MultiPlane Balancing Turbocharger Aplication

Best practices for Flexible Balancing: Application to turbocharger balancing

Turbocharger Balancing

Turbocharger rotors are small items but because of their high rotational speed, they may experience several critical speeds and should be considered as flexible. For that reason it is not possible to balance the rotor at only one speed: several speeds should be considered: so a run-up or a run-down.

The number of possible balancing planes are typically up to 4: 2 on the compressor side and 2 on the turbine side. Depending on the case, it may not be possible to add or remove weight depending on the material characteristics (drilling, scrapping, grinding) on each of the planes. Then, a selection of the planes has to be considered: typically the plate of compressor side may be a good selection. The blades stay in principle untouched.

Possible balancing planes on turbochargers:

The balancing procedure should be considered similarly to a modal analysis. The idea is to determine the relationship between the response (vibration measured at the bearing) and the exciting force (the imbalance). So adding a known weight is like impacting a structure with an impact hammer and measuring the injected force. Once the relationship is known (the influence coefficient matrix), it is easy to locate the original force (imbalance) and propose and correction weight.

The idea is to excite the 1X component with an imbalance force. The excited response should be as much as possible due to the imbalance. Any other reason (typically, a structural resonance of the bearing) will introduce errors, will decrease the precision, and will make the balancing correction less efficient.

What should be taken care of when going through flexible balancing?

The different choices that the user has to do chronologically are:

• Sensor positioning: Sensors should give results that are highly impacted by the imbalance (original or trial weight related). Typically accelerometers are used: they should preferably be positioned in a direction where the bearing is less stiff (often the horizontal direction).
• Setting of the input range: After the initial balancing, it may appear that the 1X component becomes smaller. It may be necessary to adapt the input range (doing an auto range for example).
• Trial weights positioning: The trial weights position is important as well as it should bring new information compared to the original run or other trials.
• RPM run up/down rate: It is important to have the same rotor acceleration between the initial runs and the trial runs. The structure may not behave in the same way depending on how quickly speed increases or decreases.
• Data: Data of the different runs should be synchronized from one to another. For calculating the influence coefficients matrix, the data of the different runs (Initial, Test 1, Test 2 etc. …) should be comparable at the exact same RPM value.
• RPM selection: it is probably the most sensitive point to take into account when achieving balancing. The data should be selected with care because the wrong data or too much data would get the user to biased results.
• Trim Balancing or not: To be able to carry trim balancing (further balancing after rotor weight correction implementation), one has to be sure that the system stays linear: meaning that the same matrix of influence coefficient can be used. If, it is not the case, one has to use this measurement as a new initial run.

Hopefully there are useful criteria that help the user to do the good choices! Let's go a bit more in details!

Sensor positioning

Accelerometers are often placed in the horizontal direction as they are more sensitive to imbalance.

Other types of sensors such as force sensors or proximity probes can be used.

During balancing, the purpose is to concentrate on the rotor independently from the surrounding supporting or driving structure. It is very comparable to modal analysis when a structure is hung with rubber band. In this way, one wants to measure the response of only the test item independently from the rest.

Trial weights positioning

Trial weights should be positioned in order to excite the modes that should be balanced. In the example below, the 3 first modes are excited by the weights. Run 1 excites the first rigid mode, Run 2 the second one and Run 3 excites the first bending mode.

A bad situation is for example if the trial weight is positioned exactly at the original imbalance position. In that case, it doesn't bring new information.

Data synchronization

The data of the different runs (Initial, Test 1, Test 2 etc …) should be comparable at each RPM value: one should compare for example the value of the initial run at 335 RPM, with the value for test 1 at the exact same RPM value (and not 334 RPM or 336 RPM): the larger differences, the larger errors will get. For that reason, it is very useful to interpolate the data that have been acquired.

Importing and interpolating data every 5 RPM	Interpolated data

RPM selection: Tips, Tools and criteria

When choosing the right RPMs. One should consider that each set of RPM, 1X Initial, 1X Run1, 1X Run2 is used to calculate a response matrix (called influence matrix). Having more force/response relationship information (staying in a linear domain) at different speeds helps to do a better curve fitting and calculate a more accurate matrix of influence. If for different speeds, the structure responds in the same way (same amplitude, stable phase), it doesn't bring enough new information to solve the problem and calculate the matrix of influence in the most accurate way.

In addition, the response should be representative of the influence of the imbalance: a structural mode of the bearing for example will bring erroneous information.

One should avoid

• Structural resonances of the support (impact testing of the support/bearing may be useful prior to the balancing test).
• Stationary phase (not much information).
• Resonance itself as the phase variations may be unstable (noisy phase).

One should choose

• RPM range where the phase changes are large but stable (just before or after the resonance)
• RPM range where we see clearly an impact of the trial mass over the original situation (like the green area on the example).

How to validate choices: The prognostic tool!

The prognostic tool is also a good way to check that the right choices have been made.

In every case the prognostic is calculated from the influence coefficient matrix. If the selections are the good ones, the prognostic should be close to the original when putting a correcting mass of 0 (as default). It is not supposed to be exactly 0 but should be close to it. If it is not the case it may say that something else than imbalance is taking place: like a structural resonance for example.

The example below shows the example of a proper speed selection and a bad one.

Right Speed selection	Wrong speed selection