Multivariate data analysis (MVA) is the investigation of many variables simultaneously, to understand the relationships that may exist between them. Multivariate data analysis methods have been around since the 1960’s, but until recently were primarily used in laboratories and technical groups, rarely being applied to production processes.
Today, this is changing as manufacturers looking for a competitive advantage realize that the masses of data collected during production operations offer great insights to improve product development and process performance. Because manufacturing processes are typically highly multivariate in nature – i.e. a large number of variables which are usually interactive – they require multiple measurements to fully understand them.
然而，大多数统计过程控制（SPC）系统依赖于单变量方法仅看单一变量，一次一个。 单变量统计分析趋向复杂系统的时候，因为他们无法检测到变量之间的关系，往往是工艺失常的原因失败。 多变量分析工具，使制造商能够更好地了解进程的行为，并实施更强有力的控制策略。 这使得他们能够更接近极限运行的进程，使用成本更低的组件，减少能源的使用，减少循环时间，减少浪费和返工。
Manufacturers who have adopted these tools can quickly see improvements in their operations and on their bottom line. Brad Swarbrick, Vice President of Business Development at CAMO Software, explains, “We worked with a food manufacturer who had a quality problem which they could not identify the cause of using their SPC tools. After analyzing the data with multivariate methods, they worked out what the issue was and adjusted their process accordingly, saving them around €1 million per year in scrap, rework and energy costs alone.”
He continues “Another client in the pharmaceutical sector applied multivariate analysis together with Near Infrared (NIR) spectroscopy to monitor their fluid bed dryers to achieve consistent product quality using timeline and process trends.”
CAMO’s solutions allow multivariate models to be developed by Technical Service groups or CAMO’s consultants, and then applied to real-time production processes. The solutions can be used standalone to analyze off-line data, connected to databases or scientific instruments such as spectrometers, or integrated with control systems for use by process operators.
For more information download CAMO’s free e-book Multivariate Data Analysis for Dummies at www.camo.com /杆/ MVA作傻瓜