Explore chapters and articles related to this topic
Shaft Design
Published in Wei Tong, Mechanical Design and Manufacturing of Electric Motors, 2022
Static or dynamic shaft runout measuring data can be captured using a dial indicator or an electronic digital indicator (Figure 3.17). The indicator with a magnetic base is attached to the motor (or a metallic platform) and positioned perpendicularly to the surface of interest. The readings are often expressed as total indicated runout (TIR), which is the difference between the two extreme measurements of the indicator in a full rotation of the shaft. For a runout measurement on a discontinuous surface, the measurement is usually taken at a number of positions on the outer shaft surface. Alternatively, runout measurements can be performed more efficiently by using a coordinate measuring machine (CMM) or non-contact sensors such as capacitive and eddy-current sensors, depending on runout measurement specifics and environmental conditions.
Seals for Fluid Power Equipment
Published in Anton H. Hehn, Fluid Power Troubleshooting, 1995
Only a properly trained person should attempt to repair the sealing surfaces of face seals. The condition of the real surfaces is so critical that one company provides 40 hours of training to its personnel on face seal operation, repair, and installation. With new replacement parts, do not touch the sealing surfaces with fingers or an old wiping rag. Make sure that the seal seat is perpendicular to the shaft within 0.001 in. TIR (TIR is the total change in indicator reading during one complete rotation of the shaft). Lubricate the sealing surfaces well with the fluid to be sealed before installation.
Design for Machining
Published in Helmi Youssef, Hassan El-Hofy, Non-Traditional and Advanced Machining Technologies, 2020
According to Bralla (1999), the range of tolerance recommended for broaching operation can be summarized as follows: Surface finish. Surface finish produced by broaching does not match the grinding finish. However, it is superior to the finish produced by most other machining methods. Burnished finishes can be guaranteed by employing good tool design and proper cooling oils for highly machinable materials.Flatness. Parts of uniform sections having sufficient strength to withstand cutting forces can be machined within ±0.025 mm total indicator reading (TIR).Parallelism. Parallelism of machined surfaces machined in the same stroke should be within ±0.025 mm TIR in good-to-fair machinability rated materials.Squareness. For parts that can be clamped and retained on true surfaces, a squareness of ±0.025 mm TIR is possible and tolerances of ±0.08 mm can be obtained under controlled conditions in good-machinability-rated materials.Concentricity. The concentricity error caused by broach drift should not exceed ±0.025 to ±0.05 mm for round or similarly shaped holes in good-to-fair-machinability-rated materials.Chamfers and radii. Tolerances on chamfers and radii should be as liberal as possible. Radii under 0.8 mm should have a minimum tolerance of ±0.13 mm; ±0.025 mm should be allowed for larger sizes. Large tolerances reduce broach manufacturing and maintenance cost.
A Bayesian hierarchical model for quantitative and qualitative responses
Published in Journal of Quality Technology, 2018
Lulu Kang, Xiaoning Kang, Xinwei Deng, Ran Jin
Many data collected from engineering and scientific systems contain both quantitative and qualitative (QQ) output observations or responses. For example, in the lapping stage of the wafer manufacturing process the qualitative response is the conformity of the site total indicator reading (STIR) of the wafer, which has two possible outcomes: whether or not the wafer STIR is within the tolerance. The quantitative response is the total thickness variation (TTV) of the wafer. Both of the response variables measure the smoothness of the wafers, which is an important geometrical quality index of the wafers. See Ning et al. (2012), Zhao et al. (2011), and Zhang et al. (2015) for detailed studies of these two quality characteristics. An interpretable and accurate statistical modeling approach is needed to find out how the controllable process variables and the covariates affect the two kinds of responses. Among all possible modeling techniques, the simplest approach is to model the two types of responses separately. We can use linear regression models for the quantitative response and generalized linear models or classification methods for the qualitative response. But doing so would ignore the possible association between the two responses. Deng and Jin (2015) have shown the necessity for jointly modeling the two kinds of responses for the lapping process experiment. Such association is important as it provides us with some insightful understandings of the system under study as well as a significant improvement in the prediction accuracy.
An adaptive thresholding-based process variability monitoring
Published in Journal of Quality Technology, 2019
Galal M. Abdella, Jinho Kim, Sangahn Kim, Khalifa N. Al-Khalifa, Myong K. (MK) Jeong, Abdel Magid Hamouda, Elsayed A. Elsayed
In this section, we illustrate the implementation procedures of the ALT-norm chart in the lapping process data from Li, Wang, and Yeh (2013). In wafer semiconductor manufacturing, the lapping process is used to remove saw marks and to eliminate thickness variation. However, the lapping process is described by five dimensions: (1) total thickness variation (TTV), (2) total indicator reading (TIR), (3) site total indicator reading (STIR), (4) bow (B), and (5) warp (W). See Li, Wang, and Yeh (2013) for a further detailed description of these measures.
A multisource domain adaptation method for quality prediction in small-batch production systems
Published in International Journal of Production Research, 2022
With the available data, there is historical information from two single crystals of silicon, which are denoted as Domain S-1 and Domain S-2 and contain 389 and 379 wafers, respectively. There are also 100 observations for the single crystal of silicon in process, which is denoted as Domain T. In this study, 6 important quality indices that are evaluated after the final stage are considered to be the output variables, most of which are defined by Semiconductor Equipment and Materials International as industrial standards; these indices include Centre Thickness (CTRTHK), Total Thick Variation (TTV), Total Indicator Reading (TIR), Site Total Indicator Reading (STIR), Bow and Warp. There are missing data and omitted quality inspections, and therefore, the corresponding number of input variables inspected in the first four stages of the domains are different, which are listed in Table 4. The quality variables inspected in the first four stages contain certain geometric and physical properties, such as taper and resistivity. Note that the different circumstances of the available input variables correspond to the heterogeneous assumption that the domains have different input feature spaces. In the conclusion of the numerical experiments, the advantages of the proposed method should be apparent in situations with fewer training samples in the target domain. Therefore, to evaluate the proposed model, the chosen range of the training proportion of Domain T is from 0.03–0.15 with a step length equal to 0.01, and all samples from Domain S-1 and Domain S-2 are used as the training samples of the source domains. The results are shown in Figure 8 and Tables 5 and 6, where the hyperparameters are chosen as follows: the bandwidth parameter in Equation (14) is .