Industrial Modeling Service (IMS-IMPL)
Download as DOCX, PDF0 likes669 views
The document outlines the Industrial Modeling Service (IMS) provided by Industrial Gorithms, which aims to enhance existing process models and data through improved methodologies known as 'smart modeling.' It describes various techniques, such as steady-state detection, data reconciliation, composition tracking, and excess-model regression, designed to optimize decision-making in the process industries. The IMS is positioned as a critical tool for continuous improvement and sustainability in industrial applications, aligning with modern concepts of smart manufacturing and Industry 4.0.
Industrial Modeling Service (IMS-IMPL)
- 1. Industrial Modeling Service (“Smart Modeling”) (IMS-IMPL) “Better Industrial Models and Data for Better Design, Planning, Scheduling, Optimization, Control, Monitoring and Accounting Applications” i n d u s t r IAL g o r i t h m s LLC. (IAL) www.industrialgorithms.com December 2014 Our Industrial Modeling Service (IMS) involves several important (but rarely implemented) methods to significantly improve and advance your existing models and data and can be considered as “Smart Modeling” for the process industries. Since it is well-known that good decision-making requires good models and data, IMS is ideally suited to support this continuous-improvement endeavour. IMS is specifically designed to either co-exist with your existing design, planning, scheduling, etc. applications or these same models and data can be used seamlessly into our Industrial Modeling and Programming Language (IMPL) to create new value-added applications. The following techniques form the basis of our IMS offering. Steady-State Detection of Phenomenological Variables to Assess Process/Production Stability IMPL defines phenomenological variables as flows, holdups, yields (quantities), setups, startups, switchovers, shutdowns (logics), densities, components, properties and conditions (qualities). IMPL’s steady-state detection (SSD) algorithm (Kelly and Hedengren, 2013) is useful to determine whether a process, production or plant is steady or stationary and provides an assessment of whether the system has negligible accumulation as well as its ability to be regulated and manoeuvred. Once stationary, then steady-state data reconciliation can be used to detect and identify faulty instrumentation and/or leaks. However, if a process is rarely at steady-state then this is a possible indication of poor control and/or inadequate disturbance rejection. The data required to perform SSD are a set (20 to 50 tags) of key process, production, plant or phenomenological variables (KPV’s) sampled at typically one-minute time intervals (IMPL-DataAnalysis). Furthermore, if steady-state models need to be calibrated or regressed, then not using truly steady-state data will result in biased parameter estimates due to the presence of auto-correlation i.e., the residual errors are not independently and identically distributed. Data Reconciliation of Flows, Holdups and Yields to Identify Gross-Errors/Outliers/Defects As mentioned, once the system is detected to be at steady-state then IMPL’s data reconciliation solver (Kelly, 2004) can be used to identify flow, holdup and yield gross-errors or anomalies with relatively simple quantity or material balances. Other phenomenological variables can also be included such as densities and conditions but this requires extra quantity and quality nonlinear balances. If a defect or outlier is flagged, then the field meter or analyzer needs to be re- calibrated, etc. where it is not uncommon to find many instruments that have significant measurement biases or drifts that need to be addressed or accounted for in some manner (APA-IMF, APA-FP-IMF and APA-OP-IMF). Composition Tracking of Feedstocks and Intermediate Materials to Trace their Amounts One of the most difficult and misunderstood aspects of applying process models off-line or on- line is the lack of feed and intermediate material compositions. Implementing IMPL’s
- 2. composition tracking application (Kelly et. al. 2005) essentially performs a dynamic or time- varying numerical integration of all holdups or inventories over time to track/trace the relative amounts of each material before it enters or feeds the process (APT-IMF). Pseudo- components, micro-cuts or hypotheticals used in rigorous process simulators all require this data in order to properly predict the process’ output quantity and quality variables. Without feed composition tracking, it is very unlikely that useful predictions from process models will result even with the best intentions. Data Regression of Model Coefficients with Densities, Properties, Components and Conditions Not only are feed and intermediate material compositions necessary but most process models also require as input, certain process parameters or coefficients. To provide this, IMPL’s data regression solver (TSE-GFD-IMF) can be used with actual process data to estimate or fit these coefficient values correctly. Typically these types of process parameters are heat or mass transfer coefficients, catalyst activities, etc. and require nonlinear regression techniques which IMPL employs (APM-IMF). Excess-Model Regression of Existing Models to Extend their Accuracy and Precision In some cases, it is not possible or extremely difficult to update or re-calibrate existing models with model coefficients etc. and unfortunately this most likely inhibits their ability to accurately and precisely predict the necessary process variables. Excess-model regression (XMR-IM) is a simple technique that IMPL implements to retrofit or revitalize existing models to improve their predictability. XMR uses the existing model’s predicted values as input and uses other related process variables to extend, enhance or augment the model with this information. An effective industrial example of this approach can be found in motor gasoline blending where a nonlinear or non-ideal blend law with fixed parameters (Ethyl, Dupont or Mobil Transformation Method) is extended by fitting extra parameters typically called “bonuses” using the component recipes as regressors or explanatory variables (APE-MGB-IMF). Design of Experiments for Open/Closed-Loop Dithering to Estimate Better Models Regressing industrial models using passive or happenstance data may not be rich enough to estimate good or useful models especially when feedback is omnipresent in the data. In order to improve this situation, IMPL’s unique dither signal design problem (DSDP-CLE-IMF) can be easily employed to determine experimental trials that can be run or executed on the system to significantly improve the industrial data quality. After the dither signal trials have been applied to excite or stimulate the actual process or production, then better regressed models will result. In addition, this method can also be used to verify actual first-order derivatives taken directly from the production or plant where good optimization requires good derivatives. In summary, IMS should be considered as a vital part to increasing your ability to extract more value out of your existing industrial models and data. It should also be emphasized that these methods should be maintained and applied on a regular basis for continuous-improvement and sustainability of the applications (Kelly and Zyngier, 2008) in any industrial environment. In addition, these techniques can be implemented before the installation of any new application or extension in order to provide a benchmark or reference-point for its expected future profit and/or performance benefits. And finally, the methodology outlined here is consistent with the recent concepts of Smart Manufacturing, Industry 4.0 and Smart Plant where Christofides et. al. (2007) state in their
- 3. conclusion (a) the following requirement: “the development of easy-to-use software that makes system modeling and control routine and easy-to-incorporate in the chemical engineering curriculum, as well as in an industrial environment”. IMS (Smart Modeling) is in our opinion, a step in this direction. Please contact Alkis Vazacopoulos (alkis@industrialgorithms.com) to obtain a quote for IMS and IMPL’s development and deployment licences as well as special pricing for IBM’s CPLEX LP, QP and MILP solvers which are tightly integrated with IMPL to solve industrially significant discrete and nonlinear types of problems. References Kelly, J.D., "Techniques for solving industrial nonlinear data reconciliation problems", Computers and Chemical Engineering, 2837, (2004). Kelly, J.D., Mann, J.L., Schulz, F.G., "Improve accuracy of tracing production qualities using successive reconciliation", Hydrocarbon Processing, April, (2005). Christofides, P.D., Davis, J.F., El-Farra, N.H., Clark, D., Harris, K.R.D., Gipson, J.N., “Smart plant operations: vision, progress and challenges”, American Institute of Chemical Engineering Journal, 2734-2741, (2007). Kelly, J.D., Zyngier, D., "Continuously improve planning and scheduling models with parameter feedback", FOCAPO 2008, July, (2008). Kelly, J.D., Hedengren, J.D., "A steady-state detection (SDD) algorithm to detect non-stationary drifts in processes", Journal of Process Control, 23, 326, (2013). IAL, “Advanced production accounting industrial modeling framework (APA- IMF), Slideshare, July, 2013. IAL, “Advanced property tracking/tracing industrial modeling framework (APT- IMF), Slideshare, July, 2013. IAL, “Advanced process monitoring industrial modeling framework (APM- IMF), Slideshare, July, 2013. IAL, “Advanced production accounting of a flotation plant industrial modeling framework (APA- FP- IMF), Slideshare, August, 2014. IAL, “Advanced production accounting of an olefins plant industrial modeling framework (APA- OP- IMF), Slideshare, August, 2014. IAL, “Time series estimation of gas furnace data industrial modeling framework (TSE-GFD- IMF), Slideshare, August, 2014. IAL, “Data analysis by checking, clustering and componentizing in IMPL (IMPL-DataAnalysis), Slideshare, September, 2014. IAL, “Advanced parameter estimation for motor gasoline blending (MGB) industrial modeling framework (APE-MGB- IMF), Slideshare, November, 2014.
- 4. IAL, “Excess/x-model regression to extend the accuracy and precision of existing industrial models (XMR-IM), Slideshare, November, 2014. IAL, “Dither signal design problem for closed-loop estimation industrial modeling framework (DSDP-CLE-IMF), Slideshare, December, 2014. http://en.wikipedia.org/wiki/Industry_4.0, accessed December, 2014.