Author(s):
Roberto Arnanz - Fundacion CARTIF
Anibal Reñones - Fundacion CARTIF
- - - Iberdrola Renovables
Our engineers in the Industrial Diagnosis and Predictive Maintenance Division at CARTIF are experienced developers of diagnostic, industrial-environment systems. These systems require that we capture a variety of signal types, including vibratory, electrical, and high- and low-frequency digital, as well as provide a high storage capacity because, in many cases, we continuously operate them during the full production run.
For wind turbines, data acquisition, diagnosis, and storage requirements at the feature- and design-level phases are similar to those of any rotary machine application. In our wind park application, we have to diagnose multiple machines; therefore, the quantity of data and diagnoses increases considerably and makes central processing difficult.
In collaboration with Iberdrola Renovables, we developed a solution using stand-alone diagnostic devices located in each machine, a distributed PostgreSQL database, one server for each wind park, and a central server. We decentralized data access by adapting the interface for use over a Web browser. With the LabVIEW Web server, we can develop user applications as if they were only going to be used locally and then we can publish them on the Web. As a result, we can use LabVIEW as a single development tool that integrates data acquisition, signal processing, and interface design regardless of where we installed the individual functions, how they communicate with each other, and how the user will implement them.
Data Acquisition
According to our diagnostic requirements, we have to capture multiple signals from each wind turbine. We installed the first prototype with eight ICP accelerometers, five capacitive accelerometers, three current clamps, three voltage sensors, and two inductive sensors to measure the velocity. After considering the variety of signal types, we chose to use the NI CompactDAQ system, which includes the NI cDAQ-9172 8-slot chassis, the NI USB-9233 accelerometer, the NI 9205 C Series analog input module, the NI 9423 sinking digital input C Series module, and the NI 9474 C Series digital output module.
This system met many of our data acquisition requirements including synchronized velocity signal capture with the rest of the signals used for analysis because the system under study operates at variable velocities. The system also performs continuous data acquisition over an extended period of time because the typical frequencies of the elements associated with the slow-turning shaft are very low and it captures a period for rotational velocity data that is not highly variable to facilitate analysis.
To establish the specified capture period, we stored data definitively only when the velocity variability percentage during the required capture time did not exceed a certain threshold. This method is equivalent to using trigger software in which the stored data corresponds to the pretrigger time and the trigger condition is determined by a calculation that determines the maximum velocity variation taken from past data.
In addition to the encoder signals, the system achieved a 25 kHz continuous transfer rate for eight analog channels. This allowed for continuous storage to disk and, consequently, the ability to acquire signals with this frequency for the desired amount of time.
The Diagnostic Application
Considering that the failure dynamics are relatively slow, the system acquires data in rounds and later processes the collected data. We carry out different captures (channels and frequencies) in each round, based on the scheduled diagnostics; and we store all of the results in the local database and send only the most significant results, or the alarms, to the central database.
Several modules make up the application. The supervisor module reads the database settings and, based on those settings, orders the execution of the various captures at the scheduled times as well as the subsequent processing and diagnostics when data is available. The user interface module provides access to the captured signals and the diagnostic results carries out simple analysis functions such as fast Fourier transform (FFT) display of one or several captures, and it has the ability to compare them with each other. Any user can connect to this interface through a Web browser without having to download the captured data.
The modular design facilitates run-time modification of the processing algorithms as well as the ability to add new ones without having to recompile the entire application. In this case, the algorithms are located in a dynamic link library (DLL) that can be edited at any time when a system process is not running.
The Diagnostic Network
The system includes management mechanisms to keep each of the databases up-to-date so that all of the machines continue to autonomously carry out the scheduled diagnostics even if we lose communication at one of the points.
We designed a supervisory control and data acquisition (SCADA)-type user interface, located in the central server where it can be accessed from any Web browser, so that the application user can quickly obtain all diagnostic information generated by the various machines and with a high degree of flexibility. To transparently access or analyze the captured signals, the user can connect to the interface on the machine rather than the one on the central server. We based this method on LabVIEW applications over an Apache server.
To communicate between the various network elements during the tests, we created a wireless diagnostic network that is independent from all other communications in the wind park.
Simplified Module Integration with LabVIEW
We used LabVIEW as the development software for the entire diagnostic application. The graphical programming environment simplified module integration into the R&D phase without any additional costs. While we independently designed the modules for data acquisition, processing, and diagnostics, the modular design facilitates the software development process and makes it possible to run the modules on different platforms (or on multicore systems with different microprocessors) based on the computing requirements and the equipment features. We implemented the system in wind parks managed by Iberdrola Renovables, whose collaboration has been crucial in carrying out this project.
Author Information:
Roberto Arnanz
Fundacion CARTIF
Poligono Tecnologico Boecillo
Boecillo-Valladolid
Spain
Print
PDF
