Big Data at AIChE Spring Meeting

AIChE-logoAustin, Texas, already a hotbed for technology as recently recognized by Forbes magazine, will host a gathering of chemical engineering professionals later this month. The AIChE is hosting the 2015 Spring Meeting and 11th Global Congress on Process Safety here on April 26-30.

If you’re planning to attend, make sure to catch some of the “Big Data” sessions to see how this data is being applied in our world of process instrumentation and automation. I’ll highlight two sessions featuring members of our Emerson team here in Austin.

On Monday, April 27 at 3:30pm, Emerson’s Mark Nixon, Terry Blevins, Willy Wojsznis and John Caldwell will present, Industrial Big Data Vision and Solutions. Here’s an excerpt from the session abstract:

Emerson's Mark Nixon


Emerson's Terry Blevins


Emerson's Willy Wojsznis


Emerson's John Caldwell


The process industries adopt many Big Data approaches that are applied in other industries however the Big Data implementation for Process Industries is distinctive in that it sets specific requirements for Big Data infrastructure, learning algorithms including data analytics, and presenting the results.

The presentation will address the basic components of Big Data pipeline for the process industry: hardware and software infrastructure, data streaming, data preprocessing and data learning techniques.

The core of data learning is Data Analytics (DA) which has proven its effectiveness in process fault detection and quality prediction both for batch and continuous processes. The real prospects are that Big Data based on DA will be among the leading directions for improving process effectiveness. DA requires a significant departure from the traditional thinking about how process control is implemented. Instead of the deterministic and tangible world of signals and devices, there is an abstract realm of statistical indexes, correlation factors and matrix operations. This puts a significant strain on the control systems’ developers, engineering companies, process operation and maintenance personnel. The presentation will address these major challenges for professionals working on Big Data for the process industry.

Mark, Terry, Willy and John will also present Tuesday, April 28 at 2pm on Embedded Analytics in Industry Big Data Applications. Here is the abstract:

The chemical process industry is challenged by increasing technical complexity involving large integrated facilities and rapidly evolving process automation technologies. Experienced workers are needed to make risk decisions yet much of the experienced personnel are nearing retirement and new plants are often located in areas which lack an experienced workforce. Process analytics is a powerful tool to gain better insight into these complex operations, to improve performance and reliability, and enhance operator effectiveness. However, practical implementation of on-line analytics can be problematic because of real-time data integration, operations work flows, and difficulty in implementation and maintenance of traditionally layered solutions.

This presentation describes an approach for embedding data analytics in a control system to provide on-line process analytics for improved operator effectiveness, production performance, and manufacturing reliability. The embedded solution is configured and implemented in the native control system to reduce complexity and provide a consistent operations interface. Seamless integration with real-time and historical data simplifies model training and on-line implementation using standard control system engineering tools. The presentation includes results from field trials in polymer production and oil refining facilities.

Capabilities of the embedded analytics include process quality parameter prediction and abnormal situation monitoring with fault detection. Statistical modeling technologies used for process parameter prediction include Projection to Latent Structures (PLS), Neural Networks (NN), and Multiple Linear Regression (MLR). Abnormal situation and fault detection incorporates Principal Components Analysis with enhancements for multi-state operations. The embedded analytics are executed as function blocks inside the control system controllers and are configured and trained using techniques familiar to control system users.

Embedded data analytics is a core component of big data applications in the process industries and can be used for targeted data analytic applications which benefit from residing in the control system where they can be easily integrated into other control system applications and operations interfaces. An integrated approach toward on-line analytics provides an easy-to-use set of tools for analytic model development and verification, with subsequent download to the DCS controller for on-line operation. The model development and download for on-line operation is streamlined into clearly defined and easily executed steps:

  • Defining analytic model configuration, i.e. identifying potential process parameters for PCA/PLS model
  • Creating analytic module based on the defined model configuration, downloading module to the DCS controller.
  • The module is configured to collect history data, including lab analysis that is defined by the analytic model.
  • Developing analytic models, including PCA, PLS, NN and MLR from collected historical data as defined by the analytic module or from an external historical data file created prior to the analytic module download
  • Validating and downloading model for on-line operation

On-line analytics monitors faults in the process operations and predicts product quality. The results are presented in a web based interface and can also be used to enhance an existing alarming system. The predicted property quality can be used in advanced control and optimization strategies such as model predictive control.

The continuous data analytic design has been tested on simulated data and in industrial on-line applications in the field. In one such field application, an embedded analytics PLS model was used to predict heavy components in a distillation column. When the results were compared with an online analyzer, the analytic modeling demonstrated several advantages: more reliable operation, less demanding maintenance, and integrated fault monitoring which supports quality prediction validation and process operations troubleshooting. Process engineers greatly appreciated the ability to identify abnormal distillation column operation and the potential for including this functionality into an existing alarming system.

The field trial facilitated the development of an iterative procedure for analytic model improvement. Similar results were obtained from a polybutene unit analytic model development and an on-line test. Specific to this process is the product quality (viscosity) which can be set on two distinct levels, depending on the manufacturing needs. Two alternative approaches were explored: one using a state parameter associated with viscosity and the other using two models for Low and High viscosity values. In the majority of tests two separate analytic models performed better. The primary problem with the multistate analytic model was identifying a process parameter that was defined in the analytic model as a state parameter that correlates well with quality level.

I hope you’re able to catch the presentations. If not, I’m sure they’ll have more to share at the Emerson Exchange conference in Denver this fall. You can also connect and interact with other control system analytics experts in the DeltaV groups in the Emerson Exchange 365 community.

One comment so far

  1. Process data analytics to gain better insight into the process may be done in the in control system or historian

    Asset data analytics such as equipment condition monitoring to detect fouling and predict failure will likely happen in the historian or dedicated asset management system

    Analytics correlating process data with asset data: for instance correlating relief valve releases against equipment events such as fouling or failure of heat exchangers, air cooled heat exchangers, or cooling towers etc.

    Some of this analytic will be historic in nature, looking at “Big Data” collected over long periods of time. Other analytics will be real-time in nature providing real-time indicators. Both schemes requires raw data to analyze in the first place, and distill into information and knowledge, possibly wisdom, as per the DIKW pyramid model. In any case, plants need to deploy more sensors, and staring with a plant-wide WirelessHART Pervasive Sensing infrastructure is the easiest way to tie this data into the control system and historian as no 4-20 mA and on-off wiring or I/O cards are required.

    Learn more from this article: Instrumental to Success
    http://www.ceasiamag.com/article/instrumental-to-success/11137

Leave a Reply