Systems comprising humans and machines are complex, dynamic, and evolving. It is imperative to ensure pre-emptive health (and performance) upkeep of such systems.
Traditionally, people’s approach to their well-being has tended towards being reactive, though that is definitely changing. The same pattern tends to hold true in the case of maintenance of a machine. Until some years ago, other than regular health check-ups or service schedules, people did not believe in proactive action plans to ensure that they or their machines stayed healthy and in optimum working condition.
Stakeholders in healthcare well-being and machine maintenance – be it medical professionals for human health or service professionals for a machine’s health – were trained to react or respond to something when or after things went wrong. The focus has recently categorically shifted to the proactive upkeep of the health of a person, system or machine. This is because of the ability to harness and interpret data available across such systems of men and machines, coupled with the advent and disruption in technologies available for the same – pre-emptive, predictive or prescriptive sensing analytics and upkeep are now more efficient.
This is exemplified in the move from a reactive to proactive approach adopted by people to ensure their well-being; and there is an almost uncanny synchronicity in how we now approach the maintenance of our machines!
In fact, technology advancements and the imperative nature of this predictive approach for the upkeep of human and machine health brings other elements like utility, convenience, social and economic value into the man-machine systems mix.
This predictive approach to human healthcare and machine maintenance is gaining importance. There is a marked rise in its adoption and that is being aided further by technology disruptions like the proliferation of affordable, unobtrusive or non-invasive, internet-connected sensors as part of the current Internet-of-Things (IoT) revolution.
Some recent technology disruptions can actually add value to predictive maintenance. For example, in the case of human beings, vital parameters such as blood sugar, blood pressure and ECG can be checked constantly and this data can be fed to smartphones thereby making it accessible by medical professionals who are being consulted for the said purposes. Any abnormality in those parameters can be flagged off and medical professionals will take proactive action, thereby preventing an emergency – hospitalization, serious ailments, or in extreme cases, death.
So for human health, this means the availability of 24×7 physiological data from smartphones and wearables or implantable sensors. Similar systems can be followed for maintaining machines. For machines, this not only translates to the availability of machine-mounted sensors monitoring machine health but also monitoring through unobtrusive ‘nearables’. These ‘nearables’ or wearable-sensors can be in form of optical, thermal, acoustic, or radio frequency (RF), which can sense a machine’s physical condition without any interference or disruption of operations.
The other technology disruption is in the area of sensor-based predictive analytics. The biggest challenge is the choice of appropriate features from sensor signals, which is typically time-consuming as knowledge of both signal processing and data science is needed. However, as Artificial Intelligence (AI) matures, techniques like feature engineering, and deep learning may automate feature selection too, by analyzing available sensor data. Once the right features are selected from sensor signals, statistical analysis and rule based analytics can provide the desired outcome. For example, deep learning may help decide which feature of an ECG signal indicates the onset of a certain disease or which feature of a machine vibration signal is indicative of future machine failure. The AI field is also progressing fast in cognitive computing, enabling machines to automatically learn rules from a vast unstructured text corpus – this may include medical textbooks for human health and machine maintenance manuals for machine health. Deep learning and cognitive computing have the potential to disrupt the sensor predictive analytics space.
These technology advances, along with a business model that promotes wellness and proactive just-in-time maintenance, can change the current human-dependent ‘art of diagnosis’ to a machine implementable, repeatable ‘science of prognosis’.
