检测方法偏离怎么申请与记录

结论：偏离不是禁止的，但必须评估、批准、记录、说明

检测中偶尔会遇到无法完全按标准方法执行的情形：

样品尺寸不满足方法要求。 方法要求的试样尺寸，实际样品取不出来。

设备条件受限。 设备的量程、精度或功能不完全匹配。

样品特性限制。 样品的形状、材质使某个步骤无法执行。

客户的特殊要求。 委托方要求在非标准条件下测试。

方法本身不明确。 方法条款有歧义或未覆盖的情形。

处理这些情形的原则是：偏离可以有，但必须是受控的。

受控意味着四件事：评估了影响、得到了批准、有完整记录、在报告中说明。

偏离的类型

「设备替代」需要验证等效性，不能只说「功能相近」，要有数据说明替代设备的结果与规定设备可比。

评估的要点

批准偏离前要评估：

偏离对结果的影响方向和大小。 会使结果偏高还是偏低、偏多少。

是否影响判定结论。 如果偏离使结果更严苛，不合格的结论仍然成立；如果使结果更宽松，合格的结论就存疑。

是否有技术依据。 偏离后的方法是否仍然科学合理。

是否有替代方案。 有没有不需要偏离的其他方式。

委托方是否知情并同意。 影响结论适用性的偏离要告知。

「偏离方向」的判断很实用 ——如果偏离使试验更严苛而产品仍然通过，结论的可靠性反而更强；反之则需要谨慎。

申请与批准的流程

第一步，提出。 由发现问题的检测人员提出。

第二步，技术评估。 由有相应技术能力的人评估影响。

第三步，与委托方沟通。 说明偏离内容和可能的影响。

第四步，批准。 由授权人批准。

第五步，实施并记录。 按批准的方式执行，完整记录。

第六步，报告说明。 在报告中明确说明偏离。

「与委托方沟通」不能省略 ——委托方有权知道检测是在什么条件下做的，因为这影响报告的使用。

记录的内容

偏离的具体内容。 哪一条、怎么偏离的。

偏离的原因。 为什么必须偏离。

技术评估。 对结果影响的分析。

批准情况。 谁批准的、什么时候。

委托方的确认。 沟通记录和同意的证据。

实际执行的条件。 偏离后的实际条件参数。

记录要能支撑复现 ——别人看了记录应当能知道实际是怎么做的。

报告中的表述

偏离要在报告中明确说明，表述要点：

说明依据的方法和偏离的条款。 明确到条款号。

说明实际采用的条件。 具体的参数。

说明偏离的原因。 简要说明。

必要时说明影响。 对结果解释的影响。

避免含糊表述。 「参照某标准」这类表述不够明确，应当说清楚哪些按标准、哪些不按。

「参照」这个词在报告中容易产生歧义 ——它到底是完全按照还是部分按照？建议明确写出偏离的具体内容。

常见的处理不当

偏离了但没记录。 事后无法追溯。

记录了但没评估影响。 不知道数据是否可用。

没告知委托方。 委托方按标准方法的结果理解数据。

报告中不说明。 报告使用者以为是完全符合标准的结果。

把偏离常态化。 某个偏离反复使用而不重新评估。

「把偏离常态化」值得注意 ——如果某个偏离经常发生，说明方法或条件需要系统性地解决，而不是每次都走偏离流程。

常见的准备不足

信息不全。 材料、工艺、使用环境的信息缺失，导致方案设计缺乏依据。

样品不具代表性。 试制样品或挑选样品，与量产状态有差距。

条件未明确。 测试条件没说清楚，数据的适用范围就不明确。

只看初始值。 忽略了环境和时间作用后的性能变化。

项目不完整。 遗漏了必要的项目，提交时才发现。

这五项在实际委托中出现频率较高，建议在项目启动时逐项确认，因为它们造成的返工成本远高于事前确认所花的时间。

数据的组织与保存

记录的完整性。 条件、方法、设备、样品状态、实测值。

可追溯性。 数据与样品批次、生产批次的关联。

保存期限。 覆盖产品的生命周期。

检索的便利。 需要时能快速定位。

版本的管理。 数据更新时的版本控制。

建议建立统一的数据台账，记录每组数据的来源、条件、适用范围和有效期。数据积累起来之后，后续的产品开发和变更评估都能受益。

与整体工作的衔接

这部分工作不是孤立的，它与几项内容相关：

与设计的衔接。 数据是设计决策的依据，也是设计验证的证据。

与风险管理的衔接。 识别出的风险点应当有对应的验证。

与变更管理的衔接。 变更时判断是否需要重新验证。

与生产控制的衔接。 关键参数的控制限值来自验证数据。

与售后的衔接。 失效分析需要原始数据作为对照。

建议在项目规划时就把这些衔接关系理清楚，避免各环节各做各的、数据无法互相支撑。

我们的做法

遇到需要偏离的情形，我们先与委托方沟通说明，因为偏离影响数据的解释和报告的使用。

评估时我们重点判断偏离的方向 ——如果偏离使试验更严苛，我们会在报告中说明这一点，因为这有助于报告使用者理解结论的可靠性。

报告中我们明确写出偏离的条款和实际条件，不用「参照」这类含糊表述，让报告的使用者清楚知道数据是在什么条件下得到的。

如果你的样品可能无法完全满足标准方法的要求，可以提前发过来一起评估，或者直接联系：132 4819 8029。送检要求见送检要求，流程见检测流程，其他问题见常见问题。

English version

Conclusion. Testing occasionally encounters situations where a standard method cannot be followed exactly. Specimen size may not meet the method's requirement because the actual sample cannot yield it. Equipment may be limited in range, accuracy or function. Sample characteristics of shape or material may make a step impracticable. A client may require testing under non-standard conditions. Or the method itself may be ambiguous or silent on the situation. The principle for handling these is that deviation is permissible but must be controlled, and control means four things: the impact is assessed, the deviation is approved, records are complete, and the report states it.

Types of deviation. Specimen size, where standard-size specimens cannot be taken, may make results not directly comparable. Specimen quantity, where sample material is insufficient, reduces confidence in statistical results. Test conditions differing in temperature or speed may bias results systematically. Test steps omitted or adjusted have an effect depending on the step's importance. Equipment substitution with functionally similar instruments requires equivalence verification. And acceptance criteria adopted at the client's direction limit the scope within which the conclusion applies. Equipment substitution requires verified equivalence: it is not enough to say the equipment is functionally similar, and data should show that results from the substitute are comparable with those from the specified instrument.

Points for assessment. Before approving a deviation, assess its direction and magnitude of effect on the result, meaning whether it biases results high or low and by how much. Assess whether it affects the conclusion: where the deviation makes the test more demanding, a failing conclusion still stands, while where it makes the test less demanding, a passing conclusion is open to question. Assess whether there is a technical basis, meaning whether the deviated method remains scientifically sound. Assess whether alternatives exist that avoid deviation. And assess whether the client is informed and agrees, since deviations affecting the applicability of conclusions must be disclosed. Judging direction is practical: where a deviation makes the test more demanding and the product still passes, the conclusion is if anything more robust, while the converse warrants caution.

Application and approval. The testing officer who encounters the problem raises it. Someone with the relevant technical competence assesses the impact. The client is consulted, with the deviation and its possible effects explained. An authorised person approves. The work proceeds as approved and is recorded fully. And the report states the deviation explicitly. Consulting the client cannot be omitted: clients are entitled to know the conditions under which testing was performed, since that affects how the report is used.

What to record. The specific deviation, identifying the clause and the departure. The reason it was necessary. The technical assessment analysing the effect on results. The approval, covering who approved and when. The client's confirmation, covering the consultation record and evidence of agreement. And the actual conditions applied, with specific parameters. Records should support reproduction: someone reading them should be able to establish what was actually done.

Stating it in the report. Deviations must be stated explicitly. Identify the method followed and the clause deviated from, down to the clause number. State the conditions actually used, with specific parameters. State the reason briefly. State the effect where relevant to interpreting the result. And avoid vague wording, since phrases such as "with reference to" a standard are insufficiently definite, and it should be clear which parts followed the standard and which did not. "With reference to" is prone to ambiguity in reports, since it leaves unclear whether the standard was followed fully or partly, and the specific deviation should be written out.

Common mishandling. Deviating without recording it leaves no traceability afterwards. Recording without assessing the impact leaves the usability of the data unknown. Not informing the client leaves them interpreting the data as standard-method results. Not stating it in the report leaves users believing the results conform fully. And normalising the deviation means repeating it without re-assessment. Normalisation deserves attention: a deviation that occurs frequently indicates that the method or conditions need addressing systematically rather than routing through the deviation process each time.

How we handle it. Where a deviation is needed we consult the client first, since it affects interpretation of the data and use of the report. In assessing, we concentrate on the direction of the deviation, and where it makes the test more demanding we say so in the report, since that helps report users judge the robustness of the conclusion. In the report we write out the clause deviated from and the actual conditions rather than using vague phrases, so that users know exactly the conditions under which the data were obtained.

If your sample may not fully meet the standard method's requirements, send it over in advance and we will assess it together. Phone or WeChat: +86 132 4819 8029.