listened to a fantastic presentation by a large oil and gas exploration and production company. Reliability of their fleet of electric submersible pumps (ESPs) is really important to them and, using the PI System they've built some valuable applications to monitor the performance of wells with ESPs. The speaker talked about how their next big challenge with well surveillance is to integrate non time-series (he called it reference data) with the ESP data models they've already built using the PI System’s Asset Framework. He said "we don’t want to just show users the data - we want to give them answers." In other words, he wants to treat the ”whole machine” so he can predict outcomes and prescribe a course of action - just like Dr. Sean Murphy.
The following sketch illustrates this challenge. To treat the whole patient or the whole machine, experts need to do more than just look at vital signs or time series data between 0 and 90°. The contextual data from 90° to 360° needs to be connected and must remain up to date in order to give a complete 360° view - with no blind spots.
Not having a 360° view limits the depth and breadth of questions you can ask about equipment health. The bigger your equipment context blindspot the less likely you are to get answers to questions like:
What industrial equipment is under your care? Who are your favorite patients and what are your goals for them? Do you have enough ‘patient context’ to treat them as ‘whole machines’ or are you still treating symptoms?