Research on the Detection Method of Excessive Spark in Ship DC Motors Based on Wavelet Analysis

Abstract

In order to analyze and solve the problem of excessive commutation spark of DC motor in ship electric propulsion system, which leads to a decrease in output power and low torque, this paper first establishes a mathematical model of the ship DC motor, builds its simulation model based on the mathematical model, and conducts simulation verification. Secondly, the Cassie arc model is introduced to model the commutation spark, and the Cassie arc model is connected in series in the armature winding of the DC motor to achieve virtual injection of excessive spark fault of the DC motor. Finally, the Fourier transform and wavelet analysis are used to process the data of the armature winding current and excitation current of the DC motor. The simulation results show that when an arc fault occurs in the DC motor, the ripple coefficient of the armature current and excitation current will increase, and the high-frequency component will increase. DB8 is an adopted wavelet function that decomposes the armature current and excitation current six times, and calculates the energy changes before and after the fault of each decomposed signal layer. It is found that without considering the approximate components, the D4 layer wavelet energy of the armature current and excitation current has the largest proportion in the detail components. The D1, D2, and D3 layers’ wavelet decomposition signals of the armature current and excitation current have significant energy changes; that is, the energy increase in the middle and high frequency parts exceeds 20%, and the D3 layer wavelet decomposition signal has the largest energy change, exceeding 40%. This can be used as a fault characteristic quantity to determine whether the DC motor has a large spark fault. This study can provide reference and guidance for online detection technology of excessive sparks in ship DC motors.

1. Introduction

The ship’s electric propulsion system uses an electric mechanical device to drive the propeller to rotate and propel the ship. Due to its advantages of smooth operation, low noise, safety and reliability, and easy control and speed regulation, it has been widely used in the shipping industry. DC motors have good starting and speed regulation performance, can achieve frequent stepless fast starting, and can withstand frequent impact loads. Therefore, they are often used as power sources and applied in ship electric propulsion systems. According to research, the main problem limiting the further development of ship electric propulsion systems is the easy generation of sparks between DC motor brushes and commutator segments, which can lead to insufficient output power and torque, affecting the ship’s sailing speed and increasing the difficulty of maintenance. How to detect commutation sparks in ship DC motors online has become an urgent problem to be solved.

When a DC motor is running, as the armature rotates, the components that make up the armature winding will continuously leave one branch and enter adjacent branches. During this period, the components will be short-circuited by the electric brush, and the process of changing the direction of the current in the components is called the commutation process. The time taken from the beginning to the end of commutation is called the commutation cycle. Generally, the commutation cycle is very small, only a few milliseconds, but the quality of the commutation process directly affects the performance of the DC motor. When the commutation is poor, electric sparks will be generated between the brush and the commutator. The relevant technical standards for DC motors classify sparks into levels and stipulate that DC motors cannot exceed level 11/2 during operation. When the spark is small, it has little effect on the operating characteristics of the DC motor, but when the spark is too large, it can burn out the brushes or commutator, making the motor unable to work normally. In severe cases, a long arc can form between the positive and negative brushes, causing a ring fire fault, resulting in the burning of the brushes and commutator, and even the entire brush holder.

At present, research on the detection and diagnosis of large spark faults in DC motors mainly includes the following: Ref. [1] proposes an online fault diagnosis method for DC motors based on current signal analysis. By collecting and monitoring the armature current signal of DC motors in real time, Fourier analysis and other signal processing methods are used to extract fault feature quantities, achieving online diagnosis of DC motor armature faults, excitation faults and commutation faults; Ref. [2] provides an overview of DC motor ring fires based on their own work experience, analyzes the causes and potential hazards of ring fire accidents, and provides prevention and treatment measures; Ref. [3] proposes a new method for extracting fault information of DC motor bearings. By performing wavelet decomposition on the maximum values of the bus current and three-phase current of the motor, the fault characteristic quantities of the DC motor bearings are obtained. Simulation and experimental results demonstrate the feasibility of this method; Ref. [4] studied the detection method of open circuit fault in DC motor armature winding, established an armature winding open circuit model, and achieved armature winding fault location by calculating the inter-chip resistance of DC motor. Regarding the issue of excessive spark in DC motors, current research mainly focuses on the spark level of DC motors, the electromagnetic, mechanical, and chemical causes of spark generation, the potential hazards, and corresponding maintenance measures [5,6,7]. In terms of online spark detection, methods such as spark spectrum image processing, photomultiplier tube measurement of spark energy, and peak voltmeter measurement of spark voltage are usually used. However, none of these methods can continuously monitor the spark state during the operation of DC motors online [8,9,10].

This article intends to use virtual fault injection technology to simulate and analyze the large spark fault of DC motors. Compared with physical fault injection technology, virtual fault injection technology has the advantages of safety, reliability, low cost, no damage to hardware equipment, easy data collection, and multiple repeated tests. Therefore, it has a wide range of applications. This article first establishes a mathematical model of a brushed DC motor, builds its simulation model based on the mathematical model, and verifies the correctness of the DC motor model through simulation. Secondly, based on the differential equation of the Cassie arc model, a Matlab R2016a arc simulation model was established, and the correctness of the constructed arc simulation model was verified through simulation. Next, connect the Cassie arc model in series to the armature winding circuit of the DC motor to achieve virtual injection of excessive spark in the DC motor, and observe the impact of excessive spark on the operating characteristics of the DC motor. Fourier transform and wavelet analysis were used to process the armature winding current and excitation current of a DC motor. The simulation results showed that when an arc fault occurred in the DC motor, the ripple coefficients of the armature current and excitation current would increase [11,12,13], and the high-frequency components would increase. The DB8 wavelet was used to decompose the armature current and excitation current six times, and the energy changes before and after each decomposition signal fault were calculated. It was found that in the detailed components, the D4 wavelet energy of the armature current and excitation current accounted for the largest proportion, while the D1, D2, and D3 wavelet decomposition signals had significant energy changes, with the change ratio exceeding 40%. This can be used as a fault characteristic quantity to determine whether the DC motor has a large spark fault. This study can provide a new approach and reference for spark detection of parallel excited DC motors [14,15,16].

2. Materials and Methods

2.1. Mathematical Model of Parallel Excited DC Motor

The research object of this article is a parallel excited low-speed DC motor, and its circuit schematic is shown in Figure 1.

As shown in Figure 1, the armature circuit equation of the DC motor is as follows:

$$U=R_a{i}_a+L_a{\displaystyle \frac{d{i}_a}{dt}}+E$$

(1)

$$E={\mathrm{K}}_{\mathrm{e}}\Phi w$$

(2)

$$\Phi ={\mathrm{K}}_{\mathrm{f}}{I}_f$$

(3)

In the above equation, $U$ is the terminal voltage of the DC motor, $E$ is the back electromotive force, $\Phi$ is the magnetic flux, $I_f$ is the excitation current, $w$ is the rotational speed, $R_a$ is the armature resistance, $I_a$ is the armature current, $L_a$ is the armature inductance, ${\mathrm{K}}_{\mathrm{e}}$ is the electromotive force constant, and ${\mathrm{K}}_{\mathrm{f}}$ is the excitation constant [17].

The excitation circuit equation is as follows:

$$U=I_f{R}_f+L_f{\displaystyle \frac{d{I}_f}{dt}}$$

(4)

In the above formula, $R_f$ is the excitation resistance and $L_f$ is the excitation inductance [18].

The torque and motion equation are as follows:

$$T_e-T_L=\mathrm{J}{\displaystyle \frac{dw}{dt}}+\mathrm{B}w$$

(5)

$$T_e={\mathrm{K}}_{\mathrm{t}}\Phi I_a$$

(6)

In the above equation, $T_e$ is electromagnetic torque, $T_L$ is load torque, $\mathrm{J}$ is moment of inertia, $\mathrm{B}$ is damping coefficient, and ${\mathrm{K}}_{\mathrm{t}}$ is torque constant [19].

Based on the mathematical model of the DC motor mentioned above, a Matlab simulation model was built. The parameters of the DC motor were set as shown in

Table 1. Simulation parameters of DC motor.

Serial Number	Motor Parameters	Numerical Value	Unit
1	supply voltage	220	V
2	Armature winding resistance	4	$\mathrm{\Omega }$
3	Armature winding inductance	0.15	H
4	Excitation winding resistance	120	$\mathrm{\Omega }$
5	Excitation winding inductance	4.5	H
6	Moment of inertia	0.008	$\mathrm{kg}\cdot {\mathrm{m}}^2$
7	Damping coefficient	0.01	$\mathrm{N}\cdot \mathrm{m}\cdot \mathrm{s}/\mathrm{rad}$
8	Torque constant	1.8	$\mathrm{N}\cdot \mathrm{m}/\mathrm{A}$
9	Electromotive force constant	1.8	$\mathrm{V}\cdot \mathrm{s}/\mathrm{rad}$
10	Magnetic flux proportionality coefficient	0.018	$\mathrm{Wb}/\mathrm{A}$
11	load torque	2	$\mathrm{N}\cdot \mathrm{m}$

, and the simulation time was 20 s. At 6 s, the load torque was added to obtain the waveforms of the armature current, excitation current, speed, and electromagnetic torque of the DC motor, as shown in Figure 2

From Figure 2, it can be seen that under no-load conditions, the armature current, excitation current, speed, and electromagnetic torque of the DC motor first rapidly increase and then reach stable values before 6 s. At 6 s, after adding load torque, the armature current gradually increases and then reaches a new steady-state value, while the excitation current remains unchanged. The speed gradually decreases to the new steady-state value, and the electromagnetic torque gradually increases and finally stabilizes at the new steady-state value. The above simulation results are consistent with the theoretical analysis results, verifying the correctness of the parallel excited DC motor simulation model constructed [20].

2.2. Cassie Arc Model

The mathematical model of an arc is used to describe the dynamic changes of the arc voltage, current, etc. through mathematical equations. Common mathematical models that describe the macroscopic external characteristics of arc include the Cassie model, Mayr model, Stokes model, etc. Cassie, Mayr, and other mathematical simulation models describe arcs through variable resistors and calculate the equivalent impedance of arcs using nonlinear differential equations, thereby achieving modeling and simulation of faulty arcs. The Cassie model assumes that the arc voltage is constant and can accurately reflect the resistance value at zero current crossing in the arc, mainly used in small resistance and high current circuits. The Mayr model assumes that the arc dissipation power is constant and is suitable for studying the voltage–current relationship in low-current-level arc dynamic processes, especially during the extinguishing stage. The Cassie and Mayr arc models are widely used to simulate the dynamic behavior of arcs, and many other arc models have been improved based on them. The Cassie arc model is widely used in modeling the commutation spark of DC motors due to its good description of the dynamic characteristics of the arc [21].

The Cassie arc model can be expressed mathematically as follows:

$${\displaystyle \frac{1}{g}}{\displaystyle \frac{dg}{dt}}={\displaystyle \frac{1}{{\tau }_c}}({\displaystyle \frac{u^2}{U_c^2}}-1)$$

(7)

Among them, $u$ is the arc voltage, $U_c$ is the arc voltage constant, represents the voltage drop of the arc in steady state, $g$ is the arc conductivity, ${\tau }_c$ is the arc time constant; the larger, the slower the change in arc conductivity.

Based on the mathematical model of the Cassie arc, a Matlab simulation model was built and connected in series to the AC circuit. The parameter settings are shown in

Table 2. Cassie arc model parameters.

Serial Number	Cassie Arc Model Parameters	Numerical Value	Unit
1	AC power supply voltage	220	V
2	Initial value of arc conductivity	$1.17\times {10}^{-4}$	S
3	Arc voltage constant	50	V
4	load resistance	300	$\mathrm{\Omega }$
5	AC power frequency	50	Hz
6	Arc time constant	$2.25\times {10}^{-4}$	s
7	Simulation time	0.06	s

, and the simulation results are shown in Figure 3.

From Figure 3, it can be seen that the phase of the arc voltage and arc current is the same, and the arc current has a clear “flat shoulder” phenomenon. The arc voltage has a significant peak distortion. When the arc is generated, it will continuously dissipate energy outward. When the power supply voltage decreases and the arc resistance increases, the arc will gradually extinguish. When the power supply voltage increases and the arc resistance decreases, the arc will gradually burn. With the periodic change of the power supply voltage, the arc will periodically “extinguish burn”, causing the arc voltage and arc current to change periodically and produce high-frequency components. The arc voltage waveform is similar to a square wave, and the arc current waveform is similar to a sine wave. The simulation results verify the correctness of the simulation model and also provide a basis for the subsequent use of periodic square wave signals to simulate Cassie. The arc voltage laid the theoretical foundation.

Fourier analysis was conducted on the arc voltage and arc current, and the results are shown in Figure 4. It can be seen from Figure 4 that the harmonic components of the arc voltage and arc current are odd harmonics, which is consistent with the research results in Ref. [22].

2.3. Virtual Injection for the Fault of Excessive Spark in DC Motor

This article achieves virtual injection of spark bias faults by connecting the Cassie arc model in series with the armature circuit of a DC motor. The circuit schematic is shown in Figure 5:

From the above schematic, it can be seen that when the spark is too large, only the armature circuit of the DC motor changes; that is, the armature circuit equation becomes the following [23]:

$$U=E+I_a{R}_a+L_a{\displaystyle \frac{d{I}_a}{dt}}+V_{\mathrm{arc}}$$

(8)

All other equations remain unchanged, $V_{\mathrm{arc}}$ represent the arc voltage. According to the waveform of arc voltage in Section 2, a periodic square wave signal is used to simulate arc voltage and achieve virtual injection of DC motor spark bias fault. The periodic square wave signal is shown in Figure 6:

From Figure 6, it can be seen that the voltage of the square wave signal is 20 V and the frequency is 10 Hz. Simulating a DC motor to perform 10 commutations in 1 s generates 10 commutation sparks.

Based on the mathematical model and square wave signal of the DC motor, when the spark is too large, a simulation model is built. The simulation parameters are shown in

Table 3. Simulation parameters of DC motor with excessive spark.

Serial Number	Motor Parameters	Numerical Value	Unit
1	Supply voltage	220	V
2	Armature winding resistance	4	$\mathrm{\Omega }$
3	Armature winding inductance	0.15	H
4	Excitation winding resistance	120	$\mathrm{\Omega }$
5	Excitation winding inductance	4.5	H
6	Moment of inertia	0.008	$\mathrm{kg}\cdot {\mathrm{m}}^2$
7	Damping coefficient	0.01	$\mathrm{N}\cdot \mathrm{m}\cdot \mathrm{s}/\mathrm{rad}$
8	Torque constant	1.8	$\mathrm{N}\cdot \mathrm{m}/\mathrm{A}$
9	Electromotive force constant	1.8	$\mathrm{V}\cdot \mathrm{s}/\mathrm{rad}$
10	Magnetic flux proportionality coefficient	0.018	$\mathrm{Wb}/\mathrm{A}$
11	Load torque	2	$\mathrm{N}\cdot \mathrm{m}$
12	Arc voltage	20	V

. The armature current, excitation current, speed, and electromagnetic torque of the DC motor are simulated and shown in Figure 7.

From Figure 7, it can be seen that after injecting sparks for 10 s, the armature current, excitation current, speed, and electromagnetic torque of the DC motor all decrease and begin to fluctuate. It can be intuitively seen from the image that the high-frequency components of armature current, excitation current, speed, and electromagnetic torque increase [24].

Due to the significantly faster response speed of electrical signals compared to mechanical signals, this article mainly analyzes and processes the armature current and excitation current signals of DC motors. Fourier analysis was conducted on the armature current and excitation current of the DC motor under two operating conditions: normal operation and excessive spark. The analysis results are shown in Figure 8:

From Figure 8, it can be seen that during normal operation, the armature current and excitation current of the DC motor are mainly composed of the DC component, with relatively few high-frequency components. When the spark is large, the high-frequency components in the armature current and excitation current increase significantly, which can be used as a reference indicator to determine the fault of excessive spark in the DC motor.

3. Results

Wavelet Analysis of Spark Bias in DC Motor

The wavelet transform is developed based on Fourier analysis, originating from the scaling and translation of functions. A wavelet, in simple terms, is a “small segment wave”, which is a special type of wave with a finite length and an average value of zero. It has two characteristics: firstly, it is small and has a tight or approximately tight support set in the time domain; secondly, it has alternating positive and negative fluctuations, with an average component of zero. Wavelet transform decomposes a signal into a series of wavelet functions that are superimposed, and these wavelet functions are obtained by a mother wavelet function through translation and scaling transformation [25].

When the function $\psi \left(t\right)$ satisfies the following two conditions:

$$\psi \left(t\right)\in L^2\left(R\right)$$

(9)

$${\displaystyle \underset{-\infty }{\overset{+\infty }{\int }}{\left|\stackrel{⌢}{\psi }\left(\omega \right)\right|}^2}{\left|\omega \right|}^{-1}d\omega <+\infty$$

(10)

It can be called a mother wavelet, and by translating and scaling the mother wavelet, we can obtain.

$${\psi }_{a,b}\left(t\right)={\displaystyle \frac{1}{\sqrt{a}}}\psi \left({\displaystyle \frac{t-b}{a}}\right) \left(a>0, b\in R\right)$$

(11)

${\psi }_{a,b}\left(t\right)$ is called a wavelet basis function, a is called the scale factor, b is called the scale shift factor. If a and b are continuous, ${\psi }_{a,b}\left(t\right)$ is called the continuous wavelet basis functions. For any function $x\left(t\right)$, its wavelet transform is as follows:

$$W{T}_x\left(a,b\right)={\displaystyle \frac{1}{\sqrt{a}}}{\displaystyle \underset{-\infty }{\overset{+\infty }{\int }}x\left(t\right)}\overline{\psi }\left({\displaystyle \frac{t-b}{a}}\right)dt$$

(12)

$W{T}_x\left(a,b\right)$ is called the continuous wavelet transform of $x\left(t\right)$, $\overline{\psi }\left(t\right)$ is the conjugate operation of $\psi \left(t\right)$. After the wavelet transform, the function is two-dimensional; that is, the wavelet transform transforms one-dimensional signals into two-dimensional signals [26].

From the perspective of saving computational complexity, the continuous wavelet transform factors a and b are usually discretized without losing the original signal information.

$$a=a_0^0,a_0^1,a_0^2,\dots ,a_0^j,j=0,1,2\dots$$

The sampling interval of B is

$$\Delta b=a_0^j{b}_0,j=0,1,2\dots$$

At this point, ${\psi }_{a,b}\left(t\right)$ is as follows:

$${\psi }_{a,b}\left(t\right)=a_0^{{\displaystyle \frac{j}{2}}}\psi \left(a_0^{-j}t-k{a}_0^j{b}_0\right), j=0,1,2,3; k\in Z$$

(13)

Let $a_0=2, b_0=1$

$${\psi }_{j,k}\left(t\right)=a_0^{-{\displaystyle \frac{j}{2}}}\psi \left(2^{-j}t-k\right), j=0,1,2,3; k\in Z$$

(14)

Then, the discrete wavelet transform of the $x\left(t\right)$ is

$$W{T}_x\left(j,k\right)={\displaystyle \frac{1}{\sqrt{a}}}{\displaystyle \underset{-\infty }{\overset{+\infty }{\int }}x\left(t\right)}{\psi }_{j,k}^{\ast }(t)dt$$

(15)

The wavelet transform can simultaneously perform multi-scale analysis on signals in the time-frequency domain. When there is a sudden change in the signal, the coefficients after the wavelet transform have modulus maxima, so the time of fault occurrence can be determined by detecting the modulus maxima [27].

When a DC motor malfunctions, the current signal in the armature winding often contains a large number of time-varying, short-term impulses, and sudden components. Traditional signal analysis methods, such as the Fourier transform, cannot detect sudden changes in the signal and cannot effectively extract the fault characteristics of the DC motor. In the field of motor testing, non-stationary signals are often encountered, and the Fourier transform cannot obtain effective results. Wavelet transform, as a video domain analysis method, has the ability to characterize local characteristics of signals in both time and frequency domains, making it particularly suitable for processing non-stationary signals. This provides a good technical foundation for realizing online fault diagnosis systems for DC motors and accurate testing of motor parameters [28].

The response speed of electrical signals is much faster than that of mechanical signals, and they have wide applications in the field of signal detection. Therefore, this paper mainly conducts wavelet analysis on the armature current and excitation current of DC motors to determine the fault characteristic quantities when the spark of DC motors is too large. Under normal operation and when the spark is too large, wavelet decomposition is performed on the armature current and excitation current of DC motors, respectively. The db8 wavelet function is used for six decompositions, with a sampling time of 20 s and a sampling frequency of 100 Hz. A total of 2000 discrete points are collected. The wavelet decomposition results are shown in Figure 9:

4. Discussion

By comparing the wavelet analysis results of the armature current and excitation current of the DC motor during normal operation and when the spark is large, it can be seen that the approximate and detailed components of the armature current and excitation current of the DC motor during normal operation are relatively smooth, mainly dominated by the DC component. The time point of the applied load on the DC motor can be detected through the approximate signal. When the spark is too large, the approximate component of the DC motor can detect the time point when the load is applied to the DC motor. In addition, the approximate component and detail component can also detect the node where the current suddenly changes at 10 s. Due to the presence of sparks, the high-frequency components in the armature current and excitation current increase.

Calculate the energy of each wavelet decomposition layer signal of the armature current and excitation current signals separately for normal operation and high spark fault of the DC motor, compare the changes in energy of each wavelet decomposition layer signal of the armature current signal and excitation current signal under normal operating conditions and high spark fault conditions of the DC motor, and determine the fault characteristic quantity of high spark fault of the DC motor. Define the energy of each wavelet decomposition layer signal during normal operation of the DC motor as $W$, and the energy of each wavelet decomposition layer signal during high spark fault of the DC motor as $Q$. The calculation formula for the transformation ratio coefficient $\alpha$ is as follows:

$$\alpha ={\displaystyle \frac{W}{Q}}$$

(16)

The energy and energy transformation ratio coefficients of each wavelet decomposition layer signal of the armature current signal and excitation current signal under normal operating conditions and high spark fault conditions of the DC motor are shown in

Table 4. Energy and energy transformation coefficient of each wavelet decomposition layer signal of armature current.

Signal Name	$\mathit{W}$	$\mathit{Q}$	$\mathit{\alpha }$
A6	6,998,000	6,575,700	0.966
D1	34.23	41.78	1.22
D2	308.5	392.57	1.27
D3	1634.1	2320.6	1.42
D4	23,279	23,302	1
D5	3804.2	3836.6	1
D6	1151.5	1180.1	1.02

and

Table 5. Energy and energy ratio coefficient of each wavelet decomposition layer signal of excitation current.

Signal Name	$\mathit{W}$	$\mathit{Q}$	$\mathit{\alpha }$
A6	8395.4	8071.2	0.961
D1	0.038	0.046	1.21
D2	0.343	0.437	1.27
D3	1.80	2.567	1.43
D4	26.05	26.08	1
D5	4.284	4.321	1
D6	1.372	1.405	1.02

, respectively.

According to Table 4 and Table 5, it can be seen from the energy and energy transformation coefficients of the wavelet decomposition signals of the armature current and excitation current of the DC motor under normal operating conditions and spark biased operating conditions that, without considering approximate components, the energy proportion of the D4 wavelet decomposition signals of the armature current and excitation current is the largest in the detail components. The energy changes of the D1, D2, and D3 wavelet decomposition signals of the armature current and excitation current are relatively large; that is, the energy transformation coefficient of the middle and high frequency parts increases by more than 20%. Among them, the energy transformation coefficient of the D3 wavelet decomposition signal changes the most, and the energy transformation coefficient increases by more than 40%. Therefore, the energy proportion of the D4 wavelet decomposition signals of the armature current and excitation current, as well as the degree of change in the energy of the D1, D2, and D3 wavelet decomposition signals of the armature current and excitation current, can be analyzed to comprehensively determine whether the DC motor has a large spark fault.

5. Conclusions

This article is based on the Matlab simulation platform. Firstly, based on the mathematical model of the parallel excited DC motor, a simulation model was built, and the simulation results verified the correctness of the model. Secondly, a Cassie arc simulation model was established, and the arc voltage and arc current waveforms verified the correctness of the simulation model. According to the characteristics of the arc voltage waveform, a periodic square wave signal was used to simulate Cassie. The arc is connected in series to the armature winding of the DC motor to achieve virtual injection of the DC motor spark bias fault. Simulation results show that when the spark bias is large, the ripple coefficients of the DC motor armature current, excitation current, speed, and electromagnetic torque will increase. Fourier transform is performed on the DC motor armature current and excitation current under normal operation and spark bias conditions, respectively. The results show that the spark bias fault will increase the high-frequency components in the current signal. Finally, based on the db8 wavelet function, a 6-layer decomposition was performed on the armature current and excitation current of the DC motor under two operating conditions: normal operation and high spark operation. By calculating the energy and energy transformation coefficient of each decomposed signal before and after the fault, it was found that in the detail components, the D4 layer wavelet decomposition signal had the largest proportion of energy, and the D1, D2, and D3 layers wavelet decomposition signals of armature current and excitation current had significant energy changes, all exceeding 20%. The D3 layer wavelet decomposition signal had the largest energy change, exceeding 40%. This can be used as a fault characteristic to determine whether the DC motor has a high spark fault. This study can provide a new approach and method for detecting large spark faults in DC motors, realizing real-time continuous online monitoring of such faults, and providing a reference.