Reducing Capital Cost in Process Design with Digital Twins: Two Full-Scale Case Studies

INTRODUCTION
Digital twins (DTs) are emerging as powerful technologies in the digital transformation of wastewater treatment facilities (WWTFs). While most DT applications currently focus on optimizing operations using live data and integrated modeling, to the authors knowledge no studies have explored their potential in the design phase of a project. Currently, due to insufficient data, process design assumptions often include multiple overlapping conservative parameters, leading to higher capital costs, such as assuming the coldest temperature, peak loadings, and peak flows occur simultaneously. By integrating commonly available data, DTs can provide high-quality, high-resolution dynamic data that were not previously available to reduce design uncertainty, mitigate risks, and reduce capital costs. Yang et al. proposed using DTs for soft sensing dynamic influent profiles, which addresses this data insufficiency challenge for design. High-resolution dynamics improve control design and equipment sizing accuracy. This paper investigates the benefits of using DTs in the process design phase, presenting the two full-scale case studies on incorporating DTs into process design:
1. Marine Park WWTF (City of Vancouver, Washington, US): improving existing aeration basin performance with advanced aeration control and mixed liquor recirculation to address increasing influent loads and sludge popping issues in the clarifiers.
2. Davyhulme WwTW (Greater Manchester, UK): generating a long-term influent dynamic profile that can be used by the design teams to minimize conservatisms and expansion costs for a number of planned expansion projects across the liquid streams.

METHODOLOGY
1. Dynamic Influent Profiles Estimation. Historical data were fed into the digital-twin-based Soft Sensor to generate the 15-min dynamic influent profiles along with biological kinetics, stoichiometry, and fractionation. The estimated influent profiles were collapsed into daily composite values and compared with measurements to verify their accuracy.
2. Design Option Analysis with Selected Periods. Depending on design objectives, relevant periods were selected to test design options.

RESULTS AND DISCUSSION
The measurements from the soft sensor were compared with those from both physical sensors and laboratory tests, and both comparisons showed a good match. The Davyhulme WwTW has two regularly maintained ammonia probes at the primary effluent, which allows for a direct comparison between the soft sensor and physical sensors, as shown in Figure 1 with performance metrics listed in Table 1. It is observed that the shape and magnitude of the signals (i.e., the timing of peaks and valleys, inclines and slopes) match well. With Table 1, it can be inferred that the soft sensor should be able to reproduce similar accuracy results as physical sensors. Comparison with lab measurements is provided in Figure 2 and Table 2. Additionally, the soft sensor generated dynamic biological kinetics, stoichiometry, and fractionation that can be used in design (Figure 2). The Marine Park WWTP soft sensor has a similar performance.

The dynamic influent profiles were used to refine process design by reducing uncertainty, optimizing equipment selection and sizing, and ultimately saving significant capital costs. The Marine Park case study selected four periods to analyze design options for 7-day dynamic maximum, average, minimum loading and wet weather conditions (Figure 4). Finding and design implications are summarized below:
- Influent peak loading, peak flow, and coldest temperature periods do not occur simultaneously, which can lower safety factor to reduce capital costs.
- Ammonia-based-aeration-control is proved to be a preferred advanced aeration control mechanism that is flexible for different aeration objectives with dynamic simulations.
- With dynamic influent profiles under various scenarios, the mixed liquor recirculation (MLR) design was evaluated to avoid oversizing the pumps and thus save equipment costs and energy.
- Investigations on aeration demand during minimum loading conditions eliminate the installation of aerobic zone mechanical mixers originally defined in the project scope, saving unnecessary equipment costs.

With these insights, the primary design was further refined, and the cost savings are detailed in Table 3. The reduction in MLR pump sizing and the removal of mechanical mixers, as suggested and verified by the DT and soft sensor, resulted in a total capital cost savings of $7,367,000 for this project.

CONCLUSIONS This paper presents two case studies utilizing digital twin technology to refine process design by reducing conservativeness, optimizing equipment selection and sizing, and ultimately saving significant capital costs. By leveraging a digital twin soft sensor, high-quality and high-resolution influent profiles were developed from available historical data that reduces design uncertainty. This approach unlocks process design possibilities previously restricted by insufficient dynamic data. It is the first attempt to integrate digital twin tools in supporting process design and has proven to be practical. Its most significant contribution lies in re-envisioning the use of digital twins beyond operation improvement and encouraging further exploration of their potential benefits.