All motors have vibration and noise issues, but these are more pronounced in electric drive motors. This is partly because the requirements are more stringent, as motor vibration can easily transmit and cause secondary and coupled vibrations in components such as the reducer and chassis. On the other hand, the operating conditions of electric vehicles also impose significant constraints on the motor, which exacerbate vibration and noise problems. Examples include high magnetic flux density and concentrated winding issues.
Electric drive motors prioritize high power density and high torque density in terms of performance, while also emphasizing low cost. This has led to a gradual reduction in the air gap and a gradual increase in the air gap magnetic flux density. Ultimately, the issue of high magnetic flux density will be reflected in the motor's noise and vibration levels.
In recent years, many concentrated-winding motors have emerged in the low-speed vehicle and microcar sectors. Concentrated winding motors are characterized by near-slot pole ratios and a winding pitch of 1. This design is advantageous because the small winding ends reduce copper usage, and it caters to the needs of large-scale automated production, allowing for direct automatic winding and extremely high production efficiency. However, concentrated windings present significant vibration and noise challenges. This is because the fractional-slot structure of concentrated windings generates numerous low-order harmonics, which easily pair with the fundamental frequency to produce low-space-order electromagnetic force waves. These low-space-order electromagnetic forces can excite substantial vibration responses in the housing and stator, accompanied by noise problems.
As the demand for longer driving range increases, the trend towards lightweighting in electric vehicles is evident. On the one hand, the thickness of the motor housing and stator is being reduced; on the other hand, the mass of the reducer, axle, and frame is also decreasing. This results in a decrease in the rigidity of both the main body and the system, leading to a greater vibration response and potentially more noise from the same electromagnetic force excitation.
How to analyze vibration and noise problems?
To solve vibration and noise problems, we must first solve how to analyze the problem. There are many methods and schools of thought involved, which can generally be categorized into two approaches: top-down and bottom-up.
The bottom-up approach is the primary working method of the experience-based school. The process is similar to solving a crime: first, collect evidence at the crime scene; then analyze potential suspects; investigate these suspects; and finally, establish a logical chain of hypotheses and a chain of evidence. Once a target is identified, make targeted modifications, such as rebuilding a motor to test the hypothesis. If correct, the problem is solved; otherwise, start over.
The top-down approach is the primary working method of the academic school. They believe in theory and methodology, using modeling techniques to analyze potential fault characteristics during the design phase and promptly modifying the design if problems are found. If unexpected fault characteristics are discovered during actual testing, they will gradually approximate the problem by improving model accuracy and calculation methods.
In practical work, we should strive to be well-rounded, combining and integrating these two methods. In the long run, we should prioritize the academic approach, supplemented by the experience-based one. Below, I will focus on introducing the top-down analysis process:
Step 1: Electromagnetic force calculation. Generally, two types of electromagnetic forces need to be calculated. One type is the electromagnetic force of each unit on the stator tooth tip, used for subsequent structural analysis; this is a real-time distributed force. The other type is the electromagnetic force wave in the air gap, i.e., the wave function of the electromagnetic force, used to analyze the possible causes of the electromagnetic force. Simply put: the former type of electromagnetic force is used for model calculations, while the latter is used for reverse analysis; they originate from the same source but are expressed differently.
Step 2: Modal calculation and verification. The accuracy of modal calculation is limited by factors such as constraints, impregnation, and winding, making it difficult to obtain an exact solution in one step. Therefore, it is necessary to correct various constraint and damping settings based on the results of actual testing. The experimental method for modal analysis is generally the impact test. Modal analysis is used both to correct the model and to predict the risk of resonance.
Step 3: Vibration response analysis, calculating the transient response of the casing and stator when the electromagnetic force from Step 1 is applied to the model from Step 2. Obtain the amplitude and phase of the acceleration and velocity vibrations at the points of interest.
Step 4: Noise analysis and prediction. Using acoustic analysis software, calculate the frequency distribution and decibels of the spherical noise.
How to solve the vibration and noise problem?
Regarding this, each company has its own approach. Some focus on solving the problem through manufacturing processes, some on the design of the motor itself, some on the controller, and some even adopt a comprehensive strategy that integrates various approaches.
Addressing the issue from a technological standpoint is a passive solution. Common methods include improving the quality of the impregnation process, such as increasing the amount and depth of impregnation through secondary impregnation. Similar methods include improving the quality of core lamination, reducing the number of spring clips in the stator, improving the winding end length and binding quality, etc. The essence of these methods is to increase the stiffness and damping of the structural system, thereby reducing the amplitude of vibration.
Addressing the issue through the design of the structure itself is a fundamental solution. The main idea is to start from the source of vibration, reducing the electromagnetic force where possible, and if that's not possible, ensuring the frequency of the problematic electromagnetic force avoids the structure's natural frequency. Corresponding methods include selecting the slot-to-pole ratio, which can significantly alter the frequency distribution of the electromagnetic force; and improving air gap uniformity, which can control the electromagnetic force composition, as uneven air gaps can lead to additional harmonics and generate additional electromagnetic forces. Many other methods exist, but they will not be listed here.
When the control circuitry fails to address the issue, and the motor itself cannot be the sole problem, the controller's proactive role can be leveraged; if hardware falls short, software can compensate. A common approach is harmonic management. This can solve the problem from both positive and negative perspectives. For example, if it's clearly analyzed that the 5th armature harmonic is the main source of vibration and noise, then the control system can suppress the 5th harmonic separately, thereby reducing the noise. Conversely, if it's found that the main noise is caused by the 1st tooth harmonic, and the 5th armature harmonic can just cancel out the electromagnetic force generated by the 1st tooth harmonic, then we can actively inject the 5th armature harmonic, much like fighting fire with fire. Internationally, addressing vibration and noise problems from the control side has become a research hotspot because this method is flexible and doesn't increase costs.
Below, I will first give three examples of methods used by foreign scholars to solve problems from the perspective of the motor itself, in order to broaden our thinking and inspire us.
The first example addresses noise caused by cogging harmonics in concentrated windings. Someone modified the stator teeth as shown in the image below, transforming the curved surface into a flat plane. Parametric calculations were performed using tooth width and tooth height as variable parameters. The final optimization result was a 7% reduction in motor vibration.
The second example concerns torsional vibration and skewed poles. This developer's first motor divided the rotor into two sections, with half a slot skewed between them. This reduced torque ripple and made the vehicle start more smoothly, but the motor was found to be particularly noisy at 3000-4000Hz. The problem lay in torsional resonance. The third torsional mode frequency of the motor's stator was 3725Hz, with a shape of upward and downward tension, and the phases of the upper and lower sections were exactly opposite. The two-section skewed pole design caused the cogging torque phases of the upper and lower sections to also be opposite. This resulted in a high degree of consistency with the shape of the vibration, inevitably leading to torsional resonance at a certain speed. The solution was to improve the skewed pole design, as shown in the two new skewed pole designs below. Actual measurements showed that both significantly improved noise levels.
The third example utilizes asymmetrical rotor shaping and compounding to achieve a similar skewed pole effect. As shown in the diagram, there are two types of rotor laminations: lamination A has a smaller left pole arc than its right pole arc, while lamination B is the opposite. These two types of laminations are compounded into a single rotor using an ABA (amplified rotor-absorbent composite) method, and the results in actual testing are excellent. Not only is the noise significantly reduced by 1000-2000Hz, but the cogging torque also decreases considerably.
Here are two more methods to address noise issues through the control side:
The first method is to use dual controllers to reduce high-frequency electromagnetic noise. As shown in the figure below, dual controllers can significantly suppress PWM harmonics in the motor current, achieving a noise reduction effect.
Another, more novel perspective involves a researcher who discovered that the primary vibration mode of a motor is the fourth order, and the electromagnetic force causing this vibration is significantly influenced by the direct and quadrature axis currents. Through simulations and experiments, he found that not only are the vibrations inconsistent between the motoring and generating states, but even within the second phase limit of the motoring state (Id is negative, Iq is positive), different Id and Iq distributions result in varying vibration magnitudes. Therefore, he proposed that the design of the motor and controller coupling should prioritize optimizing the electromagnetic force, thereby optimizing motor parameters and control strategies to achieve the best noise reduction through current distribution. However, no practical product applications have been observed yet.
Why strive to be a master of all things?
In dealing with motor vibration and noise problems, we naturally fall into two schools of thought. The engineering school prefers to solve problems based on experiential observation of noise and vibration, while the academic school prefers to analyze problems using theoretical models. Each approach has its advantages and disadvantages. The shortcomings of the engineering school are obvious:
Firstly, the engineering approach relies too heavily on experience, which can only be acquired through personal experience or by the teachings of experienced professionals, making it difficult to inherit and carry forward.
Secondly, many problems have the same origin but different manifestations, or different origins but the same manifestation. The symptoms are different but the causes are the same, or the causes are the same but the symptoms are different. It is difficult to distinguish them based on experience alone.
The theoretical modeling analysis methods used by academics also have many shortcomings:
Firstly, theoretical analysis often starts from steady-state ideal conditions, such as constant speed and constant torque, while in reality, vibration and noise often occur during transient processes such as acceleration and speed regulation, making it difficult to directly apply the analysis results.
Secondly, there are many theoretical sources of vibration in a motor, and many possibilities can lead to abnormal vibration and noise. These could be caused by rotor eccentricity or improper slot-pole matching. Analyzing a process is time-consuming and labor-intensive, and it is difficult to explore all possibilities.
Problems force us to think, and difficulties foster growth. Faced with real-world demands, many predecessors displayed extraordinary talent, honing their skills to the point of mastering both approaches and becoming comprehensive experts. Next, we will introduce the problem-solving methods and strategies of a top master.
Black box and white box method
When facing the vibration and noise problem of electric vehicles, this pioneer developed a novel approach. Although the theory wasn't original, its application in electric vehicles is something we can only dream of. This method is called the black-box/white-box method.
The term "black box" refers to the black box used in "black box testing," treating the motor as an object whose internal structure is unknown, focusing only on its external characteristics without considering its internal configuration. Conversely, a "white box" test presents the motor as completely open and transparent, allowing for a clear understanding of its theoretical structure and characteristics. The black-box and white-box testing methods consist of three main steps:
The first step is white-box model learning. This involves obtaining two things: 1) the theoretical spectrum of the motor's main electromagnetic forces; and 2) the components behind these spectra, which could be first-order cogging harmonics or third-order permanent magnet harmonics.
The key to white-box learning lies in clearly analyzing which electromagnetic force waves are constituted by each order of harmonics, which electromagnetic force waves are primary, and which are secondary. Generally, force waves with small spatial order and large amplitude are the primary force waves. Then, calculate the frequency variation characteristics of these force waves with rotational speed and plot them on a waterfall graph (vertical axis is rotational speed, horizontal axis is frequency). It is a diagonal line, and the frequency increases linearly with speed.
The second step is black-box physical testing. This predecessor built a motor testing platform capable of measuring motor vibration and noise data during constant speed and acceleration. This testing platform is a towing test platform, with one motor operating as the electric motor and the other as the engine, simulating various operating conditions of a car.
Black-box testing primarily yields two things: 1. the actual noise spectrum (including a noise waterfall plot); 2. the curves showing the changes in total sound pressure and sound power with velocity.
Analysis of both black-box and white-box waterfall plots revealed a richer array of noise spectrum characteristics beyond the theoretical noise spectrum. These noise spectra are largely generated by current harmonics other than the fundamental current. This researcher decomposed the complex noise spectrum into several independent components.
Deconstructing the sources of noise requires engineers to rely on their practical experience to determine which components are fundamental noise, harmonic noise, and mechanical noise. Testing also revealed that the strongest noise level occurs at 400 rpm under no-load conditions, thus establishing 400 rpm under no-load conditions as the primary analysis condition.
The third step, the four-step model analysis, involves analyzing the electromagnetic force, vibration response, and noise distribution under no-load conditions at 400 rpm. This is done to identify the source of abnormal noise.
The general process for assessing motor vibration and noise is shown in the diagram above. The following section will explain each step in detail using a real-world case study.
1. Electromagnetic force can generally be calculated using two methods: the concentrated electromagnetic force method and the nodal electromagnetic force method, each with its own advantages and disadvantages.
The process of intensive electromagnetic force calculation is as follows: First, the air gap magnetic flux density is calculated. Then, the electromagnetic force in the air gap is calculated using a formula. By accumulating calculations over multiple time periods, the complete spatiotemporal distribution of the electromagnetic force can be obtained. The electromagnetic force is then mapped onto the structure we need to study, which could be stator teeth, rotor magnets, or the surface of each pole of the rotor. Intensive electromagnetic force calculation is generally performed manually, so the selected object is usually relatively large (e.g., one tooth as one object).
In this example, the rotor surface of each pole was selected as the research object, and the temporal distribution of the electromagnetic force on each pole surface was calculated. Then, electromagnetic force spectrum analysis can be performed on each pole to obtain the frequency distribution characteristics of the electromagnetic force. It is also possible to analyze the magnitude of the electromagnetic force at a specific moment for each pole. We observe that the magnitude of the electromagnetic force at each pole is different. By performing a unified analysis of the electromagnetic forces of all poles, the spatial distribution of the electromagnetic force wave can be obtained. Having the temporal and spatial distributions of the electromagnetic force for a given object allows for a complete description of the electromagnetic force, enabling accurate application in subsequent analyses. This is the advantage of centralized force calculation. The disadvantage is the need for a significant amount of manual operation.
The calculation of distributed electromagnetic force is performed using finite element method (FE) software. This software can directly calculate the electromagnetic force within each grid cell and stores the electromagnetic force data for all step sizes. This allows for convenient acquisition of the temporal distribution of the electromagnetic force at each node, which can then be directly used as the excitation source for subsequent vibration analysis software. The advantage is that it eliminates the need for manual calculation, making it very convenient. However, this requires strong coupling between the structural analysis software and the electromagnetic analysis software. On the other hand, it makes it difficult to determine the spatial distribution of the electromagnetic force waves, hindering engineers from judging the nature and source of the electromagnetic force.
II: Modal Calculation and Correction
When performing modal calculations, it's important to note that the axial elastic modulus and Poisson's coefficient of the stator core are uncertain and need to be obtained experimentally. A senior researcher provided a set of test results, which are shown in the upper right corner of the image above. If the frequency and mode shape (spatial order) of the modal are found to be highly similar to the frequency and spatial distribution calculated from the concentrated electromagnetic force in the first step, then this electromagnetic force is extremely dangerous.
The number of orders obtainable through modal testing depends on the arrangement of sensor points; the more points, the more orders. Modal testing can obtain not only the natural frequency but also the damping coefficient.
III: Vibration Analysis
After applying electromagnetic force, this senior engineer obtained vibration data from the casing, thin end cover, and thick end cover through vibration analysis. He found that both the thin and thick end covers had large amplitudes near their first natural frequencies. He determined that these were the main vibrating components causing the abnormal noise at 400 rpm.
IV: Noise Prediction Analysis
By comparing noise simulation calculations and noise spectrum analysis from black-box testing, it was found that both have a common feature: the 580Hz and 1440Hz accessories have strong vibration spectra, which happen to be the resonant frequencies of the end caps. This further verifies that the resonance of the front and rear end caps is the cause of the abnormal noise.
Finally, based on the waterfall plots drawn during the white-box and black-box testing phases, it was found that the 3rd and 8th order force waves coincided with the natural frequencies of the two end caps at 400 rpm. The 3rd order force wave was determined to be caused by the 3rd order tooth harmonic, which revealed the root cause of the problem.
Final Improvement Plan
To optimize noise, vibration must be optimized first. Based on the formula for calculating vibration amplitude, vibration amplitude can be suppressed by reducing the excitation force, increasing the end cap mass, and increasing damping. Ultimately, this researcher chose to increase damping, reducing the overall noise by 9.72 dB.
in conclusion
In summary, the white-box and black-box approach is actually a combination of theoretical analysis, operational testing, and model simulation.
Through white-box analysis, we gained an understanding of the motor's design level and characteristics, and had a qualitative grasp of the potential harmonics that could cause noise in the motor.
Through black-box testing, we gained a complete understanding of the noise characteristics of the motor under different operating conditions, and were able to pinpoint those particularly abnormal operating conditions.
Through model simulation, we can accurately predict the vibration response and noise characteristics of various parts of the motor, and identify which electromagnetic forces cause these abnormal vibrations. This allows us to pinpoint the optimization target and provide the optimal optimization strategy.
The techniques are easy to learn, but the inner strength is hard to cultivate. Without a thorough grasp of the theory and a solid foundation, it's all just fancy footwork. Therefore, after broadening our horizons, we still need to calm down, apply and refine our skills in our work, and truly internalize them.
