Narayanaswamy Balakrishnan - Accelerated Life Testing of One-shot Devices

Source: Zelen (1959).

Table 1.3 Failure records on solder joints under CSALTs with temperature (K) and a dichotomous variable indicating if the PCB type is “copper‐nickel‐tin (CNT)” or not.

Source: Lau et al. (1988).

Zheng et al. (2018) studied grease‐based magnetorheological fluids under SSALTs with four levels of temperature and observed whether their viscosities or shear stresses decreased by more than 10% after tests. Twenty samples of grease‐based magnetorheological fluids were subject to higher‐than‐normal operating temperature. Then, each sample was inspected only once and only whether it had failed or not at the inspection time was observed, and not the actual failure time. The data collected in this manner, presented in Table 1.4, were then used to estimate the mean lifetime of grease‐based magnetorheological fluids under the normal operating temperature.

It is important to point out that one‐shot device testing data arise from diverse fields beyond reliability engineering, such as in mice tumor studies from tumorigenicity experiments; see Kodell and Nelson (1980). In such a study, each mouse received a particular dosage of benzidine dihydrochloride in its drinking water and was later sacrificed to detect whether some tumors had developed by then or not. Tumor presence can be detected only at the time of mouse's sacrifice or natural death. These data are summarized in Table 1.5. The data collected in this form were then used to measure the impact of the chemical dosage on the risk of tumor development.

Lindsey and Ryan (1993) described experimental results conducted by National Center for Toxicological Research in 1974. 3355 out of 24 000 female mice were randomized to a control group or groups that were injected with a high dose (150 ppm) of a known carcinogen, called 2‐AAF, to different parts of the bodies. The inspection times on the mice were 12, 18, and 33 months and the outcomes of mice were death without tumor (DNT) and death with tumor (DWT), and sacrificed without tumor (SNT) and sacrificed with tumor (SWT). Balakrishnan et al. (2016a), in their analysis, ignored the information about parts of mouse bodies where the drugs were injected and combined SNT and SWT into one sacrificed group, and denoted the cause of DNT as natural death and the cause of DWT as death due to cancer. These data are summarized in Table 1.6. They then estimated the chance of death without tumor.

Ling et al. (2020) were primarily concerned with the data (Berlin et al., 1979), presented in Table 1.7, on the presence or absence of two disease categories – (a) thymic lymphoma and/or glomerulosclerosis and (b) all other diseases – for an irradiated group of 343 female mice given ‐radiation and a control group of 361 radiation‐free female mice to study the onset time and the rate of development of radiation‐induced disease. All of the mice in both groups were sacrificed at various times, with the presence of a disease indicating that the disease onset occurred before sacrifice, while the absence of a disease indicating that the disease onset would occur after sacrifice.

Table 1.4 Failure records on grease‐based magnetorheological fluids under SSALTs with temperature (K).

Source: Zheng et al. (2018).

Table 1.5 The number of mice sacrificed, with tumor from tumorigenecity experiments data.

Source: Kodell and Nelson (1980).

Table 1.6 The number of mice sacrificed, died without tumor, and died with tumor from the ED01 experiment data.

Source: Lindsey and Ryan (1993).

Table 1.7 Serial sacrifice data on the presence or absence of two disease categories: (a) thymic lymphoma and/or glomerulosclerosis and (b) all other diseases.