PSI - Issue 28

Evgeniia Georgievskaia et al. / Procedia Structural Integrity 28 (2020) 836–842 Evgeniia Georgievskaia/ Structural Integrity Procedia 00 (2020) 000–000

839

The ability to identify cracks at HT elements using vibration diagnostics was considered in paper Georgievskaia E. (2019). As shown in the example of the runner there is often no correlation between driving forces of crack development and vibration parameters measured by a diagnostic system. One of the key reasons is the existence of runner’s eigenvalues modes without shaft deformation. Figure 2 presents the comparison deformed shapes of the rotating part for two different eigenvalues forms. The location of the sensors marks the letter “S” in the circle. In Fig.2a all rotating components have displacements that allow associating the shaft deformation and stress-strain state of runner or generator. In this case, it is possible to connect the rate of the crack growth in runner or generator with vibration parameters. In Fig.2b contrariwise the sensors on the shaft do not detect any distortion although runner has bending deformation. Therefore, any vibration diagnostic systems do not show some problems at such modes and do not prevent appropriate damage. 4. Description of predictive analytics system You need a convenient and effective protective tool for evaluating the actual technical condition, predicting the ultimate state, and avoid the sudden damage of HPP's equipment and increase the reliability of the machine. The proposed solution is the predictive analytics system (PAS) for hydraulic turbines as an additional means to forecast the occurrence and growth of dangerous operational defects in terms of fatigue damage. PAS realizes the analytic approach to define the limit state by fatigue criteria for different combinations of operational modes. The forecasting technology of the proposed predictive analytics system presents in Fig.3. The predictive algorithm is suitable for all types and constructions of hydraulic turbines. At first, the digital image of the turbine is developed on the base of 3D scanning data and technical documentation. Then multidisciplinary calculations are performed to simulate the stress-strain state of the unit at different operating points. Results are summarized in the matrix of the individual data as an equipment’s response to external influences. The next step is variant calculations using the matrix of the individual data and various combinations of operation modes. Input data is information about the actual operating time at every working mode and expected regime parameters for the upcoming period. The output information is an individual lifetime (actual and residual) as a part of the total lifetime. For practice purposes, the residual lifetime can be represented in working hours according to the expected combination of operation modes. The main feature of the proposed solution is dividing the complex problem into two parts: a relatively simple unaltered on-line algorithm and the modified complicated individual module developed off-line. It is possible because the basic factors in lifetime estimation of specific HT are operational conditions. So, input data for the users of the predictive system is only information about the actual operating time at every operating mode and expected regime parameters for the upcoming period.

Fig. 2. Distribution of displacements for different eigenvalues modes of hydraulic turbines rotating part: (a) bending modes of the shaft with runner and generator; (b) symmetric bending modes of the runner without deformation of the shaft