https://www.mdu.se/

Process control is an important part of any industrial system. In a wastewater reuse system this remains true. Process monitoring and fault detection (FD) are important to ensure that the control system has access to reliable data which can be used in making decisions about the operation of the process. The reuse scenario being considered in this work is that of utilizing the nutrients from the wastewater as fertilizer to agricultural soil along with using the water for irrigation purposes. This paper identifies variables that are important to the control of the process and should be a focus of monitoring and FD. In wastewater treatment these variables include temperatures, pressures, liquid levels, flow rates, pH, conductivity, biomass content, suspended solids concentration, dissolved oxygen content, total organic carbon, and the concentrations of nitrate and ammonium. The variables of interest in the reuse of nutrients and water for agriculture include soil moisture, ambient conditions, plant height, biomass content, photosynthetic activity of the crop, leaf area and leaf water content, as well as the concentrations of several ions both in the soil and in the plant. Challenges associated with process monitoring and FD specific to the two processes are also discussed, examples of these are the high dimensionality of the problem, the harsh conditions that sensors must operate in and the non-linear relationships between variables. This information will be used in future work when comparing specific FD methods to ensure that methods chosen are capable of overcoming the commonly encountered problems.

To prioritise which faults should be detected in a biological wastewater treatment process, and with what level of urgency, it is necessary to understand the effect that they have on the process. Using the Benchmark Simulation Model No. 1 and 2. (BSM1 and BSM2), several process and sensor faults were considered and their impacts on various cost, quality, and controller performance evaluation metrics analysed. Both the cost of treating the wastewater and the quality of the effluent were impacted in varying degrees of severity by the faults tested. The most influential faults in both models were decreases to autotrophic and heterotrophic growth rates, decreases to the heterotrophic death rate, and the inhabitation fault. It was shown that only larger fault sizes were significant, and the required speed of detection is dependent on the fault profile. Prioritising detection of the most influential faults was shown to have significant effects on monitoring requirements for fault detection and the subsequent complexity required of a fault detection system. A valuable takeaway was the similarity of results from BSM1 and BSM2; the consistency of the influential process faults suggests that systems that can be described by these models are likely affected by the same faults.

In a wastewater treatment plant reliable fault detection is an integral component of process supervision and ensuring safe operation of the process. Detecting and isolating process faults requires that sensors in the process can be used to uniquely identify such faults. However, sensors in the wastewater treatment process operate in hostile environments and often require expensive equipment and maintenance. This work addresses this problem by identifying a minimal set of sensors which can detect and isolate these faults in the Benchmark Simulation Model No. 1.Residual-based fault signatures are used to determine this sensor set using a graph-based approach; these fault signatures can be used in future work developing fault detection methods. It is recommended that further work investigate what sizes of faults are critical to detect based on their potential effects on the process, as well as ways to select an optimal sensor set from multiple valid configurations.

Fault detection is an important part of process supervision, especially in processes where there are strict requirements on the process outputs like in wastewater treatment. Statistical control charts such as Shewhart charts, cumulative sum (CUSUM) charts, and exponentially weighted moving average (EWMA) charts are common univariate fault detection methods. These methods have different strengths and weaknesses that are dependent on the characteristics of the fault. To account for this the methods in their base forms were tested with drift and bias sensor faults of different sizes to determine the overall performance of each method. Additionally, the faults were detected using two different sensors in the system to see how the presence of active process control influenced fault detectability. The EWMA method performed best for both fault types, specifically the drift faults, with a low false alarm rate and good detection time in comparison to the other methods. It was shown that decreasing the detection time can effectively reduce excess energy consumption caused by sensor faults. Additionally, it was shown that monitoring a manipulated variable has advantages over monitoring a controlled variable as setpoint tracking hides faults on controlled variables; lower missed detection rates are observed using manipulated variables.

The harsh operating environment in a wastewater treatment process (WWTP) makes sensor faults commonplace. Detecting these faults can be challenging due to the complex process dynamics, unknown inputs, and general noise in the process and measurements. Comparing sensor readings against predictions from a physics-based or data-driven model of the WWTP is a common strategy for detecting such faults. In this work sensor measurements are directly modelled using Gaussian process (GP) regression, a data-driven multivariate approach. These GP sensor models are, with a generalised product of experts, combined into a dedicated fault isolation scheme resembling traditional observer bank methods. The residuals are monitored with a multivariate exponentially weighted moving average chart which is used for fault detection and isolation. The method is evaluated using simulated data generated with the Benchmark Simulation Model No. 1 WWTP. Fault detection performance is reported using several standard metrics such as false alarms, missed detections, time to detection, and successful fault isolations, with emphasis on reporting across a wide range of sensors and faults to provide a point of comparison for future studies. The proposed approach performs well across these metrics. Given sufficient data representative of normal operation, this approach can easily be adapted across a wide variety of plant configurations and can be used to create operatorfriendly diagnostics resembling classical control charts.