Monitoring a Secondary Settler using Gaussian Mixture Models (original) (raw)
Related papers
Machine learning techniques for monitoring the sludge profile in a secondary settler tank
Applied Water Science, 2019
The aim of this paper is to evaluate and compare the performance of two machine learning methods, Gaussian process regression (GPR) and Gaussian mixture models (GMMs), as two possible methods for monitoring the sludge profile in a secondary settler tank (SST). In GPR, the prediction of the response variable is given as a Gaussian probability density function, whereas in the GMM the probability density function is built as a weighted sum of Gaussian distributions. In both approaches, a residual is calculated and a fault detection criterion is implemented via a recursive decision rule. As case study, GMM and GPR were tested using real data from a sensor measuring the suspended solids concentration as a function of the SST level in a wastewater treatment plant in Bromma, Sweden. Results suggest that GMM gives a faster response but is also more sensitive than GPR to changes during normal conditions.
Gaussian process regression for monitoring a secondary settler
2015
An approach based on Gaussian Process Regression for monitoring the sludge profile of a secondary settler is proposed. Gaussian Process is a probabilistic, nonparametric model with an uncertainty prediction. The approach is illustrated using data from a sensor measuring the sludge concentration in a settler as a function of the settler level at Bromma wastewater treatment plant (WWTP). Results suggest that the approach is feasible for monitoring and fault detection of the sludge settling process.
Journal of Chemometrics, 1999
A partial least squares (PLS) regression is used to model and visualize the waste-water treatment process. The score values of PLS are submitted to both a fuzzy C-means (FCM) clustering and a possibilistic C-means (PCM) clustering. In this work, four concepts are presented. Firstly, a hidden path process modeling is illustrated. Secondly, the use of and the difference between the PCM and FCM algorithms in process monitoring are shown. The difference between these algorithms is significant and should not be disregarded, because the membership values and consequently the typicality values generated by different algorithms have different interpretations. In FCM the memberships are relative and correspond to partition information, i.e. they sum to unity. In PCM the 'partition constraint' has been relaxed and thus so-called typicality values are computed that are no longer relative. Instead, these values represent a degree of typicality with the class prototypes that in turn correspond to different process states. Thirdly, a couple of possible uses of permanent and temporary cluster prototypes are given. Fourthly, a recursive cluster prototype updating is documented to follow the systematic variations, i.e. the movement of seasonal attraction points. These seasonal attraction points correspond to the process mean values during different seasons. This updating is necessary because these attraction points are dynamic in nature. The updating corresponding to adjusting the process mean to correct drifting problems of the mean values. Copyright treatment plants and those which purify effluents from pulp and paper mills. The latter differ from the former in that the influents do not contain enough nitrogen and phosphorus. Instead, these chemicals are added to the aeration basin. In this work the studied waste-water treatment plant belongs to this second category. A schematic diagram of the process is given in .
Gaussian process regression for monitoring and fault detection of wastewater treatment processes
Water Science and Technology, 2017
Monitoring and fault detection methods are increasingly important to achieve a robust and resource efficient operation of wastewater treatment plants (WWTPs). The purpose of this paper was to evaluate a promising machine learning method, Gaussian process regression (GPR), for WWTP monitoring applications. We evaluated GPR at two WWTP monitoring problems: estimate missing data in a flow rate signal (simulated data), and detect a drift in an ammonium sensor (real data). We showed that GPR with the standard estimation method, maximum likelihood estimation (GPR-MLE), suffered from local optima during estimation of kernel parameters, and did not give satisfactory results in a simulated case study. However, GPR with a state-of-the-art estimation method based on sequential Monte Carlo estimation (GPR-SMC) gave good predictions and did not suffer from local optima. Comparisons with simple standard methods revealed that GPR-SMC performed better than linear interpolation in estimating missing...
On-line multivariate statistical monitoring of batch processes using Gaussian mixture model
Computers & Chemical Engineering, 2010
The statistical monitoring of batch manufacturing processes is considered. It is known that conventional monitoring approaches, e.g. principal component analysis (PCA), are not applicable when the normal operating conditions of the process cannot be sufficiently represented by a Gaussian distribution. To address this issue, Gaussian mixture model (GMM) has been proposed to estimate the probability density function of the process nominal data, with improved monitoring results having been reported for continuous processes. This paper extends the application of GMM to on-line monitoring of batch processes, and the proposed method is demonstrated through its application to a batch semiconductor etch process.
Chemometrics and Intelligent Laboratory Systems, 1998
. Ž . In this paper, a combined approach of partial least squares PLS and fuzzy c-means FCM clustering for the monitoring of an activated-sludge waste-water treatment plant is presented. Their properties are also investigated. Both methods were applied together in process monitoring. PLS was used for extracting the most useful information from the control and process variables in order to predict a response variable, namely the diluted sludge volume index. Score values were used in FCM, which utilizes the principle of an object belonging to several classes at the same time instead of just one class. The memberships of each of the classes are defined by the membership values. Corresponding membership plots were used to help in the interpretation of the score plots. Short-term changes were considered to be disturbances and long-term changes due to drifting. q of, or in conjunction with, the quality-related variables. This is justifiable for several reasons. In almost every case, there is much more information in control and process variables than in quality-related variables, and this information can also be gained earlier. Hence, all the possible information from the process itself is used directly. In addition, one is no longer doing post mortem analysis, i.e., only monitoring the quality related variables after the product has already been completed.
Statistical monitoring of a wastewater treatment plant: A case study
Journal of Environmental Management
The efficient operation of wastewater treatment plants (WWTPs) is key to ensuring a sustainable and friendly green environment. Monitoring wastewater processes is helpful not only for evaluating the process operating conditions but also for inspecting product quality. This paper presents a flexible and efficient fault detection approach based on unsupervised deep learning to monitor the operating conditions of WWTPs. Specifically, this approach integrates a deep belief networks (DBN) model and a one-class support vector machine (OCSVM) to separate normal from abnormal features by simultaneously taking advantage of the feature-extraction capability of DBNs and the superior predicting capacity of OCSVM. Here, the DBN model, which is a powerful tool with greedy learning features, accounts for the nonlinear aspects of WWTPs, while OCSVM is used to reliably detect the faults. The developed DBN-OCSVM approach is tested through a practical application on data from a decentralized WWTP in Golden, CO, USA. The results from the DBN-OCSVM are compared with two other detectors: DBN-based K-nearest neighbor and K-means algorithms. The results show the capability of the developed strategy to monitor the WWTP, suggesting that it can raise an early alert to the abnormal conditions.
Statistical Process Monitoring of Correlated Binary and Count Data Using Mixture Distributions
Statistical Process Monitoring (SPM) is used quite extensively in semiconductor manufacturing, particularly in the areas of yield enhancement and contamination (particles) reduction. These variables are typically measured at the Integrated Circuit (IC)-level, and then summarized to the wafer and lot levels. Due to the batch nature of semiconductor processing, ICs residing in the same wafer are normally processed simultaneously. This tends to induce a positive correlation among the measurements taken at the IC level. It is known that within subgroup correlation affects the performance of control charts. In particular, it tends to induce higher than expected false alarm rates. In this paper we present 3-sigma Shewhart charts, for both yield and particle data, and CUSUM monitoring charts for yield data. These charts use mixture distributions to take into account the inherent correlation in the data; thus reducing the false alarm rate. A SAS/AF ® application has been developed to deploy...
Analytica Chimica Acta, 2005
The biomass present in a wastewater treatment plant was surveyed and their morphological properties related with operating parameters such as the total suspended solids (TSS) and sludge volume index (SVI). For that purpose image analysis was used to provide the morphological data subsequently treated by partial least squares regression (PLS) multivariable statistical technique. The results denoted the existence of a severe bulking problem of non-zoogleal nature and the PLS analysis revealed a strong relationship between the TSS and the total aggregates area as well as a close correlation between the filamentous bacteria per suspended solids ratio and the SVI.
Chemometrics and Intelligent Laboratory Systems, 1999
A Kalman filter was developed to overcome the problems caused by process drifting. Different types of models were used to predict response variables of an activated sludge waste-water treatment plant. These models were constructed using MLR, PCR, and PLS. The MLR-type regression coefficients were calculated for both the PCR and PLS models. After that, the Kalman filter was used to estimate these coefficients, recursively. Both the PCR and PLS 'inner relation' coefficient vectors were also estimated in this way and the results were then compared. The effect of the number of variables was also briefly studied. The testing was carried out using sequential process data. The prediction ability was measured by a Q 2 -value as a function of a lag in the updating of the coefficients. q