【作者】 陆爱武；

【导师】 司马锡生；

【作者基本信息】 南京工业大学 ， 企业管理， 2005， 硕士

【摘要】 控制图的一个基本假设前提是:从过程获得的观测值彼此独立。在质量要求严格的连续型制药行业中,随着设备及仪器自动化和智能化水平的提高,生产过程中的数据自动采集,采样间隔缩短,一些过程数据存在明显自相关,不能满足常规控制理论中观测值相互独立的假定,暴露出基于传统的统计过程控制理论的控制图在处理自相关数据时会出现大量的虚发警报,严重影响了控制图的使用效果,不能有效地控制和改进过程质量。过程数据自相关现象引起了国际、国内学术界的重视,并取得了有价值的成果。如,孙静的“自相关过程的采样间隔与控制效果’、“自相关过程的统计控制状态”;Zhang,N,F.的"Statistical Control for Autocorrelated data","estimating process capability indexes for autocorrelated data"等文对自相关过程控制及能力指数进行了研究,但有些是在假定时间序列模型已知的前提下进行的统计分析,有些数据是仿真生成的。对于生产现场来说,多数情况时间序列模型是未知的,需要根据采集的实际数据来获得。基于上述理论和不足,本文根据制药行业过程自身特点,通过收集、考察药品生产过程中的一系列质量指标的特性,确定以注射用水生产过程在线采集的数据一电导率为研究对象,展开自相关控制图及工序能力的研究。借助SPSS(社会科学统计包)软件为工具,采用时域分析方法,对不同采样间隔的电导率自相关性进行了分析,表明当采样间隔τ<14 min时,电导率具有自相关性,且随着采样间隔的缩短,电导率自相关性越强;基于Pandit-Wu系统建模方法,证明了不同采样间隔的电导率自相关过程模型均为AR(1);在建模基础上,通过自相关过程常规控制图与残差控制图的比较,说明了常规控制图错误预警的缺陷,论证了采用残差控制图进行自相关过程的控制可避免的质量损失;在系统无自相关影响且稳定受控的前提下,为了简化过程控制的方法,提出了重新采用常规控制图进行控制的观点,确定了最佳采样间隔τ≥14min。通过注射用水的工序能力的指数的计算及分析,表明注射用水生产过程工序能力富余,根据质量损失函数理论评估其经济性,提出了放宽公差范围、简化注射用水处理流程等措施来降低注射用水过程能力指数(CP_u)的观点。更多还原

【Abstract】 The traditional control chart is based upon the assumption that the observation data which came from the process is independent mutually. In the consecutive process of chemical and pharmaceutical industry, the requirement of quality is very strict, the process data is collected automatically with the improvement of equipment and instrument automatization and intelligentize, some process data exist obvious autocorrelated due to the sampling period is reduced , these autocorrelated data can not comply with the suppose that every data is independent which based on the traditional statistic theory. Thus fact proved that the traditional control chart was not suitable for autocorrelated data control and brought a mass of error alarm during the process control, thus badly influence the impact of control chart, resulted in losing to effectively control process and improve the quality. Autocorrelation has caused concern from internal and external academe and gained valuable achievement, for example, Sun Jing’s article of State in Control for Autocorrelated Process, also Sampling Interval in Autocorrelated Process and Control Effect, Nien Fan Zhang’s article of A Statistical Control Chart for Stationary Process Data as well as Estimating Process Capability Indexes for Autocorrelated Data etc.. in these articles, they discuss the autocorrelated process control method and process capability index calculation, but some of the statistic analysis in these articles rely on the assumption which the time series model has been known, and also some of the data is emulational. In fact, the time series model generally is unknown; it needs to modeling through collecting the actual data from the production locale. Based on the foregoing theory and its insufficiency, on account of the feature of pharmaceutical industry, by collecting and observation a series of quality indexes characteristic in the process of pharma production, confirmed the process data "WFI conductance " as the investigated object to deploy the research about the autocorrelated data control chart and process capability, with the tool of SPSS software to analysis the autocorrelation at different sampling time by means of Time Zone Analytical Method, and show that when sampling time is less than 14 minutes, the autocorrelation is existed, and the characteristic of autocorrelation is enhanced gradually along with the sampling time shorted, testify the models are AR(1) at different sampling time based on Pandit-Wu Method, and conclude that the method of residual control chart can be used to judge the autocorrelated process avoiding some cost loss by comparison with the traditional control chart which will bring an error alarm. Under the system is controlled stably without autocorrelation influence, the best sampling time is more than 14 minutes so as to simplify the control method, put forward the opinion of using traditional control chart for the process control again in case of troublesome brought by the method of residual control chart. By calculating and analyzing the process capability index, indicate that the process capability is superabundant and evaluate its economic depending on the Quality Loss Function, put forward the opinion of lowing the process capability process via ways of widen the process tolerate range and simplify the working procedure etc.更多还原