Process Data Analysis
A better process understanding is the prerequisite for many engineering tasks, e.g., for process optimization with respect to certain criteria, such as product quality or energy efficiency. One way to get such an improved understanding is to analyze process data. Today, many companies maintain process data in central archives, so-called Process Information Management Systems (PIMS), meaning that process data are in principle available in sufficient quantity. However, the potential for process optimization by analyzing these data is often not used.
On the one hand, the use of archived process data for an analysis presupposes that the data is adequately compressed during archiving, so that in particular no relevant information is lost. Besides the quantity of data, the quality of data is essential. On the other hand, the quantity of available process data is usually so large that a systematic approach and a high degree of automation is required for their analysis. These hurdles often prevent from exploiting knowledge that is contained in the process data.
Efficient algorithms for the tuning of PIMS compression parameters and for the configuration-free analysis of process data, e.g., the detection of steady-states, have been developed. These algorithms originate from research work on wavelet-based trend detection that was carried out by the Chair of Process Systems Engineering at RWTH Aachen University. Together with renowned companies from the chemical industry, the research results have been further developed by AixCAPE e.V. into algorithms, tailored to industrial requirements.
pnb Process Data Library
pnb plants & bytes GmbH makes efficient and configuration-free algorithms for analyzing process data available to customers in a library, the pnb Process Data Library, as a commercial product with support and maintenance. The library is being developed as a 64-bit compiled .NET DLL with a documented interface (API) that enables the high-performance algorithms to be easily integrated into customer-specific applications.
The following algorithms are available:
- automatic computation of tuned PIMS compression parameters
- computation of the compression result (OSI PI and IP.21)
Further algorithms planned:
- variance computation
- noise removal, also called denoising
- detection of polynomial trends, e.g., stationary process states (steady-states)
You would like to profitably use the high-performance and configuration-free algorithms for analyzing process data in your company? Contact us!