基于ART1神经网络的统计过程控制系统

A Quality Control System Based on ART1 Neural Network

统计过程控制(SPC)在改进过程水品、提高产品质量方面作出了巨大贡献。本文讨论了一种基于自适应谐振理论(ART)神经网络的SPC系统。与一般SPC系统相比,本系统不仅可以在线检测过程异常,对各种控制图异常模式还具有实时学习、在线识别功能。同时,本系统对过程的分析,无需如常规控制图一样,建立在正态假设的前提下,因此应用更方便、范围更广泛。作为一种新的SPC工具,ART1神经网络为改进控制图的应用提供了一种新的可能。

This paper presents a statistical process control （SPC） system, which is based on the adaptive resonance theory （ART） neural network. Contrast to these common SPC systems, it can not only detect the unnatural process behavior on line, but can also learn and identify these unnatural patterns on control charts at the same time. And it does not require the assumption of normal distribution assumptions. As a new promising SPC tool, ART1 network proposes a possibility of improving the application of control charts.