统计技术在第3方检验与测试领域的应用

Application of Statistical Methods to Third-party Testing and Inspection

选用国家统计技术标准，其绝大多数也是等同采用ISO技术规范，在第3方检验/测试领域的应用至少可以达到以下2个目标:优化检验/测试方案，以降低费用；和对前者结果进行统计的规范化的解释，从而获得能被公认的具有公信力的检验/测试结果。

To choose the national technical standards of statistical methods, these are identical or modified to ISO standards, the application of these statistical methods in third-party testing and inspection has two purposes at least, that is to optimality plan of the testing and inspection，so cost downing, and statistical interpretation of results from testing and inspection. So the results are of confirmation and credibility.

众所周知，样本的均值易受样本中个别极值影响，工程上更关注的是在置信度下的均值的置信下限。如图1所示，图中从上到下有3条直线，分别代表α=0.999,0.500,0.001下的疲劳循环次数与所处应力水平的关系，工程上关注的α=0.001时的耐应力水平，如在10⁵循环次数下，B点(R=0.999)所处的应力水平，而不是A点(R=0.500,即应力分布的均值)的所处应力水平。

It is common knowledge that the sample mean was influenced by max or min, so the lower confidence limit of reliability is of significance in engineering. For example, the relation with fatigue life and stress level for fatigues testing of metallic materials, see Fig.1. There are three lines with α=(0.001,0.500,0.999) respectively in Fig.1. And B point at 10⁵ with R=0.999 was more important that A point with sample mean, that was R=0.500.
图1  某钢材的疲劳寿命分布

朱先生具有30多年的技术工作经验，尤其是16年的不同行业的审核工作经历，在统计技术的应用方面积累了丰富的经验。图2是朱先生在IE专业领域的目录，图3是数据的统计解释和分析领域的子目录，图4是统计抽样检验领域的子目录。

Mr. ZHU is of technical experience over 30 years, in especial audit for different organizations since 2000, and of rich experience of SPC.Fig.2 is contents of IE scope with Mr. ZHU, Fig. 3 is sub-contents of statistical interpretation of data, and Fig.4 is sub-contents of statistical sampling procedures for inspection.

因此，统计技术方法在第3方测试与检验领域的具体应用被推出，服务方式可以是上门式技术培训，也可以针对具体的项目出具具体的统计技术实施方案。

The methods of using statistical technical to third-party testing and inspection are recommended, the ways of application maybe training in the client, and or the planning for the special item.

1  数据的统计处理和解释/statistical interpretation of data

1.1  不同分布函数的参数及参数区间的估计:包括正态分布、指数分布、极值分布等；

1.1  estimation of parameters and its interval of normal distribution, exponential distribution and Ⅰextreme value distribution, etc.

1.2  不同分布函数的离群值判断:包括正态分布、指数分布、极值分布等；

1.2  detection and treatment of outliers in the sample form normal distribution, exponential distribution and Ⅰextreme value distribution, etc.

1.3  不同分布函数的假设检验:包括正态分布、指数分布、极值分布等；

1.3  hypothesis test normal distribution, exponential distribution and Ⅰextreme value distribution, and normality tests, comparison of two means in the case of paired observations, etc.

1.4  统计容忍区间的确定/determination of statistical tolerance intervals.

1.5  本章节的应用案例/example of case

【例-1】某一的特性值在统一检验条件下测得x_i:(5.4362,5.1367,5.2175)。这组数据是否满足δ_max=0.05, α=5%的要求.

说明:检测过程是一个不增值的过程，但又是一个必不可少的过程，因此希望样本量越少越好。至少3个样本量，能否代表总体。由于存在各种误差，包括样本本身的波动性和测试技术的波动性，测试结果是否存在离群值，在无显著性离群值的情况下相对误差是否在允许的δ_max=0.05之内。

【例-2】购置一批设备的寿命呈指数分布，该设备的说明书上保证的平均无故障间隔时间,MTFB为3500h。为验收该设备质量，用随机抽样的方法抽取5台设备进行寿命试验。问在不发生一次故障的条件下，试验最少应达到多少小时，这批设备才算合格.

2  统计抽样检验方案/statistical sampling procedures for inspection

2.1  计数型抽样检验方案:包括按AQL检索的逐批检验、孤立批的抽样检验（按极限质量和不合格品百分数抽样检验方案）、序贯抽样检验方案、声称质量水平的评定方案等；

2.1  sampling procedures for inspection by attributes indexed by acceptance quality limint,AQL, for lot-by-lot inspection, limiting quality,LQ for isolated lot inspection, assessment of declared quality levels,DQL, sequential sampling plans by AQL, etc.

2.2  计量型抽样检验方案:包括按AQL检索的逐批检验、标准型一次抽样检验、序贯抽样检验、声称质量水平的评定方案、具有先验质量信息的监督抽样检验等；

2.2  sampling procedures for inspection by variables indexed by acceptance quality limint,AQL, for lot-by-lot inspection, limiting quality,LQ for isolated lot inspection, assessment of declared quality levels,DQL, sequential sampling plans by AQL, commodity applying to the case with prior information of quality, etc.

2.3  试验方案的最小样本量的确定/choice of sample size for estimating the average quality of a lot or process

2.4  本章节案例/example of case

【例-3】某轴承轴径:(100±0.1)mm，σ=0.20mm，IL= Ⅶ，试确定核查抽样方案并作判定。第3方检测，很少事先知悉σ值的。如果标准差,σ未知，又如何确定核查抽样方案。

【例-4】某钢铸件的最低屈服点规定为400N/mm²，交验批的批量为500件。采用一般检验水平,IL=Ⅱ，AQL=1.5%的正常检验，已知σ=21N/mm².从计量型抽样标准中查得样本量字码为H，由样本量字码和AQL值，可继续在计量型抽样标准中查得所需样本量为12，接收常数k=1.613。随即随机抽取12个样本产品，测得屈服点为:(431,417,469,407,452,427,411,429,420,400,445,476) N/mm².请给出案例的接收准则并确定批的接收性。

说明:本案例采用计量型抽样检验方案，故判定结果为批不合格，其原因是正态分布的均值发生了明显的偏移。而选用计数型抽样检验方案，则样本测试结果均满足规范要求。故计量型抽样检验和计数型抽样检验，其结果可能存在不同判定，而计量型具有更高的可信度。另，相同条件下，计数型抽样检验方案的样本量远大于计量型的，如例-4，计数型一次正常抽样检验方案为[N=500,IL=Ⅱ,AQL=1.5,n=50,Ac=2,R=3]。

3  老化试验、加速老化试验、可靠性试验等试验方案的策划及其试验结果的分析

3  testing design and testing results analysis of ageing testing, accelerates ageing testing, reliability testing, stress step testing, etc.

3.1  硫化或热塑性橡胶应用阿累尼乌斯图推算寿命/rubber vulcanized or thermoplastic:estimation of life-time of use from an Arrhenius plot；

3.2  疲劳试验数据的统计处理和解释/statistical interpretation of data for fatigue testing；

3.3  腐蚀试验数据的统计处理和解释/standard practice for applying statistics analysis of corrosion data;

3.4  电子元器件加速寿命试验/accelerated life test of electronic elements and components

3.5  本章节案例/example of case

【例-5】为了考察某款微型电机在50℃贮存条件下的各种可靠性特征，确定使用对数正态分布情形下的恒定应力加速寿命试验。根据以往的试验，确定用温度T作为加速应力，其加速应力水平分别为(190, 220,240,260)℃。如何安排试验及试验结果如何解读，从而得出50℃下的特性值。

附:朱立民个人简历/Annex:Resume
受教育信息:1975-1979，上海市第9中学；1979-1983: 东北大学（原校名:东北工学院），专业: 冶金过程物理化学；1995.10-1996.10，连铸连轧高级研修班，中国金属学会主办，授课地点:北京科技大学。

工作经历:1983.08-2000.05，上海浦东钢铁集团有限公司技术处（原名:上海第三钢铁公司），从事新工艺、新材料的开发与管理，高级工程师，中国金属学会会员，兼任《钢铁研究》杂志编委；2000.05-2001.10，上海海联润滑材料研究所质管部，任质量部经理；2001.11-2007.02，摩迪英联国际认证有限公司,Moody，主任审核员；2007.05-2009.04，法国国际检验局,BV，主任审核员；2010.01-2015.09，杭州汉德质量认证服务有限公司,TUV-NORD。主任审核员。
个人已经积累拥有7000个以上国家技术标准。

兴趣爱好与擅长:质量管理及其经济性；制造系统工程（MSE）和流程工程学；可信性工程（可靠性、维修性、维修保障性、可用性）；企业节能减排的方向及工业生态；数据的统计处理和解释，等等。

Education:1975.02-1979.07, No.9 Middle School, Shanghai; 1979.08-1983.07: Northeast University, Shenyang, Liaoning, Major: Physical Chemistry of Process Metallurgy; 1995.10-1996.10, The Chinese Society for Metal (through University of Beijing Tech. & Sci.), Major: Continuous Casting-rolling of Iron & Steel.

Work experience:2010.01-2015.09, TUV-NORD, lead auditor; 2007.05-2009.04,BV, lead auditor; 2001.11-2007.02, Moody, lead auditor; 2000.05-2001.10, Hiri Lubricants R&D Centre, Quality assurance manager ;1983.08-2000.05, Shanghai No.3 Iron & Steel Co., Ltd., senior engineer.
person has more than 7000 national technical standards.

Research interests: Manufacturing system engineering,MSE; Industrial engineering,IE; Dependability (reliability, maintainability, supportability); Techno-economics of quality management; statistical interpretation of data ; energy-saving and emission-reduction of industry enterprises and industrial ecology, and Quality Ecology.