Management Information Systems

Management Information Systems

Importance of Information Systems
Information systems in the paper mill exist mainly at two different levels:
1. 1. Enterprise level: To collect all information concerning the material flow in the paper mill, including e. g. quality parameters of the produced paper. The in¬formation is usually collected in a very rough time scale, i. e. reel based. The purpose is to derive figures which are required for commercial control of the mill.
2. 2. Process level: To collect quality and process data in quite high time resolution to be able to analyse and track process problems, to find the technical reasons for poor/excellent production, or to summarize the data in reports as needed by the enterprise level systems.

Whereas enterprise level systems deal mainly with production quantities (amount of good/bad production) process level systems include tools to analyse reasons for good/bad production. Comprehensive collection of process data is more and more widely used but in many older mills is still not available. This is due to
. • lack of communication standards in automation systems
 • strict border between QCS and DCS/PLC type of systems when those mills were
The earliest information systems were:

. • QCS quality reports: summarizing production and quality over reel, shift, day or grade
. • DCS process trending: trend of important process variables.

Until recently there was no simple means to show in a flexible way on one screen
e. g. how press settings (DCS data) affect moisture profile quality (QCS data). One obvious reason is, that DCS systems are (still) not built to handle “array based” information like CD profiles. They can deal only with single values like a steam pressure. On the other hand QCS systems are not built to handle thousands of process values, which requires totally different engineering and configuration tools. Modern information systems bridge the gap, by using history data bases which can deal with tens of thousands of tags (variables), with single values and with profile data as well. Additionally they provide simple access to the data via a web browser and also from the operating screens on the production floor.
Besides QCS and DCS there exist more systems at a paper mill, which have the main task to collect information for diagnostic purposes:
• Machine condition monitoring system: These systems collect information via vibration sensors, to analyse bearings, and drive components.

Advanced systems

9.3 Information Systems
include an expert system for quite detailed analysis of the drive component. A machine condition monitoring system can e. g. detect whether the outer ring or the inner ring of a bearing is damaged and is likely to fail within the next weeks. Based on this information the mill personnel can arrange countermeasures and exchange the bearing during the next planned shutdown. Typically a few 100 vibration sensors are connected to such a system. They are scanned one by one. Scanning occurs about every hour. The signal analysis uses known geometrical bearing and roll data and relates them to the actual meas¬ured signals to diagnose the condition of the bearings. The measurements have to be taken with a sampling rate of about 10 kHz to be able to diagnose bearings in detail and to give a prediction of the remaining lifetime of the bearing.

. • Process condition monitoring system (sometimes also called technological mon¬itoring system): These systems collect information from QCS sensors, vibration sensors, pressure sensors, triggers at rolls and fabrics, etc. to analyse the reasons for periodic quality disturbances. All collected signals are taken simultaneously with sample rates from 100 Hz to 4 kHz, depending on sensor type and applica¬tion. The required measurement frequency is given by the revolution time of the machine elements which can cause quality variations, like felts, rolls, pumps, etc. The correlation results are calculated using time synchronous averaging. A proc¬ess condition monitoring system can e. g. detect whether an applicator roll is responsible for periodic coat weight deviations, or whether a moisture variation is correlated with the revolution time of a press felt.
. • Barring monitoring system: This system identifies calender rolls which cause barring. Technically it works similarly to a process condition monitoring system.

. • Web inspection system: A series of cameras arranged in the CD inspect 100 % of the produced paper. Image analysis technologies are used to automatically clas¬sify the paper defects into different categories, like bright spot, dark spot, wrin¬kle, hole, etc. The web inspection systems generate reports, independently from QCS. Based on the camera based measurement and the image analysis, areas with serious defects can be marked on-line. A special winder control can stop the winder precisely at these positions, to cut off the low quality production.
. • Web break monitoring: A couple of cameras are installed at the front side and drive side of the machine to supervise positions where the web is not guided by wires or felts, and where the paper risks to break. Optical light barriers detect paper breaks. In the case of a paper break, the last minute of the camera read¬ings are stored to a hard disk for later analysis by the production personnel. Based on the camera recordings it can be analysed at which position of the machine e. g. an edge crack started to be seen, and how it evolved to a paper break, or it can be seen whether paper clipped to a roll and caused a web break. This allows the mill personnel to analyse the reasons for breaks and to optimize the runnability of the machine.

Process Analysis using Information Systems
Current development of information systems aims to automatically derive knowl¬edge from historical data (Fig. 9.5). When analysing difficult technological prob¬lems such systems still need off-line analysis by technological experts to interpret the results. In a typical application 6 to 12 months of recorded quality and process data are analysed with respect to given technological questions, like: “Most of the production in the last 6 months showed average printability but sometimes print¬ability was excellent. Why? Which are the machine settings to achieve best qual¬ity?”
Such questions are difficult to answer, as several hundred machine settings and raw material parameters have to be taken into account as potential influencing factors. Data mining technologies based on neuronal networks, expert systems or algebraic methods are used to find the requested answers. Algebraic methods are in many cases not sufficient, as these can deal only with numbers (process values) but not with cardinal data like felt supplier, ash supplier, etc. which are also of high importance.
Simpler, but still difficult applications can run online and give information of a produced paper quality which is not measurable online.

Those applications are called soft sensors. The paper quality property is estimated based on historical data. For example: Basis weight can be predicted quite well based on some process parameters in the wet end section. Historical data can be used to “learn” the rela¬tion between the measured process parameters and the produced basis weight. The learning takes place iteratively to adapt to slowly changing production condi¬tions. Such a prediction is useful during start up of the machine, while the paper is still not through the drying section and still not measured by a quality scanner.

Having a basis weight predictor the user can still adjust basis weight and therefore achieve good quality, even before the quality measurement starts scanning.

9.3 Information Systems
The challenge of soft sensors is to automatically reject data which are not ade¬quate in order to learn relationships between the quality parameter concerned and the process data, e. g. because a sensor was wrong, the machine settings were just in the process of being changed, the press felt was worn out and just before re¬placement. This is especially challenging in applications where a hundred or more process data are required to really predict a given paper quality. An example is the prediction of strength parameters. Influences on strength are e. g. the furnish composition, the jet/wire ratio, retention, dryness after press section, tension in the draws, and calendering conditions.