Characterising the Reliability of a Blower System for Medical Devices
This month's featured article is a guest submission from Deven Subramoney of Fisher & Paykel Healthcare.
The project's objective is to provide an integrated approach to reliability assessment, addressing the issues of component reliability as well as the blower system reliability. The document defines the reliability goals, sets out the reliability management approach taken, and identifies design for reliability and assurance activities executed to identify components and weaknesses that limit the blower system reliability and how to address them appropriately. The characterisation plan ties together customer requirements, a reliability goal statement, along with supporting evidence and methodologies on how this reliability goal was achieved. The blower reliability program is a subsystem of the overall system level reliability program for the medical devices and has been divided into 5 sections that integrate into product development cycles.
Identify Risks – Analyse & Assess
Quantify & Improve
Validate - Demonstrate Reliability
Monitor & Control Reliability
When characterising the reliability of the blower, we focussed on the best practice tasks that are most effective and applicable to providing highly reliable systems. This was established by clearly stated, measureable and meaningful reliability goals that have been converted into reliability activities which drove appropriate behaviour across our product team to identify and remove failures. Focus was on design for reliability tasks to confirm that the design meets reliability objectives and safety requirements, with minimal expected field problems.
A low cost medical device provides respiratory support to the patient by providing a flow of air to the patient's airway under a positive pressure (up to 30 cmH2O). A critical component of this device is the blower, which must maintain safe and effective performance for many years. Further, the blower motor must have low inertia and fast acceleration in order to maintain a constant pressure despite the varying air-flow demands throughout the patient's breath cycle.
One of the essential criterion of a Design for Reliability Program is defining goals that are needed to achieve the desired reliability. Allocation is a process for apportioning the device system target reliability (R95C95) amongst sub-systems and components. It is an essential tool at the design stage for determining the required reliability of the device to achieve a reliability target for a given system, in this case the blower. We decided to employ a series configuration equal allocation method for this medical device. Reliability allocation is always a tricky task which needs a "balancing act" to allocate reliability targets to ensure that the reliability requirements are met as a system. Therefore, as we improve over time we would adopt a more pragmatic approach for future product designs using a "Feasibility of Objectives" approach.
(Fig 1a) Medical device system reliability allocation - Displaying components of pressure subsystem only
The blower reliability requirements are derived from the overall system reliability model of the medical device. An equal allocation approach was used to determine the "reliabilities" of the five subsystems that make up the medical device. The blower is one of 3 components that make up the pressure system that has a reliability of R98.979. The blower reliability is therefore equated to R99.658 or 99.658% of blowers we manufacture must successfully function for a lifetime of 5 yrs. Furthermore, this statement must be supported by 95% confidence.
Identifying Risks – Analyse & Assess
This process was executed by filtering through a "list of failure probabilities" using FMEA, simulation and investigative type testing. This enabled us to implement several changes upfront.
Failure Mode and Effects Analysis (FMEA) is a qualitative reliability technique for systematically analysing each possible failure mode within the blower system and identifying the resulting effects. FMEA was executed by a cross functional team using their experience and judgement to determine appropriate priorities to action based on failure modes.
Lid/Volute Interface Fails to Seal Pressure Up to 30 cmH20
Lid and volute clip feature will creep. Volute diameter was increased to improve clip strength. Highly accelerated life testing (HALT) and accelerated life testing (ALT) were implemented to validate this change.
Stator Mount Fails to Isolate the Motors Vibration
Shock loading to blower dislodges stator mount isolating features. Included a clip feature on stator isolator to prevent unwanted detaching. HALT and system level drop testing were implemented to verify the change.
Bearing Mount Becomes Fully Dislodged - Impeller Contacts Pan/Lid
Creep of the shield bearing mount retention feature will result in bearing mount becoming dislodged. To reduce the risk of this failure, the radius on back of clip face was enlarged to reduce creep. Accelerated life testing using temperature would be used to verify this fix. HALT and accelerated vibration were also considered.
Weakened Solder Joints in Motor PCB
The PCB may move vertically with respect to the coil wire and fatigue the joint. A clip feature on shield was added to hold the PCB down, and the PCB was turned into a double sided through-hole plated board. HALT, accelerated vibration and thermal shock testing were used to validate this change.
Breaks in Electrical Connections of Terminations
Vibration generated from the stator can lead to fatigue failures of the terminations. To correct this, the PCB was turned into a double sided through-hole plated board. HALT, accelerated vibration and thermal shock testing were used to validate this change.
Magnet cracks & breaks into several pieces due to hoop stress in the magnet hub. A change of material was considered which has a high bonding grade of neo material and smooth internal curves without stress concentrations. To validate this change, accelerated temperature testing and HALT were executed.
The blower drive circuit generates ESD and induces a voltage on the shaft which leads to pitting corrosion on the bearing thus making it noisy. To correct this issue, ceramic bearings are now used to break the electrical circuit. Accelerated life testing, HALT and thermal shock were used to validate this change.
Impeller Hub Cracks
Impeller hub cracks due to fatigue as a result of poor tolerance limits. Fine element analysis (FEA) was employed to determine acceptable tolerance limits. HALT, accelerated temperature tests and thermal shock were used to verify this change.
High "Out of Balance"
The purpose of this test was to investigate the performance of the blower with a highly accelerated out of balance load, and identify potential failure modes and weaknesses that may occur in normal operation over its lifetime.
Failures Found During Testing
Fatigue on termination pin solder joint (Fig 2a), lower bearing mount separated (Fig 2b), wire fatigue on loom crimp (Fig 2c), bearing boss failure and stator windings started showing signs of fatigue (Fig 2d).
During the "out of balance" testing, the bearings became noisy and there was some concern as to the cause. There were a number of hypotheses about the root cause, one being high levels of temperature were destroying the grease and causing the bearing noise to become loud. Finally it was found that the motor drive circuit was generating ESD, which was travelling down the metal shaft of the blower and causing pitting corrosion (Fig 3b), which over time made the bearings noisy. All attempts were made to reduce the ESD generated by the motor drive circuit, but this proved very difficult. To resolve this issue we had to change to a ceramic bearing to break the conductive circuit. The noise levels of the 12 blowers with original bearings were measured after 2 weeks and then 2 months. The noise levels of the 6 blowers with the new ceramic bearing were measured after 2 months also. The graph (Fig 3a) below displays the worst noise levels recorded of all blowers.
Green – Blower displaying worst noise level after 2 weeks @ 70°C with impeller
Red – Blower displaying worst noise level after 2 weeks @ 70°C without impeller
Blue – Blower displaying worst noise level after 2 months @ 70°C with impeller
Purple – Blower displaying worst noise level after 2 months @ 70°C with impeller and ceramic bearing
Decisions Made on Failures Found During Investigative Testing
Cracked Termination Pins
A change to a double sided through-hole plated PCB will solve the cracked termination pin issue, and a clip was added to the stator former to provide mechanical retention on the stator PCB.
Failure of Bearing Mounts
These were verified in the HALT and accelerated vibration testing.
It is believed that the bearing failures and noise are a result of pitting corrosion caused by ESD generated from the motor drive circuit travelling down the metal shaft of the blower. We have replaced the steel bearings with ceramic bearings.
Loom and Winding Future Failures
These were verified during HALT and accelerated vibration testing.
Finite element analysis (FEA) was implemented to predict how the blower reacts to real-world forces, vibration, heat, fluid flow, and other physical effects. Finite element analysis suggests whether our blower will fatigue, wear out, or work the way it was designed. SOLIDWORKS Static simulation uses the displacement formulation of the finite element method to calculate component displacements, strains, and stresses under internal and external loads with results assumed representative up until the elastic limit. Simulations were executed on the blower clip (Fig 4a), impeller hub (Fig 4b) and rotor (Fig 4c). The maximum stress in the hub at the outer edge is 189.364 MPa. The maximum stress generated at the outer edge of the impeller hub is less than 80% of the materials tensile strength. (189.364 MPa / 241 MPa = 78.6%) Note: 241 MPa is from the material datasheet, which is not supplied in this report. Results suggested that the acceptance criteria was satisfied and therefore considered a pass. The simulations performed on the clip clearly show that changes were needed to improve the clip feature design. A force of 5.5 to 25 N was applied and the part was found to exceed the material yield strength of 127 MPa. Simulations were performed on different rotor designs and their effects on moulding stresses. The applied injection pressure was 10 MPa. In the true sense of the analysis these simulations were used as a guide as to what options we should start off with. Based on the results, the 8.5 mm design was used as a starting point for tool trials which were later validated through testing.
Failure Reporting & Corrective Action Systems
An essential part for optimising blower reliability is the reporting of failures, their causes and the corrective actions taken to rectify problems and reduce the probability of their future recurrence. When recorded appropriately, this information will be used to understand any weaknesses and apply the lessons learnt from the corrective actions that were implemented. The blower team has developed and documented a closed loop failure reporting, analysis and corrective action process for all failures occurring during development, manufacturing and infield stages. Failure reports and failure summaries have been recorded and evaluated thus far. FRACAS promotes reliability improvement throughout the life cycle of the blower and forms an integral part of our reliability program. "We must not lose the opportunity to learn from failure."
Detailed Design – Design for Reliability
The improvements made thus far were analysed and further changes were implemented on any concerning issues found. This was executed by implementing thermal profiling, HALT and several accelerated life tests to quantify the life of the blower.
5 medical devices were set up to operate under typical, average and extreme user conditions powered by a supply voltage of 85 VAC and 265 VAC whilst being subjected to an environment of 35°C. During this sequence, 40 thermal probes were placed on critical components to measure their temperatures on each device. The enclosure of blowers and the bottom of the stator windings were measured. The data were used to get some information of the temperature spread across a population of devices and their blowers, and to provide some guidance in derating analysis for other electronic circuits within the device.
HALT – Highly Accelerated Life Testing
The blower was subjected to the HALT process to uncover design weaknesses by subjecting it to progressively higher stress levels brought on by thermal dwells, vibration, rapid temperature transitions and combined environments. Throughout the HALT process the intent was to subject the blower to stimuli well beyond the expected field environments to determine the true operating and destruct limits. Failures, which typically show up in the field over a period of time at much lower stress levels, were quickly discovered while applying high stress conditions over a short period of time. In order to ruggedize the blower, the root cause of each failure mode was determined and the problems corrected to ensure design integrity. This process yielded the widest possible margin between blower capabilities and the environment in which it will operate, thus increasing the blower's reliability.
(Fig 5a) HALT Profile Log
Four blowers were mounted onto the vibration table and set up inside the HALT Chamber, with no flow restrictions. Drive boards were configured to run at a fixed speed of 19000 rpm and set up with the power supply external to the HALT chamber. Functionality was continuously monitored during testing. The blower demonstrates an acceptable level of robustness with regards to thermal and vibration stresses. Only 2 failure modes were created during this HALT test, the lids kept popping off after 125°C and beyond, and the stator PCB fatigued against the volute at 100 GRMs. We have decided to increase the volute diameter which will improve the clip strength and reduce the chance of the lid popping off.
Accelerated Life Testing - Vibration
The blower will generate continuous vibration over the 5 years of operation. Therefore, it was decided to perform some form of accelerated vibration test that would be equivalent to its expected operating lifetime of 14600 hours and thereby determine if there are any dominant failures within this life time. Four blowers were mounted onto the vibration table and set up inside the HALT chamber, with no flow restrictions. Drive boards were configured to run at a fixed speed of 19000 rpm and set up with the power supply external to the HALT chamber.
The acceleration model used for this test was based on the Coffin-Manson model described by Caruso and Dasgupta (1998) and is a representative example of Power- Law models for accelerated testing. The required life time at acceleration level g2, N(g2), that corresponds to an equivalent amount of damage to N(g1) cycles at acceleration level g1 is:
where "b" = exponent based on material and damage mechanism. The exponent b represents a fatigue characteristic of a particular material, as derived from the slope of the S-N curve for that material. As noted by Caruso and Dasgupta, beta is a function of both the material and the damage mechanism. Typical b values for ductile metals are reported as 5-6 and for plastics 4, with regards to vibration. We decided to use b = 3 to be conservative.
Vibration measured whilst blower in operation = 1 - 1.8 Grms (measured with blowers at 19000 RPM)
Vibration at "Use Level" = 5 Grms (conservative)
Expected life duration @ 5 Grms = 14600 Hours (8 * 365 * 5)
b value used = 3 (conservative)
Accelerated vibration used in test = 50 Grms
14.6 hours of vibration @ 50 Grms would have an equivalent damage of 5 Grms for 14600 hrs. Whilst being exposed to vibration, we also exposed the 4 units to numerous thermal shock cycles ramping the temp up to 110°C and down to -40°C. All 4 blowers exceeded the 15 hour test, running for 21 hours and 56 minutes. No failures or weaknesses were found.
Accelerated Life Testing - High Temperature
This test was designed to accelerate failure mechanisms brought about by steady state high temperature. Increased temperature will accelerate most degradation processes that are present in the blower design. Principal failure mechanisms may affect components such as the magnet, ceramic bearings, bearing grease, plastic components, windings, PCB and rotors. It is assumed that these dominant thermally accelerated failure mechanisms will follow the Arrhenius life-stress relationship. This test process was set up with 3 different rotor designs with all necessary changes incorporated and assembled into blower units. Testing was executed using 2 temperature stress levels. We preferred to test using 3 stress levels, but availability of test chambers proved to be a constraint. Blower units were placed into environmental chambers with no flow restrictions, and one chamber was set to 100°C and the other to 115°C. The test temperatures chosen were based on previous trial testing performed and have proven not to cause over stress failure mechanisms. Drive boards were configured to enable the blowers to run at a fixed speed of 19000 rpm whilst being exposed to the temperature stresses. The units were inspected periodically to verify functionality status. The rotors proved to be the weakest link in the blower design and, after applying acceleration factors and projecting life at "use conditions" 45°C, the 8 mm rotor design was deemed acceptable. No other failure modes were discovered during this test period, and the blower and its components have demonstrated the reliability requirements of R99.658 C95 initially set. It must be noted that an improved rotor design "A2" was also tested.
(Fig 6a) Probability plot of blowers with different rotor designs
The analysis performed on each of the 3 rotor designs employed the Arrhenius life-stress relationship, and the life distribution represented a lognormal pattern. The activation energy, calculated from the 3 sets of testing, represents the minimal necessary energy to create particular failure mechanisms and is calculated as follows:
In order to predict life at the "use" temperature of 45°C, the data from both tests from all 3 rotor designs were transformed by corresponding acceleration factors. The formula for calculating acceleration by temperature is:
Accelerated Life Testing - Temperature & Humidity
This test was designed to accelerate failure mechanisms created by humidity, which may be seen due to reverse flow over the chamber. Some of the materials used in the blower design may be susceptible to hydrolytic degradation. In particular, the components being considered were the impeller (30% carbon fibre filled Nylon 66), the stator shields (30 glass filled PBT) and the rotors. Twelve blowers were placed into the environmental chamber with no flow restrictions. Drive boards were configured to enable the blowers to run at a fixed speed of 19000 rpm whilst being exposed to a test temperature of 70°C, 95% RH for 1600 hours. The units were inspected periodically to verify functional status.
The Temperature Humidity Bias model used for this test is referred to as the "Peck model," which includes a relationship between life-and-temperature (Arrhenius model) and life-and-humidity (Peck model), so that the product of the two separable factors yields an overall acceleration factor.
M – Power parameter for humidity, normally used 2.6
Tuse – Temperature of blower during use = 45°C
Ttest – Temperature on blower during test = 70°C
RHuse – Humidity in the area of blower during use = 50%
RHtest – Humidity in the area of blower during test = 95%
Ea – Activation energy assumed = 0.6
kBoltz – Boltzmann’s constant eV / degree K = 8.60E-05
Acceleration factor = 27.28
Test duration completed = 1600 hours
The blower is expected to be exposed to 50% humidity for 12 hours per day
Projected life in days = (1600 x 27.28) / 12 = 3638
The 12 blowers were inspected and no failures were observed during this test. Our preference was to test to failure at different stress levels to extrapolate life, but chamber availability was a constraint.
Accelerated Life Testing - Thermal Shock
This test was designed to accelerate failure mechanisms brought about by thermal shock. Some of the materials used in the blower design may be susceptible to fracture, brittle components such as the magnet or ceramic bearings and rotors. Ten blowers were placed into a thermal shock chamber with no flow restrictions. Drive boards were configured to enable the blowers to run at a fixed speed of 19000 rpm whilst being exposed to 600 thermal shock cycles ranging from 110°C to a low temperature of -50°C with a dwell time of 30 minutes at each temperature extreme. The units were inspected periodically to verify functionality status.
The cyclic stress created is related to thermal expansion and contraction, which occurs in materials used in the blower design. To relate field usage to accelerated test conditions, the Coffin-Manson model was used which is actually an "inverse power law model." Fatigue exponents typically used range from 1.9 to 4. For this test we have been a bit conservative and used 1.9 as the fatigue exponent, and the gradient of the shock was ignored as part of the acceleration factor.
N – Fatigue exponent of weakest material, normally used 2.5 to 4, we used 1.9 (conservative)
Delta Tuse – Difference in operating temperatures in the field (extremes) 50°C - 25°C = 25°C
Delta Ttest – Difference in operating temperatures during test 110°C - (-50°C) = 160°C
Dwell time @ -50°C = 30 minutes
Dwell time @ 110°C = 30 minutes
Acceleration factor = 34
Test cycles completed = 600
One thermal cycle represents one day
Projected life cycles = (600 x 34) = 20400
The 10 blowers were inspected and no failures were observed during this test. Our preference was to test to failure at different stress levels to extrapolate life, but chamber availability was a constraint.
Drop Testing - Impact Resistance
The medical device will be subject to handling during transport, installation, or repair and is at risk of being dropped. The best way to ensure that the device survives its journey from the factory to the point of installation is to drop test it and verify that it survives without damage. In addition to ensuring that the device survives its journey from point A to point B, it could still be at risk of being dropped throughout its lifetime. Drop testing was executed to ensure that the medical device is capable of surviving the perils of expected use and abuse. Twenty device units were exposed to drop testing from a height of 700 mm onto a hard surface. After each fall the unit device is visually examined, and electrically and mechanically verified.
Upon inspection it was found that in several units, the power connector on the blower PCB had detached itself (Fig 7a). Markings on the air box upper indicated signs of impact on to the connector, directly resulting in failures as seen below. Double crimped through-hole connectors (Fig 7b) were then chosen because of solder joint and mechanical retention failures found in the previous connector design. To validate this change, 4 units were exposed to the accelerated vibration test profile performed earlier. No failures or weaknesses were found. Twenty new device units with new blower connectors were also executed to drop testing and no connector failures were found.
Controlled Field Trials – Residual Failures
We will be performing field trials on medical device units. Users will operate the device as they would in actual usage and regular interviews will be conducted with users in order to plot their experiences in using our product. In addition, the device has automatic data logging facilities. The results of field trials will provide valuable information to us regarding the potential for improving our device. Residual failures can be potentially expensive which can be identified and rectified. In addition, the use of field trials has good face validity, unlike our in house testing.
Validate & Demonstrate
To demonstrate and provide assurance that we have adequately designed and manufactured the blower to meet reliability requirements, the blower will be part of the device level reliability demonstration test. To determine if our reliability objective of R95C95 for 5 years expected life for a medical device has been met, a reliability demonstration test will be executed. A "non-parametric" approach will be used since it reduces uncertainty of "distribution parameter assumptions." A sample size of 59 device units will be utilised for this demonstration test. The required samples have been calculated as follows:
The blower will be part of the reliability demonstration test that simulates device life cycle loading sequential by exposing 59 device units to transportation vibration, storage in transit, shelf life and finally operating life testing.
Wilcoxon-Mann-Whitney Rank Sum Test
As a benchmark to demonstrate "better design," the blower will be compared to previous blower designs using the HALT process. This statistical comparison is a non-parametric approach using the Wilcoxon-Mann-Whitney (ranked sum) test. By subjecting four manufacturing pilot blower units to a Highly Accelerated Life Test (HALT) and comparing the results to HALT results of previous blower designs with known field reliability, the new blower design robustness can be compared with a degree of confidence.
Monitor & Control Reliability
We have also incorporated processes in place that will help sustain, monitor and control the blowers reliability, i.e., Highly Accelerated Stress Audit (HASA), ongoing reliability testing, life data analysis and reliability improvements.
System Level HASA Screening
In our efforts to maintain the designed reliability of the blower, it is necessary to monitor our manufacturing process and stop "infant mortality" related defects. Medical devices will be exposed to a HASA test screen of 5 cycles that combines high/low temperatures coupled with vibration, voltage margining and power cycling. HASA is a process where a statistical sample of devices will be screened and, based on the results, root cause and corrective actions will be pursued. The HASA profile margins will be designed from the information found in our final round of HALT testing.
Ongoing Reliability Testing & Supplier Quality
It is recommended that HALT testing be executed periodically to detect design margin drift, as HASA may not be able to detect subtle changes in our processes. Verification testing will be executed periodically and our current inspection strategies for supplier components are continuously improving. It is imperative that we maintain a good quality of components being installed into our devices.
Life Data Analysis & Reliability Improvements
Device field data will be analysed periodically and data analysis information will drive actionable decisions within cross functional teams. Our life data analysis is set up such that it feeds back into our FMEA process, providing a feedback loop into improvement strategies for design, supplier and manufacturing processes and also provides a measure of effectiveness of our HASA screen. Ideally, failures and weaknesses in the blower design and manufacturing will therefore be highlighted and addressed accordingly. All changes and improvements made that are related to failures due to design, manufacturing and supplier related issues will be validated by appropriate reliability testing.
The design reliability requirement of R99.658 C95 for the blower has been met by employing a comprehensive assessment program in a methodical manner, where reliability requirements have been well defined and appropriate tools and processes have been applied to design, demonstrate and maintain reliability. Our methodology to verify that the blower meets the reliability requirement was centered on initially identifying potential weaknesses through the application of FMEA, exploratory testing and simulation, followed by HALT testing to further challenge the product. Lastly, accelerated life testing was used to quantify the life of the blower.
The information contained within this project should serve as an overview of the techniques and requirements for implementing a reliability program, providing a general understanding of the concepts and principles applied to achieve a "robust design." For future development we realise the need to improve our simulation techniques to better understand the expected stress levels at the critical failure sites under life cycle loading profiles and accelerated test loading profiles. Such simulations, when combined with simulation-guided testing, can often provide important new insights and help quantify product vulnerabilities. The guidance presented in this project can enhance current "design for reliability" practices across the company and as such, it is hoped that all teams across the company will find the provided guidance a benefit.
About the Author
Deven Subramoney is a Senior Reliability Engineering Specialist at Fisher & Paykel Healthcare, a leading medical device manufacturer. He has over 25 years of experience in the field of Reliability Engineering for multiple product development and manufacturing companies. His experience covers a broad range of electrical, electronic and mechanical products, spearheading the setup and growth of reliability functions, focusing on the development of the reliability process, reliability engineering labs, engineers and integration into the business.