By monitoring the air gap at Igarapava hydro power plant in Brazil, major damage and costly outages were avoided on a new hydro generator

THE Igarapava hydroelectric power plant comprises 5x42MW bulb units on the Rio Grande river which borders the states of Minas Gerais and São Paulo in Brazil. The owner is the Igarapava Consortium which comprises five companies: Companhia Vale do Rio Doce (CVRD), Companhia Mineira de Metais (CMM), Companhia Siderúrgia Nacional (CSN), Companhia Energética de Minas Gerais (CEMIG) and Mineração Morro Velho.

These were the first bulb units to be installed in Brazil and the owner insisted that they had to be fully equipped with a complete, on-line machine monitoring system which can monitor various generator parameters for condition-based maintenance purposes right from the units’ commissioning. The monitored parameters are:

• Rotor-stator air gap.

• Radial and axial shaft vibration at generator and turbine guide bearings.

• Hydraulic pressure.

• Generator MW, MVAR and voltage.

• Stator current.

• Exciter current.

• Stator temperature.

• Upstream and downstream water levels and pressures.

At Igarapava, the nominal rotor-stator clearance is 11mm, therefore air gap monitoring becomes even more critical, especially since stator sag (the tendency of the stator shape on a horizontal machine to flatten under the force of gravity) leading to air gap distortion is common in bulb units around the world.


On 27 July 1999, unit 2 at Igarapava had experienced rotor-stator contact within five months of unit commissioning. This led to a lengthy and costly outage to repair the unit. At the time, the machine monitoring system supplied by vibrosystm of Canada had not yet been installed for the start up of units 1 and 2 due to project constraints.

Upon occurrence of the rotor-stator damage, the commissioning of the zero outage on-line monitoring system (ZOOM) for all five units was accelerated while, concurrently, the main contractor was investigating the cause of the rotor-stator contact. From the contractor’s perspective, the utility had purchased the monitoring system specifically to avoid such a problem, and so it was only logical to make the system fully operational as soon as possible.

In September 1999, while at the Igarapava site to complete system installation and commissioning of units 1, 2 and 5, VibroSystM took the opportunity to perform test measurements on all five units. After reviewing polar plots and trends stored in the system database, an irregularity on the previously commissioned unit 4 was noticed.

A total of four air gap sensors are installed on the perimeter of the stator core approximately 25cm from the edge of the stator iron. Sensors are installed at the 45°, 135°, 225° and 315° locations.

During the tests of the ZOOM system on unit 4, an anomaly was detected in the rotor-stator air gap at the 225° sensor location. Based upon data the presence of a ‘bump’ on the rotor rim was suspected. Using the historical display capability of the software, it was possible to isolate at a point in time the signature (ie, minimum air gap value of each pole measured over one rotor revolution) of each air gap sensor.

In order to facilitate the interpretation of air gap data, VibroSystM references air gap measurement to rotor poles rather than to time. Simplified, this means that a phase shift is performed on three of the four air gap sensors to align all four air gap traces according to their poles.

By isolating the signature of each air gap sensor it was possible to identify if the bump was permanent or transient. A permanent bump would result in all air gap sensors seeing the same signature trace. A transient bump would result in one or more air gap signals showing a different trace. By using the ZOOM software, it was possible to determine that the bump amplitude varied depending upon the angle to which the rotor was turned. The maximum bump amplitude (or most critical air gap) occurred when rotor pole 39 passed in front of the 225° sensor (see figures above).

The 225°and 45° sensors were plotted and compared with the same data from a week earlier. From the signatures it was clear that even over the course of one week there was significant deterioration of the air gap.

In studying the data, VibroSystM alerted CEMIG that the monitoring system clearly indicated a potential rotor-stator air gap failure could occur at any time.

Realising the gravity of the situation and the potential danger of imminent rotor-stator contact, the CEMIG supervising engineer immediately contacted his head office. From its Belo Horizonte headquarters, engineers were able to remotely access the ZOOM data via their remote controller and confirm that the danger of an imminent rub was real. The relevant plots showing the results were printed and faxed to the generator manufacturer’s project office. Upon further study, CEMIG shut down the machine and requested the generator manufacturer inspect the rotor rim.

Two days after the order was given to shut down the unit, the generator manufacturer visited the plant to conduct further investigations. It discovered that unit 4 was in much worse condition than unit 2 when its rotor had contacted the stator. Percussion tests were performed on the rotor rim bolts to the spider. Several bolts broke in the course of testing and it was evident that the rotor rim was loose on the spider. During machine rotation, this imbalance was overstressing the bolts and causing them to break. CEMIG and the generator manufacturer have performed a detailed generator design review to correct the problem and prevent its reoccurrence.

Preventative action

This is a clear case in which the air gap monitoring feature of the ZOOM system was able to predict an imminent rotor-to-stator contact in time for preventative actions to be taken.

CEMIG is now closely monitoring the air gap and overall machine condition of all five generators to ensure the Igarapava Consortium’s investment is fully protected. On-line monitoring is particularly useful since this case has shown that a critical air gap change can occur over a matter of weeks. Periodic, off-line measurements would be insufficient to identify and correct a problem before it turns into a costly forced outage.