PFANF8 (2001) 2:49-52 © ASM International

Failure Analysis of a Polysulfone Flow Sensor Body – A Case Study Jeffrey A. Jansen (Submitted 12 December 2000; in revised form 5 January 2001) A failure analysis was conducted on a flow-sensing device that had cracked while in service. The polysulfone sensor body cracked radially, adjacent to a molded-in steel insert. This article describes the investigative methods used to conduct the failure analysis. The techniques utilized included scanning electron microscopy, Fourier transform infrared spectroscopy, differential scanning calorimetry, thermomechanical analysis, and melt flow rate determination. It was the conclusion of the investigation that the part failed via brittle fracture, with evidence also indicating low cycle fatigue associated with cyclic temperature changes from normal service. The design of the part and the material selection were significant contributing factors because of stresses induced during molding, physical aging of the amorphous polysulfone resin, and the substantial differential in coefficients of thermal expansion between the polysulfone and the mating steel insert.

Keywords: (author please provide 3-7 keywords)

Introduction had ranged from 40 to 200 °F (4 to 93 °C). Cleaning A flow sensor was submitted for analysis because agents, including chlorine-based and acid-based the plastic body section had failed while in service chemicals had also been put through the sensor. In in an industrial application. The part was represen- addition to the failed part, a typical sample of tative of approximately 10 other sensors that had molding resin was also received for comparison with been returned from service with generally compar- the sensor material. able features. It was also anticipated that similar failures went otherwise unreported, and that the Visual Examination premature failures posed a significant risk to other A visual inspection of the failed part confirmed parts still in service. numerous cracks within the collar of the plastic body section. The cracks were oriented radially, as shown Engineering drawings specified that the body in Fig. 1. The cracks were completed in the laboratory section be made from an unfilled polysulfone resin. and further examination of the fracture surface, using Polysulfone is a transparent amorphous engineering an optical stereomicroscope, showed relatively brittle thermoplastic, which, by comparison with other fracture features. No signs of significant ductility, as resins, is generally considered to be a ductile and would be evidenced by stress whitening or permanent tough material. Like many other plastics, however, it is somewhat notch sensitive. As a class of materials, polysulfone is known for good electrical insulation properties, excellent thermal stability, and exceptional hydrolytic stability. Because of its unique combi- nation of thermal and hydrolytic properties, polysulfone is commonly used in several demanding applications including food processing, medical components, and fluid handling.[1] The polysulfone body of the failed sensor had been injection molded around a tubular stainless steel insert to form the final component. Background information obtained on the application showed that the part was routinely exposed to alternating hot and Fig. 1 The failed flow sensor shown as received exhibiting cracks cold cycles. The fluid passing through the flow sensor radially within the plastic collar around the metal insert

Jeffrey A. Jansen, Stork Technimet, Inc., 2345 S. 170th Street, New Berlin, WI 53151. Contact e-mail: [email protected].

Practical Failure Analysis Volume 1(2) April 2001 49 Polysulfone Flow Sensor Body Case Study (continued) deformation, were apparent. Additionally, the comparison presented in Fig. 3. These results fracture surface exhibited beach marks, suggesting demonstrated that the failed part was fabricated from that the crack had propagated via fatigue. a typical polysulfone resin and showed no evidence of contamination or degradation. Scanning Electron Microscopy The fracture surface was further inspected using Differential Scanning Calorimetry scanning electron microscopy (SEM). The SEM The molding resin and the failed sensor were also examination indicated that the apparent crack origin analyzed via differential scanning calorimetry was located at an area along the inner diameter of (DSC). Again, the results that were obtained on the the sensor body, which contacts the molded-in steel materials were consistent. Both thermograms insert. This area had a relatively flat morphology, showed a single transition at approximately 187 °C, characteristic of brittle fracture. Surrounding the as illustrated in Fig. 4. This is in agreement with crack origin were features of alternating cracking and the expected results for a polysulfone resin, as these arrest cycles, consistent with low cycle fatigue, as resins generally exhibit a glass transition temperature

shown in Fig. 2. The final fracture zone showed fea- (Tg) between 185 and 190 °C. No other transitions tures associated with brittle overload, as evidenced were observed, and no signs of contamination or by the significant concentration of hackle marks, and degradation were found. exhibited limited ductility in the form of isolated stretched fibrils. Throughout the SEM examination, Thermomechanical Analysis no signs were found to indicate that post-molding A sample of material was excised from the inner degradation, such as chemical attack or thermal diameter of the sensor body adjacent to the fracture deterioration, had occurred. origin and analyzed using thermomechanical analysis (TMA). The resulting thermogram Fourier Transform Infrared indicated that the material expanded steadily as a Spectroscopy function of temperature through 170 °C. At that The sample of the molding resin and the failed point, just under the glass transition temperature, part were analyzed using micro-Fourier transform the material contracted, as a result of the load placed infrared spectroscopy (FTIR) in the attenuated total upon the sample during the evaluation. This con- reflectance (ATR) mode. The results obtained on traction was significant as it indicated that the the sensor body yielded an excellent match with those material itself was not under bulk residual molded- obtained on the molding resin. Both spectra in stress in the as-received condition. The results also exhibited absorption bands characteristic of a polysulfone resin. This is illustrated in the spectral

Fig. 3 FTIR spectral comparison showing a very good match between the results obtained on the failed collar material and those representing a typical molding resin. Both Fig. 2 Scanning electron image showing brittle fracture features spectra exhibit absorption bands characteristic of and beach marks suggestive of low cycle fatigue polysulfone.

50 Volume 1(2) April 2001 Practical Failure Analysis showed that the material had a coefficient of thermal Discussion expansion of 57 µm/m °C between 0 and 100 °C. Metal insert molding certainly offers distinct The steel insert from the sensor was also tested, and manufacturing and design advantages; however, it the results were consistent with those expected for has inherent risks. These liabilities include the this type of material. A review of the thermogram potential to generate excessive strain on the location indicated that the steel material had a coefficient of of the part surrounding the insert due to a differential thermal expansion of 19 µm/m °C through 100 °C. in the coefficients of thermal expansion between the A comparison of these results, illustrating the plastic and the metal insert. thermal expansion properties that approximate actual service temperatures, is included in Fig. 5. During molding, the plastic is in a molten state, which for the current material, was approximately Melt Flow Rate 370 °C. Through the cooling process associated with the molding operation, the plastic, having a sig- Melt flow rate determinations were performed on nificantly higher coefficient of thermal expansion, the molding resin and material taken from the failed will contract to a much greater extent than the metal sensor body. The testing was performed in insert. This difference in thermal expansion causes accordance with ASTM D 1238 – Procedure B using a high level of interference hoop stress on the plastic a test temperature of 343 °C and a constant load of material at the interface. Such stresses are exerted 2.16 kg. The samples were dried to a moisture on the part as soon as it cools from the molding content below 500 ppm prior to the analysis. The process. testing indicated that the molding resin had a melt flow rate of 8.19 g/10 min. This is in agreement with Additionally, amorphous resins, such as poly- the expected results based upon the grade of sulfone, may develop cracks over time due to aging polysulfone indicated on the engineering drawing. in conjunction with stresses.[2] This aging, commonly The analysis of the failed part material produced referred to as physical aging, occurs over an extended similar results, 8.03 g/10 min., showing a slight period of time, at temperatures below the glass transi- reduction in the melt flow rate for the failed part tion, as the polymer undergoes molecular motion in material. Such a reduction is commonly observed in an attempt to achieve an equilibrium state. Physical polysulfone resins as a result of melt processing, aging is particularly prevalent when an amorphous where molecular weight can increase due to thermal resin is cooled rapidly through the glass transition induced crosslinking. The magnitude of the change, temperature.[3] The condition is aggravated when the however, is low and the melt flow rate results suggest plastic resin is molded over a metal insert, as dimen- good retention of molecular weight through the sional changes often accompany physical aging, thus molding process, without molecular degradation. producing additional hoop stress at the interface.

Collar Material

Molding Resin

Steel Insert Failed Collar

Fig. 4 A comparison of the DSC results showing comparable Fig. 5 TMA plot overlay showing the differential in the heat flow profiles for the failed collar material and the coefficients of thermal expansion between the plastic collar molding resin material and the steel insert

Practical Failure Analysis Volume 1(2) April 2001 51 Polysulfone Flow Sensor Body Case Study (continued) Because of these phenomena, namely, the differential contributing factors. It appears likely that continued in thermal expansion coefficients of the plastic and temperature changes, encountered during standard the metal and physical aging, over-molding with use conditions, coupled with the differential in the amorphous resins, such as polysulfone, in con- coefficients of thermal expansion between the junction with metal inserts should be approached polysulfone sensor body and the steel insert, along carefully. with physical aging, compounded the inherent internal stresses associated with the interference at Conclusion the interface. This combination of stresses initiated It was the conclusion of the investigation that the the stress fractures radially around the insert over flow sensor body failed within the collar section of time in service. Given the notch sensitivity of poly- the part immediately surrounding the molded-in sulfone,[4] the cracks propagated through low cycle metal insert via brittle fracture, which was brought fatigue with the repeated temperature cycles. Final about by the exertion of stresses beyond the strength catastrophic failure occurred within some of the of the material. Examination of the fracture surface cracks due to the reduced cross section of the collar. revealed features associated with brittle cracking, with additional evidence indicating alternating References cycles of cracking and arrest, characteristic of low 1. James E. Harris: Handbook of Plastic Materials and cycle fatigue. Technology, Irvin I. Rubin, ed., John Wiley & Sons, Inc., New York, NY, 1990, pp. 487- 491. The location of the cracks, which were oriented 2. Plastics Engineering Handbook of the Society of the Plastics radially around the metal insert, was a strong indi- Industry, Inc., 5th ed., Michael L. Berins, ed., Chapman & cation that the source of the principal stress was the Hall, New York, NY, 1991, p. 752. interference between the plastic collar on the sensor 3. Julie P. Harmon and Charles L. Beatty: Engineered Materials Handbook, vol. 2, ASM International, Metals Park, OH, body and the molded-in insert. Although the analysis 1988, p. 751. results indicated that the failure was primarily assoc- 4. James E. Harris: Handbook of Plastic Materials and Tech- iated with mechanical stress produced during service, nology, Irvin I. Rubin, ed., John Wiley & Sons, Inc., New the design and production of the part are significant York, NY, 1990, p. 488.

52 Volume 1(2) April 2001 Practical Failure Analysis