Pulp Production Control Based on NI LabVIEW Data Logging and Supervisory Control (DSC) Module

Create a system to monitor the pulp production process of a large paper mill to save energy and reduce costs. Solution: To increase throughput, save energy, and optimize the production process, we used the NI LabVIEW Data Logging and Supervisory Control (DSC) Module to handle complex, nonlinear modeling processes.

"The LabVIEW toolkit makes it easier to run multiple models simultaneously, which means online performance monitoring capabilities can be simple and straightforward."

Iggesund Paperboard in Workington, UK, has five paper machines in its production line, and each paper machine consumes at least 15 megawatts of energy. Saving energy is crucial to reducing our operating costs, so we need a solution to optimize our energy efficiency while ensuring that the paper machines can produce the pulp needed for the final paperboard product.

Model design

To predict the stiffness of paper, we implemented an artificial neural network (ANN) model based on a feed-forward single layer perceptron (FFSLP) structure. This model was chosen over the multiple linear regression (MLR) model because the modeling process required nonlinearity. ANNs are more accurate in making predictions for a wide range of paper materials. In addition, it is convenient to use only one model for all paper materials, making it easier for paper mill operators to use the system.

For the online monitoring function, we combined artificial neural network with multivariate data analysis (MVDA) method because this method is more adaptable to changes in the operating environment, such as changes in paper materials. The nonlinear model thus established can more accurately predict the machine direction (MD) deviation of cross direction (CD) bending stiffness.

We designed the predictive model using LabVIEW software combined with artificial neural network technology for advanced adaptive control algorithms used in closed-loop adaptive paper machine control systems. The solution includes an offline tool that can simulate paper machine loads for different pulp types. This gives operators and development engineers the opportunity to try different refining configurations and compare predicted final board quality measurements. By simulating the effects of changing parameters using LabVIEW software, we avoid expensive full-scale trials, saving us time and money.

In addition, the LabVIEW DSC module can easily run multiple models in parallel and provide direct online performance monitoring capabilities. Using the LabVIEW DSC module, we limited the model to the state range where the paper mill can operate normally. When the variables move outside the normal operating state range, the alarm will be activated to alert the operator that there is a problem in the system. At the same time, the system will indicate to the operator which variable has failed and tell the best way to solve the problem.

Online monitoring

Iggesund Paperboard is using this new model successfully in its online operations. Using a score chart approach, the system alerts the machine operator when variables deviate from the pulp set point. The score chart shows the score of the principal component analysis for each variable and is updated at one-minute intervals.

Figure 1 shows a two-dimensional monitoring score graph with the scores of two eigenvectors plotted as horizontal and vertical coordinates. The two-dimensional score graph is used to show the relationship between the two eigenvectors, which can be represented by a 95% to 99% Hotelling ellipse, which defines the normal operating area of ​​the production process. This allows operators to easily identify abnormal values.

All similar data points are gathered on this two-dimensional score graph, and each group of data points represents the working point in the production process. These graphs can be used to display a group of data points representing a single working point, or multiple groups of data points representing multiple working points. If the data points remain within the Hotelling ellipse, this indicates that the working point is within the normal operating area. Regardless of whether the basic correlation between variables is valid, any abnormal changes in process variables can be clearly displayed using this two-dimensional score graph. With the help of the load diagram, we can determine the root cause of the problem and troubleshoot it.

in conclusion

The model we built using the LabVIEW DSC Module provides real-time process information to the board machine operator, and the system clearly prompts the operator when variables deviate from the required pulp quality set point. The pulp mill operator can select the required set point and refining energy value to achieve the best results.