Abstract Electrodialysis is an important chemical process that separates pollutants from wastewater pools to produce clean water for consumption and irrigation. Initial wastewater concentration of chemical elements always differs. Chemical components are strongly dependent on the efflux origin and treatment. To optimize an electrodialysis process is congruent to improved key water quality characteristics. To predict optimal electrodialysis performance there will always be a need to conduct a small number of structured experiments. This is because wastewater conditions are usually different in each situation thus requiring reliable evidence-based design decisions to be delivered timely and low-cost. We study a real example from crucial dessert wastewater operations that aim to supply clean water for irrigation. Several issues are scrutinized that are often overlooked when carrying out multi-response multi-factorial statistical optimization in environmental screening. Programming fast-cycle trials with Taguchi-type factorial recipes reaps quick information for new development and improvement projects. But it also introduces phenomena such as saturation, unreplication and non-linearity that could undermine the optimization effort. The showcased paradigm uses popular Taguchi methods to organize a rapid and short round of trials in order to investigate the behavior of four electrodialysis controlling factors: 1) the dilute flow, 2) the cathode flow, 3) the anode flow and 4) the voltage. The three monitored water quality indices are: 1) the percentage of removed sodium cations, 2) the sodium adsorption ratio and 3) the sodium ratio. We discuss the intricacies that emerge from the synthetic type of the electrodialysis data: non-normality, non-linearity and messiness. We propose a robust and agile method to conduct the multi-response multi-factorial optimization for electrodialysis of polluted wastewater. It is based on super-ranking and distribution-free profiling. Comparison with other profiling methods is provided and main advantages are commented from a chemical engineering perspective.

中文翻译：

---

使用稳健的非线性田口分析对污染废水的电渗析过程进行并发多响应多因素筛选

摘要 电渗析是一种重要的化学过程，可将污染物从废水池中分离出来，以生产用于消费和灌溉的清洁水。化学元素的初始废水浓度总是不同的。化学成分强烈依赖于外排来源和处理。优化电渗析过程与改善关键水质特征是一致的。为了预测最佳电渗析性能，总是需要进行少量的结构化实验。这是因为每种情况下的废水条件通常都不同，因此需要及时和低成本地提供可靠的基于证据的设计决策。我们研究了旨在为灌溉提供清洁水的关键甜点废水运营中的一个真实示例。在环境筛选中进行多响应多因素统计优化时，经常会忽略几个问题。使用田口型因子配方对快速循环试验进行编程，可为新的开发和改进项目获取快速信息。但它也引入了可能破坏优化工作的饱和、非复制和非线性等现象。展示的范式使用流行的田口方法组织了一轮快速而短暂的试验，以研究四个电渗析控制因素的行为：1) 稀释流，2) 阴极流，3) 阳极流和 4) 电压. 监测的三个水质指标是：1）去除钠阳离子的百分比，2）钠吸附率和3）钠比。我们讨论了电渗析数据合成类型的复杂性：非正态性、非线性和混乱。我们提出了一种稳健而敏捷的方法来对污染废水的电渗析进行多响应多因素优化。它基于超级排名和无分布的分析。提供了与其他分析方法的比较，并从化学工程的角度评论了主要优点。