Browse Prior Art Database

RO SYSTEM PERFORMANCE ANALYTICS FOR DEGRADATION DETECTION AND ISOLATION

IP.com Disclosure Number: IPCOM000250011D
Publication Date: 2017-May-16
Document File: 4 page(s) / 107K

Publishing Venue

The IP.com Prior Art Database

Abstract

The invention discloses systems and methods to detect and predict degradation either in the membrane filtration process or the sensors of an RO plant. The system includes an RO plant fitted with sensors and instrumentation for measuring various process parameters, and a computing system to process the data, and to run the predictive model using the monitoring data. The method includes the steps of acquiring process parameters or sensor data over different timeframes, fitting trends to the data and identifying anomalies by comparing with established signature trends and thresholds that are indicative of membrane or sensor degradation. The data is extrapolated to identify time indices where the thresholds will be exceeded, to calculate remaining useful life. The developed methodology can be used to extend the operating life of membranes and reduce operating expenses. Furthermore, this methodology can be effectively applied to specific modules and to optimize maintenance strategies.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 41% of the total text.

1

RO SYSTEM PERFORMANCE ANALYTICS FOR DEGRADATION DETECTION AND ISOLATION

BACKGROUND

The invention relates generally to water treatment plant and in particular to systems and methods for monitoring and isolating operating problems in reverse osmosis (RO) plants. This invention addresses the issue of detecting faults in either the membrane or the sensors that are used to monitor the plant. Currently, remaining useful life (RUL) calculations are performed on normalized permeate flow (NPF) as well as normalized salt rejection (NSR) and the minimum value of the two is reported as the plant RUL. Additionally, an RUL for normalized differential pressure (NDP) can also be calculated and this number can be compared against the numbers for NPF and NSR and the global minimum can then be used to determine the plant RUL. Faults/changes in sensor signals are detected using certain change detection rules which search for potential outliers, signal drifts (ramp changes), signal step changes, to name a few. However, many of these tasks are currently performed with a lot of manual judgment or intervention.

BRIEF DESCRIPTION OF DRAWINGS

The invention discloses system and method for RO plant degradation detection and isolation in which:

FIG. 1 illustrates the method of monitoring RO plant and predicting remaining useful life of components.

DETAILED DESCRIPTION

The invention discloses systems and methods to detect and predict degradation either in the membrane filtration process or the sensors of an RO plant. The system includes an RO plant fitted with sensors and instrumentation for measuring various process parameters, and a computing system to process the data, and to run the predictive model using the monitoring data. The method of the invention is illustrated with reference to FIG. 1 and FIG. 2.

In a first step, data is checked for anomalies in parameters such as NaN; flow rate, pressure, conductivity and pH less than or equal to 0, negative temperature (deg C), equal feed and permeate pressure. The anomaly indices pertaining to these variables are aggregated and removed across all key performance indicators (KPIs). Outliers are then removed from the data. Outliers are those data points that are outside the bound of plus or minus k-sigma from the data mean. Statistical methods, depth, deviation, density and distance based approaches are used for outlier removal. Data smoothing is then used to remove high frequency noise. Typical smoothers that may be used are mean, median, single, double, triple exponentials, exponentially weighted moving average, recursive modified moving average, Kalman, fixed point/interval/lag or kernel based filters. In the

2

next step, in-house empirical standards may be used to eliminate operational variations through normalization and study process faults.

FIG. 1: Method of monitoring RO plant and predicting remaining useful life of components

Sensor changes are classified as outliers, steps or ramps and then analyzed by the system....