No abstract

Thermal alkaline hydrolysis is a common pretreatment method for the utilization of excess activated sludge (EAS). Owing to strict environment laws and need for better energy utilization, new methods were developed in this study to improve the efficiency of pretreatment method. Direct thermal hydrolysis (TH), pasteurized thermal hydrolysis (PTH), and alkaline pasteurized thermal hydrolysis (PTH + CaO and PTH + NaOH) methods were used to treat EAS. Each method was compared and analyzed in terms of dissolution in ammonium nitrogen (NH 4 + -N) and soluble COD (SCOD) in EAS. Furthermore, the removal of tetracycline resistance genes (TRGs)and class 1 transposon gene intI1 from EAS was investigated. The NH 4 + -N and SCOD concentrations in EAS treated by PTH were 1.24 and 2.58 times higher than those of TH. However, the removal efficiency of total TRGs and intI1 between the groups was comparable. The SCOD concentration of the PTH + NaOH group was 4.37 times higher than that of the PTH group, and the removal efficiency of total TRGs was increased by 9.52% compared with that by PTH. The NH 4 + -N and SCOD concentrations of the PTH + CaO group could reach 85.04% and 92.14% of the PTH + NaOH group, but the removal efficiency of total TRGs by PTH + CaO was 19.78% lower than that by PTH + NaOH. Thus, to reduce the financial cost in actual operation, lime (CaO) can be used instead of a strong alkali (NaOH), and pasteurized steam at 70°C instead of conventional high-temperature heating to treat EAS. This study provides a reference for the development of alkaline hydrolysis under moderate temperatures along with the removal of TRGs in EAS.

Substantial work indicates that the dynamics of neural networks (NNs) is closely related to their initialization of parameters. Inspired by the phase diagram for two-layer ReLU NNs with infinite width , we make a step towards drawing a phase diagram for three-layer ReLU NNs with infinite width. First, we derive a normalized gradient flow for three-layer ReLU NNs and obtain two key independent quantities to distinguish different dynamical regimes for common initialization methods. With carefully designed experiments and a large computation cost, for both synthetic datasets and real datasets, we find that the dynamics of each layer also could be divided into a linear regime and a condensed regime, separated by a critical regime. The criteria is the relative change of input weights (the input weight of a hidden neuron consists of the weight from its input layer to the hidden neuron and its bias term) as the width approaches infinity during the training, which tends to 0, +∞ and O(1), respectively. In addition, we also demonstrate that different layers can lie in different dynamical regimes in a training process within a deep NN. In the condensed regime, we also observe the condensation of weights in isolated orientations with low complexity. Through experiments under three-layer condition, our phase diagram suggests a complicated dynamical regimes consisting of three possible regimes, together with their mixture, for deep NNs and provides a guidance for studying deep NNs in different initialization regimes, which reveals the possibility of completely different dynamics emerging within a deep NN for its different layers.