[中文]21世纪过程管理使用基于统计过程控制的制造业分析学
时常听到这样的观点,我们处在一个新的时代,无法在制造过程中收集足够数据不再是一个问题了,但是如何有效地使用我们收集到的数据呢。这个新时代的主要战斗口号之一是,“更好地分析数据是绝对必要的!”(然后,随着那个战斗口号的常常是对某人特定风味的分析技术或研究方法的总结。)意外的是,作为解释这个技术如何促成关键行为指示(KPI)过程,术语“关键行为指示(KPI)”也被提到了。
坚持那个确定的方向,让我们以新的眼光审视一种确定的技术。
传统的质量与统计过程控制规格分析
统计过程控制(SPC)是一个由来已久且较好展示的过程管理方法。研究过现代制造业的人都知道戴明博士和他在战后日本建立统计过程控制应用标准的早期工作。统计过程控制很长时间以来被质量部门和实验室用于衡量和监控大部分工业制造设备的质量。
统计过程控制经历过周期性的改造和更新,例如过程持续改进(CPI)和全面质量管理(TQM),就像我们在每个流行周期中学到更多的东西一样。当然,统计过程控制是六西格玛的关键部分,并且很多制造业公司已经投资于精益六西格玛。(译者注:六西格玛是一种类似于ISO9000的质理管理体系)
这些统计过程控制应用中的多数提供数据分析,或者在实验室中基于生产过程末端的产品样品质量监控事例中的“犯罪之后”,或者在六西格玛过程事例中的从直接“可控告的判决”中移除的某种事物。
传统统计过程控制方法的典型应用包括成功的质量/过程管理功能,例如:
常规统计质量控制(SQC)/统计过程控制(SPC)报告
过程监控与改进
实验室中的质量分析方法
强制性测试
提供连锁客户认证
新兴的实时运行决策支持统计过程控制分析
尽管统计过程控制分析方法遇到这些功能的挑战,它也是合适的工具以符合那个新兴的战斗口号,更好地分析有效数据。对于统计过程控制方法,基于时间的比较分析和分析的可视化表示是独特的,使制造商能够更好地理解他们的过程,并且更重要的为了基于数据分析采取立即行动。统计过程控制是为满足当前制造业需要的适当的技术和方法,求助于数据分析以提供:
有能力衡量系统投资的投资报酬率
适时且有效的分析总结和报告
预测能力
可确认的效益(更低的成本,更高的利润)
支持基于结果的直接可控告的判决过程
使过程改进的高可信度。
公司参与相关的改进方案过程
使为个人的信息需求得到明确的基于角色的结果变得容易
减少集合数据的复杂计算以使带上下文的信息富有意义并可衡量
立即可测量的采取行为结果
如果战斗口号是为了更好的分析,那么实时,可控告,支持判决的分析是关键点。再者,统计过程控制技巧,方法和特性借助自身到达这些要点。提供制造业经营者用实时统计过程控制导出分析的能力,为了快速确认“失控”的参数,用足够简单的格式可视化描述,是目标。这个实时的描述可以采取嵌入人机界面(HMI)显示器的健全的统计过程控制应用程序的形式。这使大部分有效实时数据的捕获和分析变得容易,提供经营者直接访问分析结果的可视图像,允许为经营者提供预防性文件或纠正采取行为的机制,并为新兴角色提供装进整合企业分析学的人机界面。
从噪音中分离信号
确认过度变异的原因
监控持续过程改进行为的实时结果(持续过程改进是统计过程控制的实际投资回报)
对稳定过程检测的预测问题。
提供依从客户供应链需求的文档工具
基于统计过程控制的制造业分析学
随同更好分析数据的战斗口号,对商务智能和商情分析工具的使用也一起增加,并且要求把生产数据和分析与业务数据和分析整合到一起,这都在于要更好地理解企业绩效的全部内容。
这个新的机会产生对可以在接合环境中执行的分析工具的需求,并提出这些目标。以基于统计过程控制的制造业分析学的形式出现的统计过程控制方法的出现,对抓住这个机会是很恰当的。
基于统计过程控制的制造业分析学是统计学的且基于规则的,提供聚合,分析和基于角色的可视化技术与使用户能够更好地理解和改进他们过程的生产数据报告,确认并强化最佳实践,对过程事件的快速反应,在对产品质量产生影响之前,预期潜在的问题,利润,或成本。基于制造业分析学的统计过程控制的分析数据方法的主要鉴别要素是:
基于统计地
重点在于基于角色的分析和报告
识别重要的事件,隔开“噪音”
着重于为实现快速分析的可视化描述技术
提供反应的和预测的行为
足够容易实现,维护,并被工厂中的职员使用
保持统计上正当性的同时,从不同的源获得数据
提供简述健全系统需要的ISA S-95生产性能分析活动模型,方法,和为了提高做出基于广泛和不同分析方法的很知情决策的能力的工具。
合并商业级分析和制造级分析的结果是用来监控一个操作或企业的全部状态和行为的值参数。这些表示为关键行为指示(KPIs),通常是一个包括财政,操作总合的单一参数,和一个提供有意义和可信的KPI变量的实测参数。在一些基于网页的可视化工具上,监控到这些变量的值常常是“好”或“坏”的状态,例如门户或仪表板。能够增强基于网页的可视化工具功能是使用新的分析方法的关键组件的机会。
基于制造业分析学的统计过程控制方法充许一个系统创建并监控形成KPI的全部参数组件的稳定性和变化,在KPI自己显示越界之前,可以检测到主要KPI组件的变化。那么这个检测的可视化表示可能不只是显示“好”(绿)或“坏”(红)的状态,而是如“可能变得更坏”(黄)的状态。目前的KPI分析和报告系统不具有访问全部参数或用这种方式表示全部组件的能力,使得操作和管理能够快速确认,甚至预测,对早期有害征兆的变化采取应对的动作。
选择一个统计过程控制分析平台
一个好的统计过程控制平台应该能够容纳这里概述的生产环境中的全部三个级别的分析学。
传统-统计过控制质量和规格分析
新兴-支持统计过控制分析的实时操作决策
新-统计过程控制驱动制造业分析学
另外,这个平台在建筑结构上必须充许围绕不同级别的需求发展扩张和一个制造商可能需要的分阶段实施。这需要一个基于允许整合进目前的和完全不同的系统的方法的模块和组件,用一个开放而且标准的基于访问数据以对其分析的方法。一个能够提供实时统计过程控制分析功能的系统,与特定的角色耦合,可视化报告,可以支持所有级别从工厂厂房到管理部门的全范围的分析和决策。
案例分析
下面的例子说明基于制造业分析学的统计过程控制如何实施以得到实际的利益并改进操作过程。
软包装-离线和在线的实时统计过程控制分析支持持续改进方案
一个大的国际软包装制造商为不同食品工业客户生产类似形式的许多不同产品。每个客户的产品规格略有不同,为满足供应链需求针对这些规格对生产严密监控是必须的。一旦生产线开始运行,目标是保持生产线的运行,并只对设备做必要的改动以保证产品符合规格。
在实时的基础上,通过使用统计过程控制分析电子采集的测试样品,然后为操作者用一种非常基础的规格/控制的“进或出”的可视化显示方式描述结果,操作者能够检测并做出合适的调整以保持生产线运行,并可以减少变异和降低回收产品的数量。
操作者和质量检查员直接从这个界面上获得有效的信息。他们能够看到历史数据的总结,以及为了分析过程的实时统计过程控制图表。这些特定作用的报告使工作人员能够完整的访问标准操作程序,并支持统一工作流程。
统计过程控制用来建立过程控制限定并支持主动质量控制。工程师用软件做工艺研究,而且工作过程能力成为规范。减少变异,减少废弃物,并减少超额产品的方法是持续过程改进计划的一部分。
作为能动的使用统计过程控制的结果,工厂已经能够通过限制生产过程中重量的变化改进操作。在过去两年,目标重量和实际重量的差别可衡量并可定出基准。测量的结果显示从项目开始之后,累积节约的费用超过20万美元。
食品安全-统计过程控制分析用于早期预测
一个家禽处理工厂监控规定的细菌,例如大肠杆菌,沙门杆菌属。当病菌水平超过限定,要采取的行动费时且代价高。超出安全级别的事情会很快地发生,只有很少的时间做出纠正动作。设定任意的“反应”限制可以在假阳性和丢失信号之间取得平衡。
积极地使用统计过程控制和积极地管理过程使快速和知情的反应成为可能。在一个早期不稳定过程中检测应用制造业分析学的统计过程控制违反规则的事件和模式的结果,预测相似的事件,使避免重复发生的纠正动作和过程改进成为可能。
通过使用持续过程改进,在注意病菌水平的情况下,家禽工厂能够获得很高性能的生产(CPK=12.2)。这提供了大量的生产净空高度。因为高性能的过程管理,当有不稳定的过程事件发生时,工厂有实质可选择的余地,并且当工程师让过程回到控制之后,可以持续生产合乎卫生的食品。
食品包装-使用统计过程控制分析学管理联合包装机的行为
一个乳制品厂的工艺设备使用统计过程控制分析学监控他们产品的填充重量。他们用统计过程控制图表监控填充重量的最大容许变化比例的关键行为指示(KPI)。这使他们最大化产量的同时,抑制违犯最大容许变化比例的风险。
使用联合包装机包装一些特殊乳制品产品。乳制品厂用统计过程控制评定他们的过程符合规格的可靠性和能力以使联合包装机合格。产品填充重量被例行采集,并且统计过程控制结果作为他们供应链质量管理的一部分报告给乳制品厂。
因为供货商和客户都使用相同的分析方法和图表,他们可以更有效的合作以改进过程和增加收益。除此之外,乳制品厂可以使用联合包装机的可交付品质严格要求地管理他们的标签。
全局设备效率(OEE)- 使用统计过程控制以使全局设备效率计算有效
统计过程控制和过程性能方法增加全局设备效率值的价值和可用性。使用统计过程控制方法分析全局设备效率数值并将其制作成图表,比一个简单的管理仪表板上的信号器提供更多的行动信息。不仅统计过程控制图表和趋势分析为全局设备效率关键行为指示提供更多更好的决策支持,你还可以钻取并研究单个的全局设备效率组件的行为,有效性,性能和收益。
关键行为指示组件分析
一个工厂使用大的连续的过程焦炉需要开发一个有意义且可靠的包括财政,操作,可测量的参数以精确地表示全部能量耗费的关键行为指示。现有的关键行为指示系统不能监控所有必要组件的变化和可视化的表示关键行为指示,以使操作能够反应在早期变化的信号上。
通过应用使用统计过程控制为驱动的MA,基于角色的可视化报告技术,工厂可以建立一个监控所有有助于能量消耗关键行为指示的组件参数的稳定性和变化的系统,这使得关键行为指示自身显示越界之前,可以对一个主要关键行为指示组件的变化做出检测。
结论
现代控制系统,工厂厂房数据采集,和实验室生成大量过程数据。除非数据被分析并且有效地报告给全部参与了生产和工厂管理的工作人员,否则它将不会对操作管理决策的制定起到作用。积极地使用统计过程控制和制造业分析学能够使这个数据为企业管理有效地利用。紧密耦合的分析学将使控制系统成为21世纪过程和企业管理系统的一个核心组件。
[/中文]
[外文]21st Century Process Management Using SPC Based Manufacturing Analytics
We constantly hear that we are in a new era where the problem is no longer one of not being able to collect enough data on our manufacturing processes, but how to effectively use the massive volume of data we collect. One of the primary battle cries of this new era is, “Better analysis of the data is absolutely essential!” (Then, following that battle cry is usually a summary of someone’s particular flavor of analysis technology or methodology.) Chances are, the term “KPI” is also mentioned, as it is explained how this technology contributes to the KPI process.
In keeping with that established path, let us examine an established technology in a new light.
Traditional Quality and Specification Analysis SPC
Statistical Process Control (SPC) is a time honored and well demonstrated method of process management. Everyone who has studied modern manufacturing knows of Dr. Deming and his early work establishing SPC as a standard practice in postwar Japan. SPC has long been used for measuring and monitoring quality by the Quality departments and labs of most industrial manufacturing facilities.
SPC has undergone periodic recasting and updates, such as Continuous Process Improvement (CPI) and Total Quality Management (TQM), as we learn more in each cycle of popularity. Certainly SPC is a key part of the Six Sigma and Lean Six Sigma process that many manufacturing companies have invested in.
Many of these SPC applications provide data analysis either “after the fact”, in the case of quality monitoring in a lab based on product samples at the end of a production process, or somewhat removed from the immediate “actionable decisions”, in the case of the Six Sigma processes.
Typical applications of the traditional SPC methods include successful quality/process management functions such as;
•
Routine SQC/SPC Reporting
•
Process Monitoring and Improvement
•
Analytical Method QC in laboratories
•
Regulatory Compliance
•
Supply Chain Customer Certification
Emerging Real-Time Operational Decision Support SPC Analysis
While the SPC analytical methods have met the challenge in these functions, it also is the right tool to meet the emerging battle cry for better analysis of available data. The time based, comparative analysis and visual presentation of the analysis that are unique to the SPC methods, enable manufacturers to better understand their processes and, more importantly to take immediate action based upon the analysis of the data. SPC is
the appropriate technology and methodology to meet the current needs of manufacturing that call for data analysis to provide:
•
Ability to measure ROI on systems investment
•
Timely and effective analysis summaries and reporting
•
Predictive capability
•
Identifiable benefits (lower costs, higher yields)
•
Support immediate actionable decision processes based upon results
•
High confidence to make process change for improvement
•
Incorporation into a proactive process improvement program
•
Make it easy to get specific, role-based results for individual informational needs
•
Reduce complex calculations of aggregated data to meaningful and measurable information with context
•
Immediately measurable results of actions taken
If the battle cry is for better analysis, then Real-Time, actionable, decision support analysis are the rally points. Again, the SPC techniques, methods and characteristics lend themselves to these points. The abilty to provide the manufacturing operator with real-time SPC derived analysis, presented visually in a simple enough format to quickly identify “out of control” parameters, is the goal. This real-time presentation can take the form of a robust SPC application embedded within an HMI display or it could be in the form of a dedicated SPC data collection and analysis system on the manufacturing floor.
In the HMI applications, SPC has often been viewed as an accessory function. However, the emerging requirement for better analysis, to facilitate better process management and improvement, requires that fully developed SPC analysis and presentation tools be an embedded function within the HMI. This facilitates the most effective real-time data capture and analysis, provides the operator with immediate access to the visual presentation of the analysis, allows a mechanism for the operator to document preventative or corrective actions taken, and supports the emerging role of the HMI as a feed into integrated enterprise analytics.
SPC is fundamental in this emerging need of better data analysis. It is how to make more effective use of all the data in historians, MES and ERP systems. It also allow for the reduction of complex, specialized process data into graphic visualizations which operations and management can quickly understand and from which informed action can be taken.
Some of the particular derived benefits of Real-Time Decision Support SPC Analytics are;
•
Robust, easy to understand, high level of confidence
•
Identify, verify and reduce sources of variation
•
Analyzes ongoing and immediate variation, not final product quality – process control not product control
•
Applies to both the process and the product
•
Detects changes, shifts and unusual events
•
Separates signals from noise
•
Identifies causes of excessive variation
•
Monitors real-time results of continuous process improvement activities (pitch for CPI as one of the real ROI payoffs for SPC)
•
Predictive problem detection on a stable process
•
Provides tool for documentation of compliance with customer supply chain requirements
SPC Based Manufacturing Analytics
Along with the battle cry for better analysis of data, there has been an increase in the use of Business Intelligence and Business Analysis tools, and the desire to integrate production data and analysis with business data and analysis, This is all aimed at developing a better understanding of complete corporate performance.
This new opportunity has created the need for analytical tools that can perform in this joint environment, and deliver to these goals. SPC methods arise well to this new opportunity in the form of SPC based Manufacturing Analytics.
SPC based Manufacturing Analytics is statistical and rule based, providing the aggregation, analysis and role-based visualization and reporting of manufacturing data that enables users to better understand and improve their processes, identify and reinforce best practices, react quickly to process events, and anticipate potential problems before they affect product quality, yield, or cost. The key differentiating elements of the SPC based Manufacturing Analytics methodology for analyzing data are;
•
Statistically based
•
Focused on role based analysis and reporting
•
Identifies significant events, separating out “noise”
•
Emphasis on visual presentation technique to enable quick analysis
•
Supports both reactive and predictive behavior
•
Easy enough to be implemented, maintained, and used by existing plant personnel
•
Aggregates data from different sources while preserving statistical validity
•
Supports the ISA S-95 Production Performance Analysis Activity Model, which outlines the need for robust systems, methodologies, and tools to improve the ability to make very informed decisions based upon extensive and varied analysis functions.
What has been the result of the merging of these two levels of business analysis and manufacturing analysis are value parameters used to monitor the overall status and performance of an operation or enterprise. These are expressed as Key Performance Indicators (KPIs), which are usually a single parameter consisting of an aggregation of of financial, operational, and measured parameters to provide a meaningful and reliable KPI variable. These variables are often monitored for a “good” or “bad” status in some sort of web-based visualization tool, such as a portal or dashboard. The ability to
contribute to or provision this web-based visualization function is a key component of the new analysis opportunity.
SPC based Manufacturing Analytics methodologies would allow for a system to be created that monitors the stability and change of all the parameter components contributing to the KPI, which would allow the detection of a change in one key KPI component before the KPI itself shows to be out of range. The visual presentation of this detection could then be displayed not as just a “good” (green) or “bad” (red) status, but even as a “potentially getting worse” (yellow) status. Existing KPI analysis and reporting systems do not have access to all the parameters or the ability to represent all the components in this fashion, such that operations and management can quickly identify, or even predict, early signs of detrimental change to take action against.
Choosing an SPC Analysis Platform
A good SPC platform should be able to accommodate all three levels of analytics outlined here for the manufacturing environment.
•
Traditional – SPC Quality and Specification Analysis
•
Emerging – Real-Time Operational Decision Support SPC Analysis
•
New – SPC Driven Manufacturing Analytics
Additionally, this platform must be architecturally structured to allow for growth and expansion around the different levels of need and phased implementation a manufacturer may require. This requires a modular and component based approach that allows easy integration into existing and disparate systems, with an open and standards based approach to accessing the data to be analyzed. A system that can provide this function of real-time SPC analytics, coupled with role specific, visual reporting, can support the full range of analytics and decision support for all levels from plant floor through management.
Case studies
The following examples illustrate how SPC-based Manufacturing Analytics is being implemented to get real benefits and improve the operating process.
Flexible Packaging – Offline and online real-time SPC analysis supports continuous improvement program
A large international flexible packaging manufacturer produces many different products of a similar form for various food industry customers. Product specifications are slightly different for each customer, and tight monitoring of the production against these specifications is necessary to meet supply chain requirements. Once the production line is running, the goal is keep the line running, and only make necessary changes to the equipment to ensure that the product remains within specs.
By using SPC analysis of electronically collected measurement samples on a real-time basis, and then presenting the results to the operator in a very basic “in or out” of specification/control visual display, the operators were able to detect and make appropriate adjustments to keep the production line running and achieve reduced variation and reduction in returned product.
The operators and quality inspectors have information directly available to them from this interface. They are able to view summary historic data, as well as real-time SPC charts to analyze the process. These role specific reports give the staff complete access to standard operating procedures and support uniform workflows.
SPC is used to establish process control limits and support quality control initiatives. Engineering uses the software for process studies and to work process capability into specifications. The approach is part of the continuous process improvement programs which reduce variation, decrease scrap, and reduce product overages.
As a result of proactively using SPC, the plant has been able to improve operations by limiting weight variation during the manufacturing process. Over the past two years, the difference between the target weights versus actual weight was measured and benchmarked. The measured results showed a cumulative cost savings of over $200,000 since the program started.
FOOD SAFETY – Use SPC Analytics for early prediction
A poultry processing plant monitors regulated bacteria such as E. coli and Salmonella. When the pathogen level exceeds the limit, the required action is costly and time consuming. Exceeding the safe levels can happen quickly, with little time for corrective action. Setting arbitrary “reaction” limits can lead to a tradeoff between false positives and missing signals.
The fact that they actively use SPC and actively manage the process enables fast and informed response. Applying Manufacturing Analytics using SPC-based event and pattern rule violation detection results in early detection of an unstable process, predicting the likelihood of an event, and enabling corrective action and process improvement to prevent reoccurrence.
By using continuous process improvement, the poultry plant was able to develop very high capability production (Cpk = 12.2) with regard to pathogen levels. This provides a
large amount of manufacturing head room. Because of this high capability process management, the plant has a substantial leeway when a process destabilizing event occurs, and can continue to produce wholesome food while the process engineers are returning the process to control.
Food Packaging – Using SPC Analytics to manage a co-packer’s performance.
A dairy processing facility uses SPC Analytics to monitor the fill weights of their products. They use SPC control charts monitor a KPI that monitors that the fill weight to Maximum Allowable Variation (MAV) ratio. This enables them to maximize the product yield while controlling the risk of MAV violation.
The dairy uses a co-packer to package some specialty products. The dairy qualifies the co-packer by using SPC to evaluate their process dependability and capability to meet specifications. Production fill weights are routinely collected and SPC results reported to the dairy as part of their supply chain quality management.
Since both the vendor and customer are using the same analytics and charts, they can more effectively collaborate to improve the process and yields. In addition, the dairy can use the co-packer’s quality deliverables to manage their label weight regulatory compliance.
OEE - Use of SPC to validate OEE calculations
SPC and process capability methods increase value and usability of OEE values. The OEE KPI can be treated like any other process parameter with trend and capability monitoring and analysis. Using SPC methods to analyze and chart the OEE values provides far more actionable information than a simple annunciator on the management dashboard. Not only does the SPC chart and trend analysis deliver more and better
decision support on the OEE KPI, you can drill down and study the behavior of the individual OEE components, availability, performance and yield.
KPI Component Analysis
A plant uses large, continuous process ovens needs to develop a meaningful and reliable KPI consisting of financial, operational, and measured parameters to accurately represent the total energy costs. The existing KPI system was unable to monitor all the necessary component variables and visually display the KPI, such that operations could react to early signs of change.
By applying MA using SPC-driven, role-based visualization reporting techniques, the plant could create a system that monitored the stability and change of all the parameter components contributing to the Energy Consumption KPI, which allowed the detection of a change in one key KPI component before the KPI itself showed to be out of range.
Conclusion
Modern control systems, plant floor data collection, and laboratories generate large volumes of process data. Unless this data is analyzed and usefully reported to all the staff involved in production and plant management, it will not be useful for operational management decision making. Actively using SPC and Manufacturing Analytics enables this data to be effectively used to manage the enterprise. Tightly coupled analytics will make control systems a core component of 21st century process and enterprise management systems.[/外文]
|