Chang Peng

Bio: Chang Peng is an academic researcher from Beijing University of Technology. The author has contributed to research in topics: Fault (power engineering) & Moment (mathematics). The author has an hindex of 2, co-authored 6 publications receiving 12 citations.

Batch process fault monitoring and diagnosis method based on width learning system

Abstract: The invention discloses a batch process fault monitoring and diagnosis method based on a width learning system to meet requirements of real-time monitoring of actual industrial processes. A rapid neural network monitoring model is established which specifically includes two stages of 'offline modeling stage' and 'online monitoring stage'. The 'offline modeling phase' includes the steps of: first classifying the data to obtain N fault states, then establishing a monitoring submodel of the corresponding fault, and finally testing and adjusting the submodel. The 'Online monitoring' includes the steps of: reading in new time data, inputting N fault monitoring sub-models established in the offline modeling stage, obtaining all monitoring values to determine whether the fault is generated or not. The batch process fault monitoring and diagnosis method accelerates the modeling and monitoring speed while ensuring the accuracy of the monitoring, achieves real-time diagnosis, and finally obtainsa batch process fault monitoring and diagnosis method with excellent performances.

Complex industrial process fault monitoring method based on OICA

Abstract: The invention discloses a complex industrial process fault monitoring method based on OICA, The invention is used for improving the accuracy of industrial process fault monitoring and has important significance in improving the safety of a production process to ensure the safety of production equipment and production personnel and improving the product quality. The method comprises two stages of off-line modeling and on-line monitoring. The step of off-line modeling comprises the following steps: preprocessing original data; extracting independent main components and residual errors of the data by adopting OICA; and calculating a control limit through kernel density estimation in the independent component space and the residual space respectively. The step of on-line monitoring comprises the following steps: preprocessing data at the current sampling moment; and calculating the statistics of the data at the current moment, and comparing the statistics with the control limit to judge whether the fermentation process runs normally. The method does not need to assume that data obeys Gaussian distribution, is low in calculation complexity, and is not limited by a hybrid matrix form, sothat the fault monitoring effect is better.

Angle similarity stage division and monitoring method in microbial pharmacy process

Abstract: The invention discloses an angle similarity stage division and monitoring method in a microbial pharmacy process. In order to better process multi-stage characteristics in the penicillin fermentationprocess, an effective fault monitoring model based on a multi-stage division method is established. The method comprises two stages of off-line modeling and on-line monitoring. The off-line modeling comprises the following steps: firstly, expanding three-dimensional data of a fermentation process along a time axis; dividing the data into C0 sub-periods; and then establishing respective KECA modelsby using the sub-period data, finally calculating T2 and SPE statistics of the data, and determining the control limit of the statistics in each period. The on-line monitoring comprises the steps ofprocessing newly collected data according to a model, dividing the data into sub-periods, calculating the statistics of the data, and comparing the statistics with a control limit to judge whether theproduction process is faulty or not. According to the method, the multi-stage characteristics of the intermittent process are fully considered, and the fault monitoring accuracy is satisfactory.

Randomized algorithms for distributed computation of principal component analysis and singular value decomposition

Abstract: Randomized algorithms provide solutions to two ubiquitous problems: (1) the distributed calculation of a principal component analysis or singular value decomposition of a highly rectangular matrix, and (2) the distributed calculation of a low-rank approximation (in the form of a singular value decomposition) to an arbitrary matrix. Carefully honed algorithms yield results that are uniformly superior to those of the stock, deterministic implementations in Spark (the popular platform for distributed computation); in particular, whereas the stock software will without warning return left singular vectors that are far from numerically orthonormal, a significantly burnished randomized implementation generates left singular vectors that are numerically orthonormal to nearly the machine precision.

Development of Soft Sensor Based on Sequential Kernel Fuzzy Partitioning and Just-in-Time Relevance Vector Machine for Multiphase Batch Processes

Abstract: Batch processes are important manufacturing approaches widely used in modern industry. During the manufacturing process, quality prediction is essential. Data-driven-based soft sensors have widely used for quality prediction because of the advantages of simple application and high flexibility. However, they do not perform well due to the characteristics of dynamic, nonlinear, and multiphases in batch processes. To address these issues, a soft sensor based on sequential kernel fuzzy partitioning (SKFP) and just-in-time relevance vector machine (JRVM), named SKFP-JRVM, is proposed in this work. First, an SKFP algorithm is proposed to divide batches into phases. In the SKFP algorithm, the inner and the sequential membership are constructed and compared to obtain highly accurate phase partitioning results. Meanwhile, a partitioning evaluation index is applied to automatically determine the optimal number of phases. Moreover, a soft sensor model based on the JRVM is built for each phase. The JRVM model not only considers the modeling samples of adjacent phases but also effectively addresses the problem of soft sensor modeling of process data with highly nonlinear and dynamic characteristics. The effectiveness of the SKFP-JRVM method is demonstrated on a numerical simulation and a penicillin fermentation process.