March 19, 2014 - Ultramax enables users to generate more business value from production and manufacturing operations with the existing assets. It complements and guides the existing Process Control System (including those with various levels of Advanced Process Control).
In particular, from the stream of operating data obtained from regular operations, Ultramax continually learns how to refine the adjustments (tuning) of setpoint values to increase the effectiveness, efficiency and productivity of operations. Operating performance is defined by multiple metrics and the criterion selected by the users. The metrics are usually related to: process safety, product quality, costs, throughput, and environmental impact. With this comprehensive evaluation, it is notable how often current controls underutilize or waste process capabilities.
The technology used is Sequential Empirical Optimization (SEO). Because SEO uses available data immediately – rather than gathering data for later off-line analysis – it gets improvements faster than other technologies; and for the same reason it provides safer performance than the experimental runs and the model validation runs used in other technologies. Because SEO is self-learning (it does not require engineers to create or update process models) and the optimization analytics are fully automatic, it is simpler and less expensive to install and use. Ultramax is applicable to most production processes: continuous, batch, discrete, bulk. It depends on the existing Regulatory Process Control and its sensors to implement the setpoint values.
It can be used with manual data entry for temporary studies to determine optimal SOPs; and on-line to compensate adjustments for on-going changes in conditions that affect operating performance (such as measured product / fuel characteristic, demand on the process, environmental conditions), where it can run in advisory or closed-loop mode.
Technical advances and improvements over 7.2.f include:
- Better explanations and user insights: It now explains some reasons for the sequential adjustment advices, such as forced changes in adjustments due to changes in inputs constraints, and limits to changes from previous adjustments due to user-defined maximum allowed (gradual changes to maintain process performance stability).
- The algorithm to balance the risks of violating output constraints and the extrapolation speed towards the optimum has been improved, and it explains the risks it takes.
- The Data Pattern report indicates the existence of combinations of adjustments which are highly correlated (i.e., which span a subspace).
Applications with many inputs (over 20) and with noisy data now converge to the optimum more consistently, and thus on the average faster.
For the analysis of historical run data this feature indicates probable lack of independence among adjustments.
Selecting independent adjustments is necessary for the analytics of empirical prediction models and optimization to work.
The new version is available free to current owners of 7.2 licenses first released on April 2011, and at a discount to owners of older versions.
