Study On Robust Data Reconciliation Method For Chemical Process

Reliable and accurate process measurements are crucial for the control, simulation and management of process. However, errors are often contained in the measurements, which can be classified as random error and gross error. Random error is independent to each other and conforms to normal distribution, while the gross error rarely occurs and doesn't conform to normal distribution. A gross error in a measured variable causes"smearing", contaminating the estimates for other measured variables when reconciling process measurement. So the methods avoiding the effect of gross error during data reconciliation were proposed, which is called robust data reconciliation methods. But the existing robust data reconciliation methods are not completely efficient to avoid the effect of the gross error.In this paper the mechanism and research progress of traditional robust data reconciliation methods were studied. To overcome the deficiency of existing robust data reconciliation methods, a new method named bi-weight with stronger robustness was proposed. Defining the difference between measurement and its rectified values as relative residual, the objective function of bi-weight method was constructed, in which the corresponding influence function value of the relative residual contained gross error is zero. Compared with classic robust data reconciliation methods upon their mechanism,the bi-weight method was proved to be more robust.To show the benefits of using bi-weight method for data reconciliation, a linear problems and a nonlinear one were chosen as examples, and the results show that the robustness and stability of bi-weight method are better than other methods. Additionally, the effect of parameter in the bi-weight model on the reconciliation performance was studied and proved using a nonlinear example.Leakage of unit is different from the gross error occurred in sensor, and constraints of data reconciliation model will change if a unit leaks. Based on the mechanism of bi-weight method and the peculiarity of leakage, a new data reconciliation model was proposed, in which the bi-weight function of relative residuals of the constraints was added into the objective function of bi-weight method and the constraints changed to the ones with leakage. A classic example was used to prove the efficiency of the new model and the results show that the problem with no gross errors contained in the flow adjacent to the leaking node can be solved, while the other situations cannot be solved. This paper analyzed the mechanism of new model and suggested that compensation should be given to the gross error before detecting the leakage, which needs to be studied further.