Skip to main content

Bringing The Best Of Both

Published by
World Cement,


Dirk Schmidt, KIMA Process Control, explains how to best combine artificial intelligence and predictive advanced control for fully autonomous grinding control.

A key objective of today’s industry is attaining ‘Industry 4.0’, and everywhere one looks ‘big data’ and AI are being hailed as the future. The goal is to convince cement producers to utilise the latest technologies, similar to those made available day-to-day in mobile phones. Of course, even in very conservative industries, new technologies will have to be considered sooner or later. Particularly when it comes to process optimisation and automation, other industries have shown how production processes can be enhanced. It might surprise some therefore, to learn that fully autonomous mill operation (including the use of AI) has been taking place since 2009, as this article will show. There are actually business consultants that seriously want to ‘copy-paste’ the tools and control loops of Industry 4.0 from chemical plants and refineries into cement production with rotary kilns and ball mills/VRMs. A recent example came from a report regarding the first successful conversions of plant control to AI. Under the headline it was stated that it was a breakthrough to use these technologies in the most conservative building materials industry. Caution is needed here – the capabilities of AI are still limited as the history shows. However, alternative routes have already been found and AI delivers these last lost puzzle pieces; KIMA Process Control is enthusiastic that it has found the right mixture of AI and APC.

How AI started in cement production

After a short discussion with these new suppliers it becomes clear that this ‘revolution’ in cement plant control systems is simply operating with Model Predictive Control (MPC) and soft sensors. The ‘intelligence’ is (still) limited to a prediction of a non-measurable process signal or a measurable one predicted in the future. Instead of MPC, the system gets a new name and the talk is now about AI. To clarify: supporting serious stable control systems, such as Advanced Controllers using fuzzy logic predictions, uses only elements of AI, but nothing more. It is not the ‘self-thinking’ and ‘learning’ machine everybody thinks about when AI is mentioned, it is simply adaptive MPC.

Self-learning adaptive MPC was introduced in 2001 by Powitec Intelligent Technologies (Germany), the first comprehensive black-box controller that operated a rotary kiln fully autonomously for more than 24 hours without manual interaction. The core was an image processing camera looking into the kiln and an online prediction of the free lime. Various AI components were used as far back as 2002. Using adaptive MPC, the energy input to the kiln main burner and calciner was automatically adjusted. Pioneers in the use of AI during these days were the cement producers LEUBE (Austria) and Maerker Zement (Germany).

Soon after, companies such as ABB, FLSmidth, Pavillion, KIMA Echtzeitsysteme and Rockwell entered the market with similar model-based controllers – and again not soft sensors (like today), but serious black-box control engines. Today, many companies have returned to the more robust Advanced Process Control using complex fuzzy logic control, and use predictive models for soft sensors. The reason for that is the serious differences in the cement manufacturing process compared to all other common production processes.

Special process conditions

Producing clinker is a uniquely complex operation! It involves two competing process circles plus a significant amount of wear during each operational campaign. In a so-called ‘multi-dimensional non-linear process model’ of a kiln or a mill, tomorrow is already different to today. There were attempts at implementing software features, such as self-adaption and self-learning, but a host of changes have to be considered for each period of maintenance: liners, balls, chutes, feeders, valves, refractory, fuels, and (last, but not least) the raw materials.

And then there are even further changes to consider: the quarry components and additives, the fuel heating value, water and ash changes, the change of coal and pet coke particle size distribution from their mills and the related change of combustion (ignition point, out burn, flame shape, etc.). All these changes directly affect the chemistry of clinker and cement – a significant challenge for a controller.

If a multi-dimensional process model is supported with signals that have a drift, or are not stable, they fail. They may work for some days and then permanently lose performance/quality. And the most significant process signals, which are necessary for the control of a mill or a kiln, show exactly this phenomenon and avoid continuous, fully-autonomous operation. This is, of course, the problem with MPC and/or AI.

The most robust control solution for such permanent exchanges is still a rule-based control structure that follows the experience of the most intelligent source in the plant: the operators. Accordingly, the mixture of APC and AI is key as it most accurately represents the human element.

Picking the best parts: AI and APC

APCs have been continuously developed over the years, and the line between APC and AI has become blurred. APC should not be described as ‘old fashioned’ or ‘historic’ – this would ignore the developments made by the earlier mentioned companies using these tools some 10 – 15 years ago. Until recently, APC using fuzzy logic was the most reliable and robust solution for the cement manufacturing processes. A plant operating its mill or kiln with this technology is not ‘old fashioned’. Assisted by modern modules of AI, these systems reach a new level in automation: Predictive Advanced Process Control. This article will detail the first plants which drive their mills fully autonomously over several days. Furthermore, it should be mentioned that the level of the ‘auto-pilot’ function is not limited to smooth operation only. The KIMA Process MILLMASTER allows a fully automated start of the mill, automated stop, emergency recoveries and fully automatic switching between cement types. The following case studies will discuss some plants that increased their performance using SMARTCONTROL from KIMA Process Control. The software-platform was supplied globally in nearly 200 rule-based expert systems known as ‘MILLMASTER’ and using state-of-the-art fuzzy logic and model predictive soft sensors.

Predictive APC

In 2008, KIMA Echtzeitsysteme (the previous name of KIMA Process Control) published an article about a project to supply 30 SMARTCONTROL APC packages for ball mills (inclusive the fill-level measurement SmartFill) to a selection of Holcim group plants in Eastern and Central Europe. After the commissioning, and later during steady operation of these plants, the development of MILLMASTER continued separately in the Holcim group as well as in KIMA Process Control.

New designs of the human machine interface, programming logic, and new software modules were developed to follow new trends in automation and advanced process control. The following case studies show the results and experiences of the current users of this product. New sensor devices, like the V-Sens (vibration sensors) and T-Sensor (a contactless torque sensor), build upon the success of SMARTFILL from ball mills to all VRM applications. The combination of SMARTFILL with APC systems is not only successful with MILLMASTER, but also with the control software platforms from LafargeHolcim, HeidelbergCement and BuzziUnicem. SMARTFILL has made KIMA P.C. a market leader in this field of instrumentation. As it provides robust and drift-free process signals, it can be used in combination with other process variables for predictive solutions – as Soft Sensors for fineness in the MILLMASTER system.

AI plus APC

The best performance available today is achieved with a mixture of the well-established ‘Expert Systems’ using robust fuzzy control modules as well as modules/tools from AI. The following three case studies show a wide variation of tasks/targets and successful integrations that are being offered in cement plants today.

You can continue reading this article and access the June issue of World Cement HERE: https://bit.ly/304Pa2f

Read the article online at: https://www.worldcement.com/special-reports/10062020/bringing-the-best-of-both/

You might also like

 WCT2020

WCT2020

WCT2020 provides a unique online forum for cement industry professionals to hear first-hand from experts through a series of exclusive presentations from cement producers and industry experts.

Find out more and register for the series »

 

 Spotlight

World Cement Spotlight with Rockwell Automation

World Cement Editor, David Bizley, sat down with Michael Tay, Advanced Analytics Product Manager at Rockwell Automation to discuss his recent article in World Cement.

Entitled ‘Smooth Sailing’, this article explains how machine learning can help save energy, reduce downtime and predict equipment failures, thus enabling the smooth running of cement plant operations.

Watch the interview now »

 
 
 

Embed article link: (copy the HTML code below):


 

This article has been tagged under the following:

European cement news Construction news