Advanced Analytics for the IIoT

In this chapter, we will cover the concepts of advanced analytics and how it helps in IIoT use cases to provide insights, and to efficiently run manufacturing process and control systems for better productivity. Analytics that come under the category of complex data analysis, using techniques such as machine learning, are referred to in this chapter as advanced analytics. By the end of this chapter, you will have learned some of the IIoT business use cases and how machine learning techniques help in solving them. From the technical side, we will cover machine learning concepts at a high level, and frameworks such as Apache Spark ML and TensorFlow, as well as other tooling that is available to us.

In this chapter, we will cover the following topics:

• IIoT business use cases and analytics
• IIoT analytics classification—reliable, efficient, and...

IIoT business use cases and analytics

In this section, we will discuss some of the IIoT use cases and the analytics needed for measuring the outcomes. Some of them are common and are used across industries including mining, aviation, oil and gas, power, healthcare, and semiconductors, and can be broadly classified into areas such as asset performance, operations optimization, and supply chain management.

In general, enterprises measure the performance of plants and run them efficiently by monitoring the health of the equipment, as well as business needs and targets. The following are some of the measures that provide an ability to get an insight into plant performance and efficiency:

• Asset reliability and availability
• Monitoring mission-critical events
• Reducing unplanned downtime
• Optimization of the manufacturing process
• Optimization of fleet operations

...

IIoT analytics types

IIoT analytics primarily falls under the following types:

• Reliable analytics
• Efficient analytics
• Profitable analytics

Reliable analytics

Reliable analytics provides insights into metrics such as failure events, and maintenance events over a period of time to determine the reliability of the plant. Each process of the plant collects a set of metrics that feeds to continuously running analytics, so that the respective analytics can identify anomalies and generate alarms and notifications. Based on the notification, the plant operators can schedule maintenance events for tuning the system, with or without any downtime.

...

IIoT analytics – cloud and edge

Based on the IIoT use cases, the analytics need to be executed on the edge device closer to the controller so that it can react using a standard operation procedure based on the mission -critical events. On the other hand, there are analytics that typically run for a longer time and need to process large amount of datasets, which is at a fleet level. Such analytics fall under big data analytics and require high computing power to distribute the data computation generally preferred to run in a cloud environment for scalability and cost reasons. In this section, we will discuss both the cloud-based and edge-based analytics and technologies that can be used.

Cloud-based analytics

As mentioned...

IIoT data for analytics

In this section, we will discuss some of the core datasets of IIoT, such as time series, and asset data, and its role in analytics.

Time series data

Sensor and control systems are the core elements of the IIoT that produce enormous amounts of data through which we can determine various aspects of the physical system in various dimensions, such as performance, health, and anomalies. Typically, the sensor data is stored in a time series format. For example, temperature sensor reading at a timestamp of 12p.m. with a value of 15° can be referred to as a tag (temperature sensor), timestamp (time at which the reading happened), and value (actual value observed) respectively. In IIoT, there will be thousands...

IIoT analytics – architecture

In this section, we will discuss architecture for building analytics in IIoT, using popular open source technologies as a reference for running the analytics at ultra scale on the cloud and on-premise.

Big data and analytics – technology stack

Based on the needs of the business, the big data platform needs to be run in various deployment models, such as public and private clouds, or sometimes it needs to be run on the same stack in a customer environment. Occasionally also, platforms need to be built that are cloud agnostic, so that the stack can be portable across the cloud environments. In various business cases such as this, it is always good to have a strategy of building...

Advanced analytics – artificial intelligence, machine learning, and deep learning

Advanced analytics is classified as the set of analytics that requires complex statistical analysis, physics-based models, neural networks, and so on; in other words, the analytics that falls under the category of artificial intelligence (AI) for building machine learning models and predicting outcomes based on the machine models using observed data. AI-based analytics are all about learning from the observed data and eventually predicting the outcomes for the new data, or classifying the data based on the models built using the knowledge-based systems. In this section, we will primarily discuss machine learning and deep learning methods in building the AI algorithms.

Building a model

...

Summary

In this chapter, we covered the concepts of advanced analytics and how it plays a major role in the IIoT use cases for running the industrial equipment and manufacturing plants efficiently. We also discussed some of the real-world challenges in running the legacy analytics at scale and how big data plays a major role in building better models to predict the outcomes with more accuracy. We also looked at some of latest technologies, such as Apache Spark ML, the Keras framework, and the APIs available for the development of machine learning models. In the next chapter, we will look at how to develop an IIoT application.