Elsevier

Microelectronics Reliability

Volume 58, March 2016, Pages 113-118
Microelectronics Reliability

Observer based dynamic adaptive cooling system for power modules

https://doi.org/10.1016/j.microrel.2016.01.020Get rights and content

Highlights

  • Advanced cooling methodology proposal

  • State-space thermal modelling and use of observer for advanced/non-invasive implementation

  • Observer validation under fast transient response

  • Demonstrated reduction of operational ΔT

  • Demonstrated reduced degradation of power modules subjected to realistic mission profiles

Abstract

This paper presents an advanced dynamic cooling strategy for multi-layer structured power electronic modules. An observer based feedback controller is proposed to reduce a power device or module's thermal cycle amplitude during operation, with the aim of improving reliability and lifetime. The full-state observer design is based on a developed Cauer type thermal model. The observer enables estimation and control of the temperature at reliability critical locations only measuring one accessible location. This makes the method particularly powerful and suitable for application in power systems. The designed strategy is confirmed experimentally. Although the experiment is developed for a specific application scenario, the proposed strategy is of general validity.

Introduction

In power electronics, the failure mechanisms generally can be grouped by random and wear-out failures [1], [2], [3]. Wear-out mechanism failures make up the majority of failures in power electronic modules [4]. In wear-out mechanism, thermo-mechanical stress plays a very important role in affecting power electronic devices/modules reliability [1], [5], such as fracture propagation and degradations in solder layers [6], wire-bond lift-off [7] and emitter metallization [8]. The failure mechanisms are influenced by both environmental and load conditions [9], [10]. To address this issue, research has addressed different aspects, for example, new semiconductor and material technologies [11], [12], package architecture [13], interconnection [14], control of power electronic modules [15] and advanced cooling technologies [5], [16].

Fig. 1 describes a summary of the results of extensive reliability tests on IGBT power modules [17]. These results clearly indicated that, over the considered temperature range, a power module operational lifetime depends mainly on two parameters: 1) the amplitude of the thermal cycles, ΔT, that the module experiences; and 2) the average operational temperature, Tm. Fig. 1 clearly shows that if ΔT is reduced by even the same amount that Tm is increased, a much higher number of cycles to failure can be achieved. For instance, moving from point 1 to point 2, as ΔT is fixed at 50 K, increasing Tm by 20 K from 80 (353.15 K) to 100 (373.15 K), the cycles to failure will be reduced by 3 × 105 cycles from point 1 (5 × 105 cycles) to point 2 (2 × 105 cycles). However, moving from point 2 to point 3, keeping the same Tm, a reduction of 20 K in ΔT increases the number of cycles to failure to 2 × 106, that is, even better than the starting point 1. In other words, quantitatively, ΔT has a much more significant effect on the reliability of power modules than Tm.

Presently, typical power device thermal management only aims at ensuring that the maximum operating temperature is kept below a safety critical value at full-load or worst-case conditions and the cooling device is based on fixed designed parameters. In view of the close considerations, from a reliability point of view, this is clearly not optimum. Some temperature regulated thermal management strategies have been proposed with consideration of maintaining device operation temperature variation as small as possible [18], [19], [20]. A temperature control system is presented in [18], where the device under test (DUT) is sandwiched with a heat sink and heater. By electrically controlling the heater power, heat flow to/from the electronic device is quickly adjusted; and that in turn regulates the device temperature. This strategy is actually a heating strategy instead of a cooling strategy and it costs extra power for heating the device to a certain temperature value. The patent in [19] demonstrates a temperature controlled cooling method. In this method, the DUT is cooled by mechanically swinging the cooling fluid direction (e.g. the cooling fan facing direction) towards the heat dissipation element to regulate the temperature to a target value. However, this method requires several parallel mounted cooling fans and each fan needs a motor for swing functions, which increases the complexity of the cooling system and limits its thermal response time. In [20], a method to control the fan speed used in cooling integrated circuits is presented. In this method, a thermal diode is used to monitor device temperature, and the fan speed is adjusted by looking up a pre-defined temperature-speed table. There are two main shortages for this method: 1) a temperature sensor must be mounted inside the device; and 2) controlling cooling fan by look-up table is an open loop control, thus it is sensitive to system variations and easy to have temperature errors. Therefore, considering the reliability and temperature control issues, an observer-based adaptive cooling strategy with multi-variable feedback control technique is proposed here.

As shown in Fig. 2, the temperature with constant cooling power will vary as load changes. In order to decrease ΔT, the cooling power can be adjusted according to the load variations and this can be achieved simply by reducing the cooling power.

This paper presents an advanced dynamic cooling strategy for multi-layer structured power electronic modules. An observer based feedback controller is proposed to reduce a power device or module's thermal cycle amplitude during operation, with the aim of improving reliability and lifetime. The proposed methodology is schematically illustrated in Fig. 3.

The temperature at a reliability critical location of the power assembly is controlled against variations in the actual load and power losses Pdiss (i.e., power dissipation) and boundary condition Tamb (i.e., ambient temperature). The feedback control loop monitors the temperature of the desired location Tout and intervenes on the cooling parameter Vcooling to eliminate temperature errors Terr to control the temperature output and decrease temperature variations. The control parameter Vcooling is the controller output signal used to control the cooling devices. It can be the bias voltage applied on the fan for a forced air convection cooling, or the voltage on the pump in a liquid cooling system. By controlling the cooling device, the thermal impedance of the system, Zth, is adjusted to meet the temperature regulation. An observer based feedback controller is proposed to reduce a power device or module thermal cycle amplitude during operation, with the aim of improving reliability and lifetime. The full-state observer design is based on a developed Cauer type thermal model. To ensure the accuracy of the developed model, FEA (Finite Element Analysis) method is applied to derive a Cauer type thermal network where the observer is modelled on. The observer enables estimation and control of the temperature at reliability critical locations only measuring one accessible location. This makes the method particularly powerful and suitable for application in power systems. The designed strategy is confirmed experimentally. Although the case-study experiment is developed for a specific application scenario, the proposed strategy is of general validity.

Section snippets

Temperature estimation

A common way to estimate junction temperature is building a real-time thermal model and match the model to experiment data to get a reference look-up table for temperature estimation [21]. This requires high accuracy physical parameters, proper initial conditions to ensure the precision of modelling, high initial efforts to build up the look-up table and basically only for junction temperature estimation. Here, the proposed temperature full-order observer is a system that provides an estimation

Temperature control system design

Based on the proposed thermal modelling method and the temperature observer, the adaptive cooling system controller can be designed. Usually, the equations of the thermal model involving the controllable thermal impedance Zth (or using thermal resistance Rcooling for simplicity) and the cooling parameter Vcooling are non-linear. However, modern control system design is mainly based on linear system design: analysis methods are much easier for linear models than for non-linear models. Therefore,

Reliability testing: proof of concept demonstration

A photograph of the test vehicle is shown in Fig. 7: it is a commercial IGBT module mounted on a fanned heat-sink.

It was subjected to 3000 temperature and power cycles derived from a realistic mission profile for wind applications, shown in Fig. 8. The temperature at baseplate was measured during operation, a certain part of two cycles is shown in Fig. 9. With the collected baseplate temperature and the proposed observer, the reference temperature at the solder layer beneath the substrate is

Conclusion

In this paper, an observer based dynamic adaptive cooling strategy has been proposed and validated. Considering the effect of temperature variation ΔT and average temperature Tm on device reliability, the proposed cooling strategy uses multi-variable feedback control to regulate the device temperature against load variations (e.g., power and ambient temperature change). With the developed feedback controller, a higher temperature reduction in ΔT is achieved than the increase in Tm. This

References (27)

  • X. Perpina et al.

    Analysis of clamped inductive turnoff failure in railway traction igbt power modules under overload conditions

    IEEE Trans. Ind. Electron.

    (2011)
  • G. Buiatti et al.

    An online and noninvasive technique for the condition monitoring of capacitors in boost converters

    IEEE Trans. Instrum. Meas.

    (2010)
  • L. Ngwendson et al.

    An overview of the recent developments in high-voltage power semiconductor mos-controlled bipolar devices

  • Cited by (16)

    View all citing articles on Scopus
    View full text