How can AI take Process Industries into a new age?

Andrej Vasiljevic, Siemens | June 2022

Artificial Intelligence or AI represents an idiom that’s been with us for more than 65 years. It was initially coined and used by American computer scientist John McCarthy [1], as a part of a research workshop which took place at Dartmouth College in 1956 [2]. Yet, the idea of autonomous logic-based operations performed by a man-made machine had existed for much longer prior [3]. Over the past decades, the expression has slowly entered our daily lives through popular fantasy and science fiction; mostly cinematography as well as literary works of Arthur Clarke and Isaac Asimov, amongst others. In parallel, with the development of computing power, so did the application of AI within technology continue to develop. While there is no uniformly established definition of Artificial Intelligence yet, some of existing and acclaimed research would define AI as a “continuous evolving frontier of computing capabilities” [4], and “the science of making machines do things that would require intelligence if done by men” [5]. In fact, the definition of AI will depend on the outlook upon which it is approached.


A Smart Computer or Something More?

One of Artificial Intelligence’s main elements is defined as Machine learning—a set of algorithms and statistical models that helps computers obtain knowledge from patterns of repetitions, or “the computation paradigm where the capacity for solving the given problem is built by previous examples” [6]. There are 3 basic types of Machine learning: Unsupervised, Supervised and Enforced learning. The acquired data patterns or “wisdom” is then processed and used to predict future behaviors, optimize operations, and prevent potential anomalies. The next step comes in the form of Deep learning—a comprehensive structure of algorithms which empowers computers to function more independently and are modelled in the likes of human brain’s neural networks, while being capable of making data-driven decisions [7]. Finally, Cognitive computing [8] focuses on the use of computerized algorithmic models to simulate the human thought process with a set of subsystems. Amongst the main deployment areas of AI is the industrial sector, where hundreds of complex and interconnected operations occur each second. This specifically refers to industries with advanced application of automation, and primarily—Process Industries.

AI Meets Process Industries, But What About the Benefits?

Manufacturing constitutes a fundamental basic industry for economies, and as such remains a critical force that supports and propels national economic growth, as well as global outlooks. Furthermore, Process industry forms a backbone of the entire industrial sector, by delivering a wide scope of products which provide the base and support for many other industries: plastics, adhesive agents, lubricants, pesticides, acids for mining operations, colors, components for pharmaceuticals and cosmetics, to name a few. However, what makes Process Industries specific is a set of characteristics that cannot be digitalized with ease. For instance, raw materials change constantly, while production processes combine complex chemical and physical reactions, whilst supported by comprehensive mechanisms and integrated systems. Considering that the production process is continuous and cannot simply be stopped, problems in any stage of operation will consequently impact the entire production line and the quality of products, but also lead to diminished plant efficiency, increased pollution, and losses. The characteristics and challenges of Process Industries are hence displayed through difficulties in measurement, modelling, process control, and optimization. It is therefore crucial to understand these challenges from the perspective of plant operators and engineers, and implement solutions which incorporate Artificial Intelligence, which precisely target weak links within the production process, thus providing a basis for sustainable plant operation.


In the case of Process Industries, there are numerous benefits that AI could bring to focus, thus acting as a key towards achieving sustainable production:

  • Reduced maintenance cost and unplanned production downtime
  • Improved inventory and logistics management
  • Track & trace focused improvement of product quality
  • Increased production yield through advanced process monitoring
  • Protection of people, assets, and environment
  • Reduced energy consumption and CO2 emissions

These goals can be accomplished through the implementation of Process AI Services such as:

  • Data Analytics Expert Services: leveraging the potential and value of plant and production data
  • Predictive Analytics: applying Machine  learning to predict asset and process deviations
  • AI Anomaly Assistant: utilizing AI engines as collaborative platforms to increase the business impact of processes
  • Batch Performance Analytics: quality optimization within batch production

Batch Performance Analytics is focused on optimizing Batch processes, which are challenging due to their non-stationary process nature, nonlinear, and dynamic data behavior. There are also variabilities with regards to cycle times and the time between batches, as well as uneven batch lengths [9]. As a solution, integrated multiple data sets from different sources, data-driven modelling with the application of machine learning and domain experience, and anomaly detection with help of AI algorithms based on trained models can be applied. An anomaly in time series data is defined as a point or sequence of points that deviates from the normal behavior of the data. At the same time, the methodology of anomaly detection fundamentally compares the expected prediction errors to the actual prediction error [10]. The method of creating an AI-based anomaly detection system can be thus depicted in 5 steps:


Data sourcing → Data contextualization → AI model creation → Model training → Anomaly detection


Some of the likely benefits of Batch Performance Analytics, amongst others, would feature: Reduced batch variability and off-specification batches, increased production efficiency by reducing waste and batch cycle time, as well as an overall increased understanding of the batch process. In other words, by refining the batch process, a sustainable and efficient production process becomes the norm.


Looking Ahead

There is no doubt that Artificial Intelligence will make a decisive impact on Process Industries, the overall industrial complex, and many other if not most aspects of our daily lives in the coming decades. Whereas the Sci Fi scenario of parallel consciousness is not something that can be expected any time soon, if ever, a new digital revolution is on the rise. Its main purpose is not to challenge humanity, but provide support, inspiration, and help us tackle some very burning problems such as carbon emissions and waste. AI can greatly assist in ensuring energy efficiency, resource efficiency and ultimately provide us with a power boost on the path to Sustainability. We have the knowledge, we have the tech.

It’s time to create value.



[1] Original link / accessed on 03-05-2022: | 

Archived link / 29-04-2022:


[2] Original link / accessed on 03-05-2022: | 

Archived link / 03-05-2022:


[3] Samuel Butler | Darwin Among the Machines | The Press | Christchurch, New Zealand | 1863 |   

Original link / accessed on 03-05-2022:


[4] Nicholas Berente, Bin Gu, Jan Recker, Radhika Santhanam | Managing Artificial Intelligence | 

MIS Quarterly Vol. 45 No. 3 / 09-2021 | p. 1433 |


[5] James Luke, David Porter, Padmanabhan Santhanam | Beyond Algorithms: Delivering AI for 

Business | CRC Press | 2022 | p. 7 | DOI: 10.1201/9781003108498


[6] Taeho Jo | Machine Learning Foundations: Supervised, Unsupervised, and Advanced Learning | 

Springer Nature | 2021 | p. 3 | DOI: 10.1007/978-3-030-65900-4


[7] John D. Kelleher | Deep Learning | The MIT Press | 2019 | p. 1 | ISBN: 9780262537551


[8] Vijay V Raghavan, Venkat N. Gudivada, Venu Govindaraju, C.R. Rao | Cognitive Computing: Theory 

and Applications | Elsevier | 2016 | p.p. 6-7 | eBook ISBN: 9780444637512


[9] Anicia Zeberli, Sara Badr, Christian Siegmundb, Markus Matternb, Hirokazu Sugiyamaa | Data-

driven anomaly detection and diagnostics for changeover processes in biopharmaceutical drug

product manufacturing | Elsevier | 2021 | p. 54 | DOI: 10.1016/j.cherd.2020.12.018


[10] Michael Daniel DeLaus | Machine Learning for Automated Anomaly Detection in Semiconductor 

Manufacturing | Massachusetts Institute of Technology | 2019 | p.p. 17-18 | 

Original link / accessed on 06-05-2022:


* All dates are displayed as DD-MM-YYYY