通过追溯可靠性实验的发展起源，指出了现有以故障时间概率分布为核心的可靠性研究范式的历史局限性，即仅适用于有故障观测数据的情况，而对创新产品的研发束手无策。具体表现在，可靠性指标具有小样本、动态性、属人性特征，而质量特征却是大样本的、静态的、客观的，因此传统可靠性统计试验错误的使用了质量管理中概率抽样验收方法；而加速寿命试验和加速退化试验，由于采用基于微观失效物理构建的经验性加速模型使得试验方法并不适用于真正具有功能的系统级产品。为此，从可靠性科学原理的角度分析了以验证因果规律为目标的可靠性科学实验的存在性，进而定义了可靠性实验为旨在验证系统裕量与性能和性能要求之间的机会因果规律的受控实验，并分析了包括规律清晰性、黑箱认识论和机会因果律在内的可靠性实验的内涵，在此基础上给出了可靠性实验方法的基本原则：系统综合、分类判别和优化均衡。进一步的，基于可靠性实验原理，构建了以模型为核心的可靠性时空验证与测试体系，展示了该体系才是唯一能够从可靠性特征内涵出发，同时满足工程需求的可靠性验证与测试方法。最后，探讨了可靠性科学实验的使用范围，提出了可靠性时空验证与测试体系的发展设想。

This article reviews the development history of reliability testing， and points out the historical limitations of the existing reliability research paradigm， which is centered on the probability distribution of failure time. This paradigm is only applicable in situations with failure observation data and proves ineffective for the development of innovative products. Specifically， reliability indicators exhibit characteristics of small samples， dynamic behavior， and subjectivity， while quality characteristics are large-sample， static， and objective. Consequently， traditional reliability statistical testing has erroneously adopted the probability sampling acceptance methods for quality management. Moreover， accelerated life testing and accelerated degradation testing， due to their reliance on empirical acceleration models based on failure physics in micro dimensions， are not suitable for true system-level products with function requirements. This paper analyzes the necessity of reliability science experiments aimed at validating causal laws from the perspective of reliability science principles. Reliability experiments are then defined as controlled experiments intended for verification of opportunity causation laws between system margins and performance and performance requirements. The connotations of reliability experiments， including the clarity of laws， black-box epistemology， and opportunity causation， are explored， and fundamental principles for reliability experimental methods—system integration， classification judgement， and optimization equilibrium—are proposed. Furthermore， a model-centered spatiotemporal reliability verification and testing system is established based on the principles of reliability experiments， demonstrating that this system is the only one that is derived based on the connotation of reliability characteristic and meets engineering demands. Finally， the scope of application of reliability scientific experiments is discussed， and developmental ideas for the spatiotemporal reliability verification and testing system are proposed.