The Sicement MCO expert system optimizes the operation of Südbayerisches Portland-Zementwerk Gebr. Wiesböck & Co. GmbH and reduces both operator workload and energy costs.

The grinding process accounts for almost 45 percent of total power consumption in cement works. Siemens developed the Sicement IT MCO (Mill Control Optimization) expert system to help mills achieve higher energy efficiency, productivity, product quality, and availability. Sicement IT MCO is based on components of the Advanced Process Control (APC) library of Simatic PCS 7. The expert system makes precise forecasts of the quality parameters from the mill, controls the mill according to the nominal values, and can be integrated into an existing Simatic PCS 7 control system without additional hardware.

The first pilot project for the expert system was Südbayerisches Portland-Zementwerk Gebr. Wiesböck & Co. GmbH, located in the small town of Rohrdorf in southern Germany. Founded in 1930 by Georg Wiesböck and his brothers Ludwig and Andreas, it is today part of the Rohrdorf Materials Group with 40 production sites in Germany, Austria, Hungary, and Italy. Ball mills grind the raw material from local quarries for cement production. The raw meal is then fired to cement clinker, the basic material for all types of cement.

Grinding is the number one power- consuming process in cement works (All photos: Südbayerisches Portland-Zementwerk Gebr. Wiesböck & Co. GmbH)

The most important quality parameter is the fineness of the raw meal. Previously, samples were taken hourly, analyzed in the laboratory, and used for process correction. Recently, however, the company decided to implement a new control system to record and analyze the data automatically and control the grinding process. Simple integration into the existing control technology, an increase in throughput, and a reduction in the operator’s workload were further requirements for the new system.

System architecture and implementation

With the components of the Simatic PCS 7 APC library, complex relations can also be written mathematically and used for automatic, flexible plant operation. In order to master the complex control tasks within the grinding process, the optimized mill control has a neuronal soft sensor that measures the process input variables from which the MCO makes a fineness forecast. Model Predictive Control (MPC) reduces process fluctuations.

The Rohrdorf MCO system was implemented in three phases: The analysis of the plant and its existing control solutions in phase one were followed by comprehensive plant tests. After evaluation of the production data, Siemens modeled the neuronal soft sensor and the MPC system in the second phase. Software engineering and integration into the existing PCS 7 control system followed in phase three.

Convincing results

In comparison with the previous mode of operation of the mills, the cement works achieved a 5 to 8 percent increase in performance, and the investment paid off within three months. Moreover, the integration into the PCS 7 control system prevents additional service and maintenance costs.The MCO system ensures the optimum quality of the raw meal as well as stable production conditions. The reaction time from the entry of a modification at the system input to the appropriate response at the system output of the controlled system has been drastically reduced. The uniform grinding process increases throughput and relieves the workload of the operator. Energy consumption has been considerably reduced and the life of the plant mechanics prolonged. Following the excellent results of this project, Siemens has begun optimization of the other three mills.

A sensor measures the filling level in the cement mill

Optimized mill control: how it works

The new mill control operates in three steps: First the neuronal soft sensor measures nine process input variables and generates a fineness forecast. These process variables include the sifter speed, fresh product quantity, twist flap position, and representatives for the filling level of the mill (bucket chain flow, power of the mill drive, Electric Ear chamber 1 and 2) as well as the mill temperature and the selected recipe type. The particle size forecast is then compared with the laboratory values as a setpoint parameter. The compared value for the particle size (quality) and the measured value grits (returned material) are the control variables of the MPC. The values are optimized by changing the batch feed and the sifter speed based on a predictive calculation of the appropriate control processes using a complete process model to shift the control variable as close as possible to the desired nominal values. This process yields external set points for the individual subcontrollers to optimize the future behavior of the plant over a specific time period.

info:

contact: