Once the key parameters discussed in section 4.1 are chosen, the remaining variables necessary to collecting data can be calculated by the methods covered in this section.
4.2.1 Current Methods
The current method used to store data from the National Instruments DAQ is con- tinuous 100% data retention. The DAQ is configured to a certain sampling frequency and run time, then records and saves data for that entire span. The starting of the data acquisition is based off a command from the user interface app. While the DAQ is trying to sample data, Matlab is breaking that data up into pre-selected win- dow sizes of data and then performing calculations on every window that has been recorded. However, no data is being removed or consolidated to just measurements. This results in a data file that contains the time waveform of the entire signal as well as the spectral components of every window for the entire signal. This utilizes a large amount of data storage and computing power.
4.2.2 Batch Processing and Triggers
As opposed to continuously collecting data over a long time period and then trying to cut it up in manageable pieces for data processing, a more efficient method is batch processing. Batch processing takes short collections of data and then processes that data. The length of each data collection depends on the inputs described in the previous section 4.1. So depending on the frequency resolution the time span for data collection could be long or short. How often data is collected depends on triggers, which are just a set of conditions. When the conditions are met a data collection begins. Common triggers are time or rotor speed. For instance triggers could be defined as every 5 seconds and a change of 25 rpm. The DAQ will begin taking data when the DAQ is started however it is only recording data into a buffer to evaluate the condition of the triggers. If 5 seconds hasn’t past or if a 25rpm change hasn’t occurred then the data in the buffer is deleted. If the conditions of the trigger have been met then a data collection will begin and the data will be saved and processed. This method has the advantage of reducing the total amount of data trying to be transmitted from the DAQ to the computer and can be configured to ignore certain states such as before the rotor begins to ramp. An example showcasing data taken with batch processing at a collection rate of every 2 seconds can be seen in the cascade plot in figure 4.43.
4.2.3 Vector and Waveform Data
One method commonly used in industry to maximize data storage efficiency is to dictate the level of data storage rates for vector and waveform data separately. This can be done by defining different triggers for vector and waveform data. Vector data can be a small set of complex values which represent either an amplitude and/ or
phase. Examples of vector data would be overall amplitude, 1X amplitude and phase, 0.5x amplitude and phase, 2X amplitude and phase etc. This data can be stored more at a higher frequency because the entire spectral components of the window and the time waveform do not need to be saved. Vector data can be presented in plots such as bode, polar, trend, and shaft centerline plots. Waveform data consists of an entire time waveform collection and the spectral components calculated by the fft. The waveform data can be represented by half and full spectrum, cascade, waterfall, orbit, and time waveform plots. This data requires much more storage and typically is stored less frequently.
The application of this method allows the user to customize storage rates depending on the intended use of the data. For example in the ME318 lab student perform a single plane balance on a RK4 rotor kit. This requires the use of the 1X amplitude and phase in the form of a bode or polar plot. In this case only vector data is being used; therefore, the collection of waveform data could be eliminated entirely to optimize data storage use.
4.2.4 State Based Acquisition
An alternative means of data collection to continuously sampling data is to define sampling rates based of the state of the machine. Typical states are Stopped, Slow Roll, Transient (ramp-up, ramp-down), and Running Speed. These states are based off the running speed of the rotor. By utilizing several if-then statements the parame- ters for data acquisition can be changed to record desired information while avoiding unnecessary data storage and account for different requirements for data acquisition during each state. The conditions for these statements are known as set points. For example, if rotor speed is less than 300 rpm, then the state is in slow roll. If the rotor speed is between 300 rpm and 2900 rpm then the state is Transient. These
speed values constitute the set points for each state. Then within each state, indi- vidual triggers can be defined as described in sections 4.2.2 and 4.2.3. For instance when operating at Slow Roll, the measurements such as amplitude and phase are rel- atively constant and contain no dynamic information so very few measurements are need. During transients the rotating speed is changing so the acquisition time of each collection must be shortened so the frequency stays relatively constant within each collection period. Once at running speed is reached the frequency remains constant and higher resolution data with longer sampling times can be used.