IoT Analytics Enabling Industry 4.0
Published on : Saturday 06-02-2021
Shylaja Sabbani and Arun Venkatesh elaborate upon how IoT analytics is helping manufacturers in building digital and resilient supply chains.
Manufacturing industry is about production efficiency. The three main questions that every manufacturer needs to answer – what to produce, how much to produce and how long does it take to produce. They face numerous challenges with increase in complexity of product and product life cycles, and demand for a lot size of one (customisation) being the main ones. Manufacturers are now leveraging Industry 4.0 powered by smart technologies such as Internet of Things (IoT), Machine Learning and Artificial Intelligence to answer these questions. IoT is one of the foundational steps towards smart production. Sensors provide real time information on assets, machines, production lines and processes. However, sensors just help in collecting the data. The real benefit is realised when manufacturers analyse the data in context of business data to derive meaningful insights. Many manufacturers are moving towards decisions based on IoT analytics to derive the most business benefits.
Preparing for the IoT analytics
Before we get into the topic of IoT analytics, we need to understand some of the capabilities of IoT platforms that enable IoT analytics. First is the concept of Digital Twin. Digital twin is the digital representation of physical devices. It enables users to view incoming sensor data in context of the physical device.
In most scenarios, the sensor data itself might not be usable. In the case of monitoring the level of coolant in a CNC machine, the ‘fill level’ sensor measures and sends the distance between the sensor and the surface of liquid. This distance can mean different things for containers of different shape and size. This distance is first transformed to available volume and is then used to analyse when it needs to be replenished. Hence, the second important part is transformation of data.
In a typical IoT scenario, data from multiple sensors is ingested at high speed, in high volume, and of different variety into the IoT platform. Noise is something that is inherent when we talk of such data. Any analysis done on this data is likely to mislead with false positive or negative events. Hence, it is essential to first cleanse the noise from this data before it can be consumed in analytics.
Another factor that makes sensor data usable is contextualisation of sensor data. In the same example, along with the size and shape of the container it is important to know what is stored in the container, sometimes it could be sugar, pulp, etc. Adding this business context to sensor data makes it consumable. The sensor data which is cleaned, transformed and contextualised, is now ready to be analysed.
IoT analytics
Analysis of the contextualised sensor data is done for various reasons which range from real time identification of exceptions to managing operations in a plant to making strategic decisions. It is a lot more than just a visualisation of data.
In-stream analytics, also known as real time analysis of sensor data, is done via the use of frameworks such as rules. This helps in identifying critical events and taking appropriate action. Say for example, if a machine is overheating, monitoring the temperature of the machine and setting up a rule to notify the operator or switch off the machine if the temperature crosses a threshold. This helps in ensuring efficient running of the machines without the operator having to do regular checks. Extending this further, the operator can also find the root cause of problems by looking at sensor data surrounding an event that occurred. For example, if a battery failure was registered, it would be interesting to have a look at the temperature fluctuations during the time period of battery failure.
The real potential of IoT data can be realised if this data can be analysed over a period. Say the average running temperature of the machine helps to predict the operating condition of the machine, thereby getting insights into failure modes and leading indicators for the same. This information helps the production engineer to monitor the production process and optimise the same. Usually this is referred to as time series analytics.
Operational analytics is used for monitoring and managing operations that have a shorter time horizon with a focus on tracking real-time operational processes. This data is generally aggregated data unlike the time series data. The aggregation helps in reducing the volume of data analysed and mainly focuses on the trend. This data when contextualised provides insights which help in improving the efficiencies of the machines. These are generally made available for the production managers to check the trends and compare the results. For example, in the case of the packaging machine, the trend of daily consumption of inputs such as dyes will give the sales manager insights on consumption trends such as daily consumption and seasonality in consumption. When combined with the business data of the location or business partner where the machine is located, the sales manager can now optimise the supply of dyes to different business partners.
Industry 4.0 is enabling supply chain practices to be responsive and adapt to the external situation. Based on the telemetry collected from the assets, manufacturers can collect a lot of information and combine it with the results of what-if-simulations. AI technology is used to come up with the predictive models and bring back the intelligence to the machines and making Artificial Intelligence of Things (AIoT) way forward for the manufacturers. These predictive models are based on sensor data such as tool speed, vibration, environment data such as humidity, temperature and business data such as material produced, tool used, machine type, machine operation. The results of these models are used in predictive analytics helping the manufacturers to adapt preventive maintenance compared to run-for-failure maintenance.
In more sophisticated use cases, images and video feed is used as inputs to predictive models to identify quality defects early on. With the advancements in technology to analyse the videos on the go, manufacturers are leveraging the advanced analytics (video analytics based on AI/ML) to understand the exact steps performed by humans or cobots and take the necessary actions. These are also widely used for operator training.
Strategic Analytics involves KPIs, dashboards, and analytical metrics for monitoring the long-term company strategy with the help of critical success factors. These dashboards are usually more complex and provide an enterprise-wide view on the business. For strategic analytics it is important to combine the aggregated data across multiple facets of the organisational information. For example, based on the usage behaviour of the cars collected from the sensor data, decisions can be taken on the design of the automobile which can be customised during the production of the car. This kind of information helps the companies to reimagine their business models like moving from OEM sales to subscription, from batch processing to lot size one, etc.
Discussion about IoT analytics is never complete without discussing Edge Analytics. Many businesses operate in areas with network connectivity issues or have use cases where immediate response is critical. They cannot depend on IoT to solve these business problems. In such situations, edge analytics is deployed, where processing is done at the source of data and only required data is sent to the cloud. Capabilities of edge analytics range from simple analysis of streaming sensor data to very complex predictive models deployed at the edge. Automation of quality processes in the production line via use of image processing and predictive models deployed at the edge is an example of the value that edge analytics brings in.
The world today has been brought to a standstill by a virus. It has posed supply chain challenges like never. Manufacturers are relying on technology to bounce back. No manufacturer wants to be left behind. The ability to analyse the operational parameters of the machines, track and trace the products, meet the required contractual agreements has been promising for the manufacturers, in building digital and resilient supply chains.
The views expressed in this article are those of the author and may not reflect those of SAP.
Shylaja Sabbani is Director of the Engineering Unit of IoT in SAP Labs Bangalore. She is responsible for building talent, product delivery and customer success. Her areas of expertise include IoT, Analytics, Retail & Fashion Domain, Customer Success, Product development, Partner and Stakeholder management. She has been a trusted advisor for many fortune 500 customers across multiple industries. Shylaja is a thought leader in the space of Career Transformation and has delivered multiple talks at NASSCOM, SWE and GHC (US and Bangalore).
Shylaja has taken up leadership roles for GHC for technical and soft skills. She has evangelised SAP products and strategy at multiple SAP customer facing events like Sapphire, ASUG, TechEd.
Shylaja is an alumnus of Indian Institute of Management, Kozhikode (Executive MBA) and Harvard Business School (Driving Digital Strategies).
Shylaja Sabbani (https://www.linkedin.com/in/shylajasabbani/)
Arun Venkatesh is a Senior Product Manager of IoT Product Management at SAP Labs India focused on product strategy, definition and customer success. He has diverse experience in agile product management, marketing and business strategy. His analytical ability combined with his customer centric mind-set and excellent communication skills has led him to deliver products and launch them in new markets in a complex cross functional environment working with industry leaders like SAP, Microsoft, L&T and startup ecosystem.
Areas of expertise: IoT, Edge Computing, Analytics, Business Strategy, Product Marketing, Marketing research, Go to Market Strategy, Process Improvement, Product Management, B2B marketing, Partner/Vendor Management, Market Entry Strategy, Data Analysis.
Arun is an MBA graduate from Carnegie Mellon University.
Arun Venkatesh (https://www.linkedin.com/in/arunvenkatesh1/)
