 "Internet of Things invasion")
The industrial uses of Internet of Things (IoT) devices promise to revolutionalise the way firms think about product maintenance, product development, and resource rationalisation in the times to come. IoT applications could be found in manufacturing, transportation, resource extraction, agriculture, and the military. Cases of industrial IoT use would involve designing optimal sensor systems that work reliably and are secure, handling the streaming data that would emerge from these sensors, and pragmatic economic assessment to ensure long-term business success.
IoT data currently seems to be put to limited use as suggested by some analyst reports. It is very likely that not more than one per cent of the data emanating from an oil rig is used for analysis. That is because this data is used largely to detect and control anomalies rather than for optimisation and prediction purposes. Similarly, data from devices embedded in machines and industrial products do not seem to be used for predictive maintenance that could lead to performance optimisation.
It is expected that more than 50 billion machines will be connected to the internet by 2020. To add to the complexity here, a jet engine could possibly have around 25 sensors each emanating data and a locomotive could have 250 sensors or so. Thus, the 50 billion machines with multiple sensors each streaming data would pose a challenge that even a distributed file system like Hadoop spread over multiple nodes could struggle to handle from storage and processing points of view.
The challenge the streaming data from IoT devices poses are several - developing intelligent filters that could process and tune out data not essential for gathering insights, developing an online analytic capability to process the filtered streaming data, integrating the processed IoT data with the enterprise data, and finding large storage capability for storing a large number of images and videos in addition to numeric and descriptive data.
Intelligent filters would be application-specific and are still a work in progress. In-memory analytics requires tools such as Spark, probably with significantly enhanced libraries to draw from and large sized memory banks. Integrating with enterprise data and hence adding few dimensions to the existing enterprise data could be a large programme management task by itself for sizeable enterprises. The ever-increasing storage demand would probably require the next breakthrough in storage technology to satisfy it.
What is likely to be the IoT architecture for firms to consider? A broad architecture of an IoT system is likely to include mechanical systems such as sensors and micro controllers connected to machines, gateways to collect and funnel sensor signals located either on cloud or on premise, an IoT application that takes in the streaming data, and outputs from an IoT application either populating dashboards or triggering workflow actions.
Thus, it seems appropriate to ponder if a complex, rather expensive, and still evolving schema of IoT is worthwhile for firms to bother with. Let us start with predictive maintenance. It has been argued and proven that preventive maintenance results in positive cash flow when compared with unexpected breakdowns resulting in repairs and delays to various activities.
Analysing product usage data would certainly give insights into future product development, avoid unnecessary experimentation with features and could even result in a lot more precise specifications for greater customer adaptations. The above examples, if followed, would result in higher degrees of resource rationalisation for firms and, if institutionalised, would considerably enhance their efficiency.
We have largely considered applications in manufacturing; however, productive applications exist in transportation and mineral extraction. As the IoT ecosystem evolves, using such a schema may become a de facto requirement. However, shaping this IoT schema and using it productively would likely bestow competitive advantage on the early movers.
IoT data currently seems to be put to limited use as suggested by some analyst reports. It is very likely that not more than one per cent of the data emanating from an oil rig is used for analysis. That is because this data is used largely to detect and control anomalies rather than for optimisation and prediction purposes. Similarly, data from devices embedded in machines and industrial products do not seem to be used for predictive maintenance that could lead to performance optimisation.
It is expected that more than 50 billion machines will be connected to the internet by 2020. To add to the complexity here, a jet engine could possibly have around 25 sensors each emanating data and a locomotive could have 250 sensors or so. Thus, the 50 billion machines with multiple sensors each streaming data would pose a challenge that even a distributed file system like Hadoop spread over multiple nodes could struggle to handle from storage and processing points of view.
The challenge the streaming data from IoT devices poses are several - developing intelligent filters that could process and tune out data not essential for gathering insights, developing an online analytic capability to process the filtered streaming data, integrating the processed IoT data with the enterprise data, and finding large storage capability for storing a large number of images and videos in addition to numeric and descriptive data.
Intelligent filters would be application-specific and are still a work in progress. In-memory analytics requires tools such as Spark, probably with significantly enhanced libraries to draw from and large sized memory banks. Integrating with enterprise data and hence adding few dimensions to the existing enterprise data could be a large programme management task by itself for sizeable enterprises. The ever-increasing storage demand would probably require the next breakthrough in storage technology to satisfy it.
What is likely to be the IoT architecture for firms to consider? A broad architecture of an IoT system is likely to include mechanical systems such as sensors and micro controllers connected to machines, gateways to collect and funnel sensor signals located either on cloud or on premise, an IoT application that takes in the streaming data, and outputs from an IoT application either populating dashboards or triggering workflow actions.
Thus, it seems appropriate to ponder if a complex, rather expensive, and still evolving schema of IoT is worthwhile for firms to bother with. Let us start with predictive maintenance. It has been argued and proven that preventive maintenance results in positive cash flow when compared with unexpected breakdowns resulting in repairs and delays to various activities.
Analysing product usage data would certainly give insights into future product development, avoid unnecessary experimentation with features and could even result in a lot more precise specifications for greater customer adaptations. The above examples, if followed, would result in higher degrees of resource rationalisation for firms and, if institutionalised, would considerably enhance their efficiency.
We have largely considered applications in manufacturing; however, productive applications exist in transportation and mineral extraction. As the IoT ecosystem evolves, using such a schema may become a de facto requirement. However, shaping this IoT schema and using it productively would likely bestow competitive advantage on the early movers.
The author is associate professor, Complexity Research Group, Institute for Financial Management and Research