11 Industrial AI Trends that will Dominate the World in 2021

These trends broadly cover the three themes of: Where will businesses adopt AI in 2021? How will AI become more accessible? How will AI capabilities evolve?



By Swati Giri, Product Marketing Associate at Flutura

With the onset of the pandemic, there was one thing that multiple industries adopted overnight – digital transformation. Rapid transformation that was to happen in 5 years took place within 6 months. The accelerated adoption will continue through 2021 in the areas of AI, ML, intelligent automation, VR, AR, and such advanced tech-based solutions. 

The proliferation of digital transformation will be widespread. Moreover, it will also proactively prepare industries to manage future uncertainty, complex security and privacy concerns, the ethical use of Artificial Intelligence, and the increasing impact of climate change. AI will see omnipresent applications in 2021. 

These trends identified are from the trenches of the frontline employees of Flutura, that broadly covers three themes of Where will businesses adopt AI in 2021 - seeing exceptions in a dynamic environment, closed loop actions at the edge, making operations more autonomous; How will AI become more accessible - democratization of AI, explainable AI;  How will AI capabilities evolve - non intrusive detection of process deviations using video, reinforcement to learn from human experiences as well

 

Trend #1: AI Driven Dynamic Operating Models

 
Many industrial companies still have not accepted the reality of the rapid societal shifts happening across the globe. COVID will be seen by future generations as a significant inflection point in human history. Human behavior and human perception of opportunities and challenges will all go through a dramatic change in the years to come and companies must put in place a more dynamic operating business model. In a dynamic operating model AI will help spot exceptions sooner for companies to react in a dynamic environment and reduce the cost of decision making, increase quality and speed of decision making.

 

Trend #2: Acceleration of edge intelligence to avoid adverse events

 
Industries such as Mining, Marine, Oil & Gas Offshore operate in rugged environments with complex cascaded machinery.  With an increased sensor fabric to ensure safe, reliable, and efficient operations it is critical to establish closed loop actions to keep the operations in the most optimal operating envelope. Given the fast -paced nature of operations on the field and a challenged and cost- effective environment for high speed communication network, the idea of real “real-time” decisions has proliferated into these industries and will see an acceleration in the adoption of edge intelligence. A mining accident in Brazil led to flooding and loss of lives. The incident made the newly replaced CEO to institutionalize an edge intelligence team to detect HSE adverse events upfront. 

Figure

Fig 1: Edge intelligence on a  LNG carrier to continuously assess and predict containment risks or any significant event detection in advance giving sufficient time to respond

 

 

Trend #3: Rise of adoption rate for AutoML for reduced task time

 
Instead of letting the user guess which model to deploy, AutoMLs automatically compare models, deploys for faster and easier insights. In a post-pandemic world, where digital leaders are insisting on reduced time to impact, there is a movement from "handcrafted" AI models to "augmented AI models". According to Forrester, 61% of the data and analytics decision-makers firms that are adopting AI, have already implemented AutoML software in 2020. Google, Microsoft has released AutoML products primarily targeting AI workloads mainly for consumer use cases. While companies like Flutura are automating ML for enterprise AI.

Figure

Fig 2: A glance at Google AutoML Table for building and deploying ML models

 

 

Trend #4: Autonomous Operations for Increased Operational Safety and Efficiency

 
Increasingly the OEMs Controls and Automation players will seek collaborative alliances with Deep Sensor AI players to close the loop on automated- low-risk decision making. For high-risk operations, the decision-making loop will have a human to oversee execution. Companies will pick up the pace to look at every nook and corner of their operations to simplify and make it autonomous. The convergence of multiple technologies will enable this adoption and AI will take center stage. 

The use of systematic error propagation enables to learn from past measurements what are normal and abnormal states of the drilling system, therefore fulfilling the main requirements of an autonomous system.

Figure

Fig 3: Offshore drilling in Oil & Gas

 

 

Trend #5: Enrich first principle models with data driven models

 
The characteristic of industrial processes is the very complex interplay between variables and influencing factors. The first principle model alone has inability to gauge the secondary effects’ influence on the process. In the real industrial scenarios, observed behavior can deviate remarkably from the typical correlations we expect to see. Without application of domain knowledge, there can be incorrect conclusions based on ‘spurious correlations.’ While the data models provide real time recommendation of process parameters after accounting for variations, recognizes patterns and detects faults from historical data. With Hybrid models you overcome the disadvantages and combine the best of first principle and empirical models.  Finding the right combination of models to model the real-world operations is key to achieving the best results. Flutura has a unique readiness assessment methodology called AXON Pre-Cog Framework to choose the right models and assess viability of use cases for organizations.

Figure

Fig 4: The precise usage of first-principle models and data-driven models

 

 

Trend #6: Democratizing AI

 
Having experienced several deployments of AI solutions for Industrial Facilities, Flutura has concluded that empowering Industrial Engineers with data science and providing effective collaboration mechanisms between data scientists and domain experts is the key to deploying effective AI solutions to achieve the desired business outcomes. We foresee several new tools and technologies being launched in the market to address this. Flutura has been working on one such tool which we believe is the Industry’s first in this category called Engineer’s Workbench. 

Figure

Fig 5: Engineers creating heatmap by performing root cause analysis using Engineers’ WorkBench

 

 

Trend #7: AI explainability for deepening of trust

 
AI-ML models have a widespread application from a simple vacuum cleaner to medical diagnosis. The accuracy of prediction has ascended to 98-99% leaving most of the business heads impressed. However, understanding the reasons behind predictions is equally important in assessing trust in a model. To fulfill this requirement there are novel ML-models explaining the technique of predictions in an interpretable and faithful manner. These machine learning models provide the reasoning behind the predictions made without subjecting them to human interpretation. In the medical field, AI-ML is used extensively to diagnose complex diseases and even use technology for operating on the patient with precision. For this instance, AI-explainability provides an assurance to have a higher degree of trust as it provides them with the reasoning to arrive at such conclusions.

 

Trend #8: Integrating AI-driven microeconomics of engineering operations into their financial outcomes

 
Many industrial companies have not modeled the microeconomics of engineering operations into their financial outcomes. That is what is leading the companies to still take tentative steps towards the adoption of new technologies. This will change very soon given the highly competitive landscape, constant pricing pressures, and fluctuating demand cycles. Flutura has a pragmatic business case establishment methodology called AXON Micro Operational Economics Framework to identify and establish critical value levers and set ROI goals for AI investments. Next-generation integrated operations will see a highly microeconomic AI-centric operating model.

 

Trend #9: High-dimensional data analytics using Computer Vision

 
Technology has played a key role in facilitating changes enforced, by COVID-19, to overhaul everyday operations. Forrester predicts that in 2021, more than a third of companies in adaptive and growth mode will look to AI to help with workplace disruption for both location-based, physical, or human-touch workers and knowledge workers working from home.’ Computer vision provides a non-intrusive method of detecting process inconsistencies, quality deviations, safety non-compliance, surface inspection, asset downtime, and other event detections.

One of India's largest electronics manufacturers that has adopted the technology to automate PCB production monitoring and detect detection. 

Figure

Fig 6: High-quality image data of produced printed circuit boards (PCBs) is used to check for 20+ anomalies and defects.

 

 

Trend #10: Platform of platforms to have best-of-breed solutions

 
Companies rely on several vendors to implement digital solutions. The strategy lets companies assemble a best-of-breed solution by integrating products from several different vendors - software, platform, hardware, and infrastructure. The idea of platform-of-platforms is to leverage the best of the best, as well as customers' previous IT decisions and investments, and can focus on delivering the most valuable customer engagement solutions. 

 

Trend 11: Reinforcement learning for foolproof operations in an asset-intensive industry

 
With a combination of the field operators’ expertise, supervised and unsupervised learning, the AI models have reached an acceptable accuracy level in predictive and prescriptive analytics. The companies will now move towards reinforcement learning using experience-driven sequential decision-making. This method interacts with the environment to learn and drives decisions towards a goal that rewards the actions taken. Without the need to specify all the asset failure scenarios, the algorithm learns repeatedly by exploring all possible options.

Figure

Fig 7: Reinforcement learning - A feedback system learning

 

 

Conclusion

 
AI might not have predicted the pandemic, but in the future, it will help executives rethink the future of work considering systematic risk in the long-term plan; drive more efficiency, reliability, elasticity, and scale in operations. The growth of AI continues as we enter 2021 and the focus would not only be on new technologies and applications in the industry but also how it intersects with society. This is only the beginning and artificial networks will open upopen truly limitless possibilities year-on-year. 

 
Bio: Swati Giri is a Product Marketing Associate at Flutura, an Industrial AI and IoT company. She works on their flagship AI product Cerebra, which powers operational excellence and reliability for connected systems in industries like Oil and Gas, Manufacturing, Heavy Machinery, etc.

Related: