智慧机械制造业噪音源多目标识别分类数据

为了降低机械制造工业作业场景的噪声污染，优化人员作业的环境状态，我们通过CATBoost多目标分类算法进行不同噪音源进行检测分析，从而根据不同噪音的特性进行多目标分类，通过此操作，可以帮助我们筛选出高噪音污染性的运行设备，紧接着再做进一步的环境噪音优化，从而大大提高了从业人员在工作环境的舒适度。1.数据搜集：收集机械制造作业场景的声学信号数据，并将其作为特征变量。同时，收集对应噪音源的器件数据，并将其作为分类的目标变量。2.预处理：利用归一化公式x=（xi-min）/（max-min）；其中xi是样本字段中第i条数据，x是归一化后的值；并且利用同类均值插补法进行缺失值补充，首先用层次聚类模型预测缺失变量的类型，再以该类型的均值插补。3.数据分析：由于不同设备的声源信息不同，因此我们将他们每个频率的最大值和最小值都搜集出来，紧接着将最大值中最大的那个值作为上限，最小值中最小的那个值作为下限，因此构造出了频率值区间，将区间内的所有值平均分成[0，4000]、[4000，8000]、[8000，12000]、[12000，16000]、[16000，20000]这5类，记为A,B,C,D,E字母，其他数据字段如：声压级、持时、尖锐度、响度的值也同理进行一样的操作，紧接着根据采集到的设备信息，我们可以将它的各个信息值对标字母类别，最后再将各个转换为字母类别的信息按顺序串成一段字母(如：CBAAE对应的就是电锯Electric Saw),最后我们将该字母串导入到交互式文本框中，它就会反馈出对应的声源类别，其中交互式文本框的制作流程如下：首先根据每一个设备信息的上下限确定好可能存在的数字类别，紧接着将所有可能的字母类别进行串联，最终出现了所有属于该设备的字母串类型，最后将这些字母串导入到交互式系统中（也就是所谓的穷举法）。

To mitigate noise pollution in machinery manufacturing workplace scenarios and improve the working environment for on-site workers, we utilize the CATBoost multi-objective classification algorithm to detect and analyze various noise sources, and conduct multi-target classification based on the characteristics of each noise. This approach allows us to screen out high noise-polluting operating equipment, followed by further environmental noise optimization, thereby significantly enhancing the comfort of workers in their workplaces.
1. Data Collection: Collect acoustic signal data from machinery manufacturing workplace scenarios as feature variables. Meanwhile, collect device data corresponding to each noise source as the target variables for classification.
2. Preprocessing: Use the normalization formula $x = frac{x_i - ext{min}}{ ext{max} - ext{min}}$, where $x_i$ represents the i-th data point in the sample field and $x$ is the normalized value. Additionally, adopt the class-specific mean imputation method for missing value supplementation: first predict the type of missing variables using a hierarchical clustering model, then impute with the mean value of that type.
3. Data Analysis: Since the sound source information varies across different equipment, we first collect the maximum and minimum values for each frequency band. We then take the largest value among all maxima as the upper limit and the smallest value among all minima as the lower limit, thereby constructing a frequency value interval. This interval is evenly divided into five categories: [0, 4000], [4000, 8000], [8000, 12000], [12000, 16000], and [16000, 20000], labeled as A, B, C, D, E respectively. The same processing is applied to other data fields including sound pressure level, duration, sharpness, and loudness. Subsequently, based on the collected equipment information, we map each of its data values to the corresponding letter category. Finally, the information converted into letter categories is concatenated in sequence to form a letter string (for example, CBAAE corresponds to an Electric Saw). We then input this letter string into an interactive text box, which will output the corresponding sound source category. The development workflow of this interactive text box is as follows: first, determine all possible numerical categories based on the upper and lower limits of each piece of equipment's data; then concatenate all possible letter categories to generate all valid letter string types for that equipment; finally, import these letter strings into the interactive system, which is also referred to as the exhaustive method.

搜集汇总

数据集介绍

特点

该数据集包含501条机械制造业噪音源数据，用于多目标识别分类，通过CATBoost算法优化工作环境噪音。数据包括频率、声压级等关键声学参数，并通过特定算法进行预处理和分析，以识别不同噪音源类别。