Aspen Process Statistical Analyzer
(formerly Aspen IQmodel Powertools)
Empirical model-building environment for process analysis
applications
Aspen Process Statistical Analyzer is a tool set for data
manipulation and identification of non-linear models such as the
Bounded Derivative Network (BDN). The export function allows these
models to be used directly in inferential applications or the
models can be used indirectly to provide gain predictions at a
given operating point to a control application. Aspen Process
Statistical Analyzer is a core element of AspenTech's aspenONET
Advanced Process Control applications.
Features
- Data Management – Data import from
Microsoft® Excel or MATLAB
- Data Manipulation – Data matrix
management including combining data sets, variable transforms,
re-sampling, removing or adding variables, and combining variables
through calculations
- Data Visualization – Data plotting
including trending and scatter plots
- Data Pre-Screening – Data cleaning
through a variety of a variety of mechanism including filtering,
multivariate outlier detection, and manual cutting
- Multivariate Analysis – Powerful
multivariate statistical analysis for outlier detection and/or
process monitoring applications (includes PLS, PCA and highly
interactive 2D/3D multivariate charts, Q/T2 Stats, Contribution
Plots)
- SPC Support – Q statistics allow users
to view contributions from key variables over time to investigate
which variables are causing the Q statistic to violate previously
defined upper control limits
- Cause and Effect Investigation –
Correlation analysis for determining cause and effect relationships
and for time aligning data
- Training/Testing Data Selection –
Complete flexibility over data used for building and validating
models
- Regression – Support for a variety of
regression model types including Multiple Linear Regression, SERVA
analysis, Partial Least Squares, and Principal Components
Analysis
- Advanced Modeling – A variety of
nonlinear model types including static or dynamic (first order
Hammerstein) Gain Constrained models (Bounded Derivative
Networks)
- Nonlinear Controller Modeling –
Capability to build and export Aspen Nonlinear Controller
models, which are gain-constrained general order nonlinear Wiener
models with guaranteed bounds on the steady state gains (licensed
separately with Aspen Nonlinear Controller)
Benefits
- Reduces the time and effort required to transform process data
into suitable data sets for building empirical models
- Simplifies the task of building model types whose structure and
configuration are especially suited to online applications
- Improves manufacturing operations by enabling better
predictions of how the process will perform in the future through
greater model accuracy