Prior Knowledge for Predictive Modeling: The Case of Acute Aquatic Toxicity.

Early assessment of the potential impact of chemicals on health and the environment requires toxicological properties of the molecules. Predictive modeling is often used to estimate the property values in silico from pre-existing experimental data, which is often scarce and uncertain. One of the ways to advance the predictive modeling procedure might be the use of knowledge existing in the field. Scientific publications contain a vast amount of knowledge. However, the amount of manual work required to process the enormous volumes of information gathered in scientific articles might hinder its utilization. This work explores the opportunity of semiautomated knowledge extraction from scientific papers and investigates a few potential ways of its use for predictive modeling. The knowledge extraction and predictive modeling are applied to the field of acute aquatic toxicity. Acute aquatic toxicity is an important parameter of the safety assessment of chemicals. The extensive amount of diverse information existing in the field makes acute aquatic toxicity an attractive area for investigation of knowledge use for predictive modeling. The work demonstrates that the knowledge collection and classification procedure could be useful in hybrid modeling studies concerning the model and predictor selection, addressing data gaps, and evaluation of models' performance.

Integrated waste management in batch chemical industry based on multi-objective optimization.

UNLABELLED: A multi-objective optimization methodology for hazardous liquid waste management is presented in this paper using industrially based LCA models and operating constraints. This approach is used to optimize the handling of waste streams introducing flexible mixing policy scenarios compared to the rigid policy scenarios of the industrial system. It is shown that increasing the degrees of freedom for the waste mixing reduces significantly both the operating cost and the environmental impact by avoiding the use of utilities. Moreover, the influence of waste availability as function of production planning without waste storage is analyzed in several multiperiod optimizations. There, it is demonstrated that this saving potential can be further increased by integration of multiperiod production planning with waste management policies, up to the level of 40% for the environmental impact, and more than 50% for the operating cost, compared to the industrial base case. In some specific cases, a proper matching of production planning and waste mixing policies can also turn the waste treatment into a source of profit exploiting energy production from the incineration process. IMPLICATIONS: This study reveals the savings potential of more flexible policies in waste management, in particular waste mixing of liquid waste in batch chemical industries treated in incineration, wet air oxidation, wastewater treatment plants, or recovered by distillation. Through a multi-objective optimization framework including models for operating costs and life-cycle inventories based on industrial data, operating constraints from industrial practice, and terminal constraints from legislation, savings potentials up to 50% for the operation cost and 40% for the environmental impact are demonstrated in two case studies.

Molecular-structure-based models of chemical inventories using neural networks.

Chemical synthesis is a complex and diverse procedure, and production data are often scarce or incomplete. A detailed inventory analysis of all mass and energy flows necessary for the production of chemicals is often costly and time-intensive. Therefore only few chemical inventories exist, even though they are essential for process optimization and the environmental assessment of many products. This paper introduces a newtype of model to provide estimates for inventory data and environmental impacts of chemical production based on the molecular structure of a chemical and without a priori knowledge of the production process. These molecular-structure-based models offer inventory data for users in process design and optimization, screening life cycle assessment (LCA), and supply chain management. They can be applied even if the producer is unknown or the production process is not documented. We assessed the capabilities of linear regression and neural network models for this purpose. All models were generated with a data set of inventory data on 103 chemicals. Different input sets were chosen as ways to transform the chemical structure into a numerical vector of descriptors and the effectiveness of the different input sets was analyzed. The results show that a correctly developed neural network model can perform on an acceptable level for many purposes. The models can assist process developers to improve energy efficiency in all design stages and aid in LCA and supply chain management by filling data gaps.

Comparison of recent methods for inference of variable influence in neural networks.

Neural networks (NNs) belong to 'black box' models and therefore 'suffer' from interpretation difficulties. Four recent methods inferring variable influence in NNs are compared in this paper. The methods assist the interpretation task during different phases of the modeling procedure. They belong to information theory (ITSS), the Bayesian framework (ARD), the analysis of the network's weights (GIM), and the sequential omission of the variables (SZW). The comparison is based upon artificial and real data sets of differing size, complexity and noise level. The influence of the neural network's size has also been considered. The results provide useful information about the agreement between the methods under different conditions. Generally, SZW and GIM differ from ARD regarding the variable influence, although applied to NNs with similar modeling accuracy, even when larger data sets sizes are used. ITSS produces similar results to SZW and GIM, although suffering more from the 'curse of dimensionality'.