Terzetto of Industrial AI Orchestra

Three is an influential number in symbology and considered as sign of perfection in mythology. Apparently, three is also the smallest number of elements needed to create a pattern. Human brain, which is a pattern completion machine, can grasp visual content as combinations of elements, colors and fonts.?

Artificial Intelligence (AI) is often motivated by study of neuron functions in the brain and neuroscience in general. As AI bridges the gap between humans and machines, three key buzzwords—predictive, prescriptive, and generative AI—are gaining attention as the trio of Industrial AI symphony. Despite being distinct concepts, the three technologies are raised from a common digital gene - Data. This article aims to delve and provide insights on this trifecta of Industrial AI technologies.

Digital Oracle in shopfloor

Predictive AI serves as a digital oracle, foreseeing behavior patterns of machines and processes by leveraging historical data and machine learning algorithms to predict future outcomes accurately.

For instance, in the case of turbomachinery, predictive AI enables proactive maintenance by analyzing historical and real-time data to interpret data patterns by using clustering algorithms to predict machine anomalies. This technology provides early warnings for potential issues, minimizes unscheduled downtime, leading to improved maintenance planning, reduced costs, and sustained equipment operation. While predictive AI excels in uncovering hidden patterns and connections within extensive data indiscernible for humans, it's important to acknowledge its limitations that can arise from potential biases in training data and that the technology may find challenging in handling unforeseen conditions or abrupt behavioral shifts in process operation.

Digital Doctor for Industrial Operations

Prescriptive AI, often likened to a digital doctor, recommends optimal actions based on predictions by utilizing algorithms and machine learning models to simulate scenarios and predict potential outcomes. It not only anticipates what will occur and when, but also why, offering decision options to leverage opportunities or mitigate risks.

For instance, in the context of process engineering; for heat exchange equipment, prescriptive analytics monitors cooling durations of fluids and prompts maintenance when necessary, ensuring optimal performance. Considering current asset operating conditions and loading, prescriptive analytics can provide insight of what would be the remaining useful life of the asset before it must be taken into maintenance. Prescriptive AI incorporates both structured and unstructured data, predictive analytics, optimization, and business rules (heuristics). The solution through deep learning neural network can continuously learn and improve prediction accuracy and decision options considering various variables and potential outcomes simultaneously. While it excels in complex scenario analysis, it's essential to emphasize that prescriptive AI serves as a decision-making aid and should not entirely replace human judgment, especially in critical situations.

Wizardry in Manufacturing

Generative AI is the digital wizard in the realm of AI. While predictive AI forecasts and prescriptive AI makes decisions, Generative AI focuses on creating entirely new content using machine learning models. It's the force behind chatbots like ChatGPT, capable of engaging in human-like conversations and crafting original and revamp designs for optimized engineering.

ChatGPT operates as a large language model (LLM) employing a transformer architecture, specifically OpenAI’s generative pre-trained transformer (GPT). When given a prompt, it dissects the text into tokens and analyses their relationships using an "attention" mechanism, akin to interconnected nodes in biological neural networks. Trained on extensive text datasets, GPT develops parameters to discern patterns and relationships, enhancing its language processing capabilities. Below is an example.?

One of the prevalent areas where Generative AI for Industries is picking up fast is in development of chatbots. These AI engines simulate human-like conversations, offer personalized recommendations, and can potentially automate various tasks in the manufacturing industry. Many end users have adopted and acknowledged operational efficiency improvements, enhanced customer service and satisfaction through a more interactive platform for internal and external users.

Another area that has shown promise is in development of process design and simulations for manufacturing operations. Various proof of concepts and studies indicate that Generative AI based on language models is capable of translating text prompts into designs, creating design variations, converting designs into instructions, and assessing Process design and Operational performance. In engineering design, Generative AI can improve engineering data inputs, help scenario creation, drive process design optimization, and generate synthetic data to enhance accuracy of conceptual design cost estimation and decision making.

It is with little doubt that Generative AI’s adoption can reduce time to market, while providing a competitive edge aligned to customer requirements. While the technology shows promise, human understanding remains crucial, and in current context these tools can only complement, not replace, engineers' skills in engineering practice.

Unlocking the Power of Hybrid AI

Combining technologies such as Predictive, Prescriptive, and Generative AI unleashes their full potential, also known as hybrid AI. Along with these AI triads used in combination, a popular approach also involves integrating simulation with these technologies. This integration enables soft sensing inputs to enhance the measured data, providing a more comprehensive predictive model with estimated unknown factors and quantified uncertainty in projections. Simulation, being deterministic, requires lower data acquisition and processing power compared to pattern recognition, as it leverages available data supplemented with knowledge, eliminating the need for exhaustive data searching for correlations. Moreover, simulation developed with expert’s process knowledge of cause and effect offers a reliable prediction accuracy thereby helps reduce false correlations common in big data analysis. The combination of AI technology with deterministic simulation has been successful in various scenarios and case studies, demonstrating its effectiveness in enhancing predictive models and decision-making processes.

For process industries, the combination of Predictive, Prescriptive, and Generative AI shows tremendous promise. For instance, in oil and gas production facility, the hybrid AI approach can be used to predict equipment failures (Predictive AI), recommend assess remaining useful life of assets advise optimal operating conditions to maximize output while minimizing energy consumption (Prescriptive AI), and help generate synthetic data for what-ifs, scenario planning and HAZOP analysis (Generative AI). They can provide a comprehensive approach to improve operational efficiency, reduce downtime, and offer better decision-making in complex process industry environments.

Synergy of AI, Data and Human skills

A significant factor in the success of Digital adoption is the interplay of AI, Data and Human factors. As Artificial Intelligence continue to expand in industrial applications, it's important to recognize that the key drivers behind their effectiveness are data and human inputs. The relationship between AI and data is synergistic: AI relies on large volumes of data (Big data) for robust model training and enhanced data analysis. This enables AI to provide more accurate predictions and valuable insights for Industry. It's clear that without substantial data, AI's potential is limited, and without AI, much of the value in data remains untapped.

It is also essential to emphasize that human understanding remains integral in this equation. Even as AI tools provide suggestions, it's crucial for engineers to comprehend the underlying reasons, especially in cases where AI may generate less optimal recommendations based on insufficient or inaccurate datasets. Role of human skills in discerning the AI outputs is vital for preventing potential shortcomings in system analysis or design. These tools are not meant to replace the skills of Industrial workforce but rather to empower them with greater efficiency, flexibility, and deeper insights for optimizing manufacturing operations.

Conclusion

As we ride the AI revolution, it's evident that predictive, prescriptive, and generative AI technologies among others will assume increasingly significant roles in various aspects of our lives and businesses. Yet, while we embrace their potential, it's crucial to remain mindful of their limitations and ethical considerations. In this AI-driven era, industries who can adeptly harness these technologies—comprehending their strengths, constraints, and interactions—will be best positioned to prosper. The industrial future has arrived, and it's being shaped by predictive, prescriptive, and generative AI.