44a Wind Turbine 13m Fiberglass Blade Failure Analysis and Solution

Field Failure On March 2, 2011 at least one blade failed during what should have been normal operation of an Enertech 44A wind turbine near Forked River, NJ. All three blades ultimately separated from the turbine hub. Despite several requests, the owner would not allow the blades to be removed from the site for laboratory analysis. Based on photos of the failed blades it was obvious that the blades failed at the blade root. It is likely that one blade failed and the imbalance created gyroscopic forces and/or impact loads that broke the other two blades. The turbine with the failed blade had winglets (aerodynamic stabilizers near the tip of the blade) installed in an attempt to aid yaw stability (this turbine is one of two Enertech turbines with winglets).

E-13 Blade Project Objectives • Replace original wood blades – Composite – Maximize Efficiency • Modern airfoil • Design Simulation • Precise Tooling

– Self Starting and Stall Regulated – Centrifugal tip brake – Compatible with old hubs

– Eventually applicable to new upwind turbine – Economical manufacturing system Old New

Old

New

Prior to this incident, during maintenance procedures, our dealer observed cracks on the trailing edge root area of the blades of this turbine (See Fig. 1). Enertech personnel inspected the cracks and found no evidence of exposed fibers even after gently probing the cracks. It was our opinion that the cracks were not structural, but rather shrinkage cracks in resin-rich areas near the surface of the blade. After the failure on March 2, Enertech required all turbine blades inspected for similar cracks. Any turbine exhibiting similar cracks was to be shut down until further instructions from Enertech (See Enertech Product Safety Bulletin PSB11-2, March 4, 2011). Inspection results indicated that not all blades possessed cracks. Some turbines have continued to run without developing cracks.

Fig. 1

This blade design was completed in the spring of 2010. A static flapwise test completed on a blade sample in June 2010 resulted in flapwise moment capability in excess of two and one half times the survival load predicted by WT Perf (Ref. 1) for a 120 mph wind. The test moment was created using a distributed force along the length of the blade that approximated the calculated wind loading along the blade. Failure occurred by localized buckling approximately 40 inches from the root.

Failure Cause Investigation In an attempt to understand what caused the field failure, Enertech repeated tests of production blades at the factory and performed an audit of the manufacturing process. Static tests were performed on production blades March 5&6 (See Figure 2) . Static test results were inconclusive as one blade (SN B1057) failed at the blade root at a moment less than the June 2010 blade but in excess of 1.75 times the calculated survival load. The failure mode exhibited slightly different characteristic than appeared on the field photos. A second test (SN B1067) resulted in failure by compression skin buckling approximately 50 inches from the blade root at loads similar to the June 2010 blade. There were also differences in the deflections and flapwise natural frequency values of both blades as compared to the June 2010 blade. We assume the test blade that failed at the root to be the same mode as the blade that failed in service.

Fig. 2

Manufacturing Audit The manufacturing audit revealed that a step in the foam core preparation process may have been omitted on some blades. A fillet of foam is formed at the steel plate as a result of the foam core molding process (See Figure 3). This fillet needs to be removed to provide proper encapsulation of the steel root plate and load transfer from the surrounding composite laminate (See Figure 4). On some blade cores this fillet was not removed. This omission results in a change of the load path for compressive forces due to bending. The result is an altered stress field in the root portion of the blade creating higher in plane stresses and additional out-of-plane stresses at the root. These unanticipated stresses would likely cause cracks in the blade root that could propagate until blade separation. Both test blades exhibited varying degrees of foam fillet at the root plate.

Fig. 3

Foam Fillet

Fig. 4

Potential Cause Exploration A multitude of potential causes of failure were considered. We directed our efforts towards the causes we considered most probable. Nine potential causes of failure were identified including inadequate fiber resin saturation, water - freeze/thaw cycles, winglet affect on blade natural frequency, yaw instability, blade harmonics, foam core preparation, installation induced strains due to blade root end geometry, and coefficient of thermal expansion differences around the perimeter of blade root. Each potential cause was evaluated individually and in combination by analysis or testing. Individual and combination tests were performed to evaluate some of these potential causes.

Resin saturation was investigated by inspecting all blades in house for dry fibers. Holes through the laminate allowed visual inspection of the sections and dry fiber can be distinguished and quantified. No significant amount of dry fiber was found. The blade with the most dry fiber (SN B1067) was the second test blade static tested in March and no significant impact on ultimate strength was found.

Freeze/thaw cycles could create substantial loads. However, all turbine blades are subject to this condition and in fact the foam core should reduce susceptibility to water buildup of any significant thickness. Water between the foam core and the root laminate would only crush the foam core during freezing. The foam core is not a structural element of the blade. Freeze/thaw cycles is not considered to have contributed to the blade failure.

The blade harmonics were evaluated with and without winglets. The natural frequency of the tested blades was 3.30 Hz. Addition of a winglet lowered the natural frequency to 3.17 Hz. Winglet installation does shift the natural frequency slightly, however, it is still well between the harmonics of cyclic excitation force such as blade passage frequency. We do not believe blade natural frequency nor the presence of winglets contributed to the blade failure.

The gyroscopic loads from yaw motion were estimated based on yaw rates experienced in field and were small when compared to other anticipated service loads. Subsequent aeroelastic simulations of the 44a turbine confirmed load values. Gyroscopic loads are not considered to be a likely cause of failure.

To understand how a blade fails, Enertech developed a series of dynamic tests. The tests cycled a blade in blocks of flapwise and edgewise bending. The flapwise bending loads were approximately equivalent to a 100 MPH wind alternating direction 2 times each second based on deflection measurements. In addition thermal and installation stresses were imposed on the specimen. Under these extreme load conditions, Enertech was able to initiate and propagate cracks to imminent failure, indicating that the cause(s) of failure may have been identified.

The primary cause of failure is considered to be the presence of a foam fillet at the root plate. However, installation and thermal loads are believed to also contribute to the failures. The compressive interaction between the foam fillet, the fiberglass encapsulation and the steel plate is complex and appears to vary among blade samples. There is no simple method to determine the degree of foam fillet present inside the root of a blade. Consequently, all previously produced blades are assumed to have some degree of foam fillet and require a solution.

Solution for Previously Produced Blades (Bolster) To eliminate the structural problems created by the causes listed above, Enertech developed a steel bolster that surrounds the root of the blade reaching along the blade beyond the areas of the structural concerns. This bolster clamps around the blade root and secures to the blade with a structural adhesive. The bolster includes a steel base that inserts between the blade root and the cast rotor hub being clamped by the blade attachment bolts (See Figure 5).

Fig. 5

The design of the bolster includes specific elements to make the stiffness of the bolster compatible with the fiberglass blade and the adhesive. The modulus of elasticity of steel is about ten times the value of fiberglass. Bonding substrates with such different stiffness values can create severe demands on the structural adhesive. Enertech used finite element analysis modeling of the bolster side plate to optimize the stiffness of the shape and configuration while retaining sufficient strength for the loads. The structural adhesive is specifically chosen for its bonding and strength characteristics and a controlled thickness is maintained during application.

Validation of Bolster All of the blades tested to date have exceeded a flapwise ultimate design load in excess of 1.75 times the value predicted by aeroelastic analysis. Prior to April 2011 only static (ultimate) load tests had been performed. To validate the bolster installation design Enertech implemented a design and test program based on IEC 61400 (Ref. 2) standards as referenced in the American Wind Energy Association's "AWEA Small Wind Turbine Performance and Safety Standard" (9.1-2009)(Ref. 3) . This process involves calculating design loads based on turbine parameters, then calculating target test loads from the design loads and appropriate partial safety factors.

http://www.youtube.com/watch?v=79Z4fewYzoU

Testing and Results Enertech blade SN 1021-1 was removed from field service after approximately 300 hours of runtime. The blade showed small cracks at the root area. The blade was placed on the cyclic test stand and cycled aggressively edgewise for approximately 1.5 hours. At this point the crack in the trailing edge at the root had grown significantly to the point of imminent failure (See Fig. 6). The steel bolster was installed using the structural adhesive and the blade was remounted to the test stand (See Fig. 7). The test stand was fitted with strain gages giving the ability to monitor bending moments about perpendicular axes.

Fig. 6

Fig. 7

Conclusion The Enertech 44a wind turbine blade manufactured prior to March 2011 when modified by the installation of the Enertech blade bolster meets the design and testing recommendations of IEC 61400-2 and -23.

NWTC Design Codes (WT_Perf by Marshall Buhl) http://wind.nrel.gov/designcodes/simulators/wtperf/

2) IEC 61400 - Wind Turbine Generator Systems International Electrotechnical Commission PO Box 131 CH-1211, Geneva, Switzerland

3) AWEA Small Wind Turbine Performance and Safety Standard, AWEA 9.1 - 2009 American Wind Energy Association, 1501 M Street NW Suite 1000, Washington, DC 20005

4) FAST - Turbine Simulation Code is provided by the National Wind Technology Center and is available at: NWTC Design Codes (FAST by Jason Jonkman, Ph.D.) http://wind.nrel.gov/designcodes/simulators/fast/

5) Sandia Report SAND2002-0771, March 2002 Sandia National Laboratories