Output-relevant Variational autoencoder for Just-in-time soft sensor modeling with missing data

Main challenges for developing data-based models lie in the existence of high-dimensional and possibly missing observations that exist in stored data from industry process. Variational autoencoder (VAE) as one of the deep learning methods has been applied for extracting useful information or features from high-dimensional dataset. Considering that existing VAE is unsupervised, an output-relevant VAE is proposed for extracting output-relevant features in this work. By using correlation between process variables, different weight is correspondingly assigned to each input variable. With symmetric Kullback–Leibler (SKL) divergence, the similarity is evaluated between the stored samples and a query sample. According to the values of the SKL divergence, data relevant for modeling are selected. Subsequently, Gaussian process regression (GPR) is utilized to establish a model between the input and the corresponding output at the query sample. In addition, owing to the common existence of missing data in output data set, the parameters and missing data in the GPR are estimated simultaneously. A practical debutanizer industrial process is utilized to illustrate the effectiveness of the proposed method.     

Coordinated scheduling of production and delivery from multiple plants

This paper deals with the scheduling of orders and vehicle assignment for production and distribution planning in a scenario of no-wait, immediate delivery to the customer site. We first describe the problem and then present an integer programming model that maximizes the weighted value of the orders served. We consider a special case of the problem which can be solved in polinomial time by a minimum cost flow algorithm. Based on this approach we develop a heuristic procedure for the general case. Comparisons with an exact graph-based method attest that our heuristic produces good-quality solutions in short running times.     

Scale-up in the 4th dimension in the field of granulation and drying or how to avoid classical scale-up

In the pharmaceutical industry, the production of granules is based on a batch concept. This concept offers many advantages, as a batch can be accepted or rejected. However, the scale-up of the batch size may lead to problems. The variety of the equipment involved does not facilitate the scale-up process and the capital invested in space and equipment is high. An alternative approach is the use of a continuous process. However, continuous processes have the disadvantage among others that the batch size is not well defined. Thus, a special quasi-continuous production concept was developed, taking into account the advantages of a batch type and a continuous process. This concept was developed in cooperation with the Institute of Pharmaceutical Technology of the University of Basel, Glatt CH-4133 Pratteln and F. Hoffmann-La Roche, CH-4070 Basel. The equipment allows to implement a “Just in Time Production Concept” as a large batch B consists of n subunit (SU) batches b, i.e. B=nb. The subunit batch b corresponds to, e.g. 7-kg material for the production of pharmaceutical granules for further processing such as tabletting. At the Roche production site, this novel process equipment was used to manufacture batch sizes B with n=10, n=100 and so far up to n=600 subunits. This leads to an optimal use of capital invested in GMP space and equipment. The difference to the classical scale-up is the following: with classical scale-up, the dimensions of the equipment x, y, z is enlarged and the process time is more or less kept constant. With this novel concept, the dimension x, y, z of the equipment is kept constant and the process is repeated in the 4th dimension “n times”. Thus, for the scale-up in the 4th dimension, i.e. in the time, the equipment needs to show a “self-cleaning” property and appropriate formulations. The novel concept is of special interest, as the quality of the product is not changed during scale-up.     

An iterated-local-search heuristic for the resource-constrained weighted earliness-tardiness project scheduling problem

This paper deals with the weighted earliness-tardiness resource-constrained project scheduling problem with minimum and maximum time lags (WET-RCPSP/max). The problem consists of scheduling the activities of a project subject to prescribed resource and temporal constraints such that the total weighted deviation of the activities' completion times from prescribed due dates is minimized. Key applications are planning of just-in-time production and reactive scheduling. For the (approximative) solution of the WET-RCPSP/max, we present a population-based iterated-local-search heuristic. We also report the results of an experimental performance analysis where this heuristic outperformed state-of-the-art methods.     

Maximizing the weighted number of just-in-time jobs in flow shop scheduling

In this paper we consider the maximization of the weighted number of just-in-time jobs that should be completed exactly on their due dates in n-job, m-machine flow shop problems. We show that a two-machine flow shop problem is NP-complete. When job weights are all identical, we show that the problem can be solved in polynomial time. We also show that a three-machine flow shop problem with identical job weights is NP-hard in the strong sense by reduction of the 3-partition problem.     

基于JITL的电量预测高斯软测量建模研究

为了解决工业过程受本身结构特征、外界因素等影响而存在严重的非线性和时变性等问题,本文提出了一种基于输入输出综合性相似度指标的即时学习高斯过程软测量建模方法。在该方法中,将样本数据进行归一化处理,首先利用传统的基于距离和角度的相似度指标分别对样本输入输出变量进行相似度计算,进而对相似度进行综合,最后选择出最终的相关样本集,建立高斯过程回归软测量模型,将所提基于输入输出相似度指标的即时学习高斯工程软测量模型应用于城市日用电量数据的预测。研究结果表明,所提出的软测量建模方法可以实现对日用电量数据的高精度预测且预测结果具有较小的误差。因此可表明该方法可在电量预测中具有一定的应用可靠性,可以在电力市场预测分析中得到广泛的应用。     

基于延时开工的单件准时调度问题研究

在现代制造企业中,准时调度已成为企业获得利润和效益的关键.针对准时加工的要求,代价度的概念被首次提出,并利用遗传算法的嵌套使用,对单件调度问题给出了以代价度最小为调度目标的延时开工调度方法,通过计算机的模拟得出了结果,并对交货期和延时开工/提前完工系数比两个重要参数变化而引起的调度结果的变化进行了研究.