1. Introduction

In recent decades, electric powertrain systems have attracted more and more interest due to the requirements of future trans- portation, energy conservation and emissions reduction around the world.

Industrial application drives and electric vehicle have several requirements related to the electric propulsion control. It is the ability to operate at constant power over a wide speed range, good overload performance and high efficiency etc in the following that ⁽¹⁾ :

∙ High instant power and high power density

∙ High torque at low speeds for starting and climbing, as well as high power at high speed for cruising

∙ Very wide speed range including constant-torque and constant- power regions

∙ Fast torque response and

∙ High efficiency over wide speed and torque ranges

∙ High efficiency for regenerative braking

∙ High reliability and robustness for various vehicle operating conditions and reasonable cost

The high power density and the capability to operate continuously under fault conditions are necessary characteristics for a motor used for EVs.

These kinds of typical faults in traction motor include: failure of one or more power electronic components of the drive system, opening or shorting of one or more of a stator phase winding and the other mechanical components faults etc. Especially, the open-phase current fault affects over the stator current and then introduces excessive torque oscillations leading to mechanical load defect and bearing premature failure etc ^(2,15).

This paper proposes the open-phase current fault conditions of motor control system for electric vehicle and the applied fault diagnosis is performed with a simple hardware signal conditioning circuits without a internal complex detection software algorithm.

Fig. 1. Typical case (a) droid (b) e-cart (c) autonomous vehicle (d) electric traction system

The proposed method is faster than the conventional method because it can instantaneously detect a fault condition that defines the open-phase current fault using the change in pulse width.

Fig. 2. Control block diagram for detecting open phase fault conditions

2. Fault condition and diagnosis

2.1 Conventional method

In order to obtain high-reliability motor drives as shown in Figure 1, modern control strategies like field-oriented vector control should be employed. These techniques are inherently dependent on the measurement device, or sensors that should operate properly.

However, if such a sensor fails, the control logic must com- pensate for the fault function properly or run independently of the fault condition.

In Figure 2, CASE Ⅰ shows the conventional vector control block diagram for traction drive module and the open-phase fault condition. In this block, two motor phase currents are generally measured and the other phase current is calculated by the minus sum of two currents as shown in Equation (1).

(1)

$i_{c}= -(i_{a}+ i_{b})$

These measurements feed the Clarke transformation and the outputs of this projection are designated $i_{\alpha}$ and $i_{\beta}$ in the stationary reference frame. These two components of the current are the inputs of the Park transformation that gives the current in the d, q rotating reference frame.

In this paper, as described in ⁽¹⁵⁾, it is proposed that the single-phase failure detection system may determine whether the fault occurs or not using the waveform of the three-phase current as below Equation (2).

(2)

$\left | i_{a}\right |{and}\left | i_{b}\right |{and}\left | i_{a}\right | <\dfrac{\sqrt{3}}{2}\alpha\sqrt{(i_{d}^{*})^{2}+(i_{q}^{*})^{2}}$

where $i_{a},\: i_{b},\: i_{c}$: phase current & $i_{d},\: i_{q}$ : d-/q-axis current & $0<\alpha\le 1$

However, since this equation assumes that the sum of the currents is zero if there is an error in the signal conditioning circuit, the algorithm does not recognize whether the resulting error signal is a true open-phase current error or a signal condi- tioning error.

Fig. 3. Three phase motor current and vector diagram (a) 3 phase current waveforms (b) the sum of instantaneous current (c) vector diagram at normal operation (d) vector diagram at signal conditioning circuit fault (e) vector diagram at an open-phase fault condition

2.2 CASE study: fault condition & detection

Before suggesting a main topic, it is necessary to limit the scope of application. That is, when the over-current flows to the traction drive due to

Fig. 4. fault conditions | (upper) 3 phase current waveforms (middle) operation flag (under) the sum of instantaneous current (a) over-current condition (b) signal conditioning circuit fault (c) open-phase current fault external load under normal operation condi- tion, the control logic can generate an alarm, not a fault signal.

Therefore, the functions that can protect the system using the related alarm or protection devices in the proposed 3-phase inverter were excluded from the case study topic.

Figure 3 shows the three-phase motor current and vector diagram in the ideal condition. When 3 phase current has a periodic waveform of 120º, the sum of the currents is zero and has the vector diagram shown in Figure 3(c). However, if one of the three signal conditioning circuit has a fault, it will look like Figure 3(d).

Although Figure 3(d) is not the real open-phase current fault, the control logic based on the mentioned Equation (2) may determine this condition as the single phase open fault.

Figure 3(e) shows the real open-phase fault due to the electromechanical open-circuit, which has a 180º two-phase current vector diagram.

This condition corresponds to CASE Ⅱ as shown in Figure 2 and the mentioned control logic can correctly diagnose the open-phase current fault condition.

Figure 4(a) shows the waveform of over current fault, where the sum of each phase current is zero.

Figure 4(b) shows the waveform of signal condition- ing circuit fault, where the sum of each phase current is not zero but sinusoidal value. Therefore, the control logic can easily detect this fault.

Figure 4(c) shows the waveform of the real open-phase current fault, where the sum of each phase current is zero, but the resultant current become higher than before in short time. Therefore, the control logic can easily detect this kind of fault condition.

3. Proposed method

This paper proposes the open-phase current diagnosis method that can correctly and easily detect both of CASE Ⅰ and CASE Ⅱ. This method uses the simple signal conditioning circuit to detect three-phase currents based on an additional H/W circuit without a complex diagnostic software algorithm.

As shown in Figure 2, the proposed method corresponds to CASE Ⅱ.

Figure 5 shows the method of monitoring phase current which is based on CASE Ⅱ as mentioned in Figure 2. When the three-phase current alternates periodically at a constant speed, the $I_{a}$ pulse has a pulse waveform if $I_{a}$ flows through the additional H/W circuit as shown in Figure 2 and Figure 5(b), where the pulse width depends on the impedance in the H/W circuit. The other two currents $I_{b}$ and $I_{c}$ also have a pulse-like waveform in the same manner in Figure 5(c) and Figure 5(d).

In general, the mentioned pulse waveform depends on the frequency of each phase current and so the pulse width also is determined. Therefore, when an open-phase current fault occurs, the corresponding pulse waveform is different from the normal pattern.

In order to improve the diagnosis response, logical sum for the 3-phase pulse - that is - XOR is used to detect the open-phase fault condition and as a result the response time for diagnosis can be 3 times faster than before as shown in Figure 5(e).

As shown in Figure 5(e) and 5(f), the diagnosis methods of the open-phase current are the logical sum (XOR) and the arithmetic sum ($I_{s um}$). The proposed method can detect the error conditions mentioned using XOR.

Fig. 5. Proposed method of monitoring phase current (a) 3 phase current waveforms (b) pulse output, Ia_pulse (c) pulse output, Ib_pulse (d) pulse output, Ic_pulse (e) XOR output (f) the sum of instan- taneous current

Figure 6 shows the XOR detection method applied to the fault case of signal conditioning circuit. When a fault occurs, the arithmetic sum has a periodically changed value but the logical sum has the changed pulse width. But the proposed method can quickly detect the fault condition due to the result of XOR.

However, this fault condition is not real the open-phase fault condition but the problem of signal processing circuit. The proposed method for the same situation can signal that the control unit must shift to another fault-tolerant operation instead of stopping the total control system.

Fig. 6. Application of proposed method to Case Ⅰ(a) 3 phase current wave- forms (b) operation flag (c) the sum of instantaneous current (d) XOR output

Fig. 7. Application of proposed method to Case Ⅱ(a) 3 phase current wave- forms (b) operation flag (c) the sum of instantaneous current (d) XOR output

Figure 7 shows the XOR detection method applied to CASE Ⅱ. When a fault occurs, the arithmetic sum has zero value but the logical sum has the changed pulse width. But the proposed method can quickly detect the fault condition due to the result of XOR. In this case, the conventional method needs more time to determine whether the phase current change due to this fault corresponds to the fault level or over current level etc.

5. 결 론

This paper proposes a method of the open-fault condition and diagnosis of control system using a traction drive module. For this, we have proposed case studies for the open-phase current fault diagnostic method as follows:

∙ Design fault, that is, signal conditioning circuit has an internal fault.

∙ CASE Ⅰ - line fault, that is, an open-phase fault is detected by S/W diagnostic algorithm.

∙ CASE Ⅱ - line fault, that is, an open-phase fault is detected by an additional H/W diagnostic circuit.

This method uses the simple signal conditioning circuit to detect three-phase currents based on optical coupling devices without a complex diagnostic software algorithm.

In order to improve the diagnosis response, logical sum for the 3-phase pulse, that is, XOR is used to detect the open-phase fault condition and as a result the response time for diagnosis can be 3 times faster than before as shown in Figure 7(e).

Through the application of the proposed method to the conventional fault conditions, we know that the proposed method can quickly detect the open-phase current fault condition.