Operation Lifetime Prediction for Industrial Cables by Deep Learning Algorithm

Flexible cables are widely used in the industrial equipment, such as crane, heavy machinery and robot. These cables, as one of the important parts, transmit power, signals and control commands to the place where manufacturing and transportation happened. Due to the heavy duty and harsh working environments, the cables’ quality and operation life are major concerns for the users and producers. In order to make sure the design and manufacturing are all in good shape when these cables were made, Manufacturers spend tremendous amount of time and money on testing the finished goods. However, the complexity of customer requirements, difficulty of cabling process and diversity of materials’ combination make the service life of flexible cables challenge to forecast.

Deep learning algorithm is a branch of machine learning technology and widely used in vision recognition, speech recognition, and natural language processing. The power of this technique is able to identify the unseen pattern in the huge and disordered data, combine these patterns to form a usable model, and use this model to forecast the future event. This deep leaning algorithm has proven itself in junk mail detection, automatic pilot car, and voice control assistant.

Lots of data from flexible cable life testing are collected and used the deep learning algorithm to generate the multilayer perceptron model. This model consists of three functional layers: one INPUT layer, some HIDDEN layers, and one OUTPUT layer. There are 11 input parameters acting as the design and manufacturing features in the input layer. Two hidden layers are chose in this model, the first and second hidden layers have 40 neurons and 30 neurons respectively for machine learning purpose. The output layer just has one neuron as the probability percentage of surviving chance after the rigorous life testing. This model can help the designer evaluate his/her design before the cable is manufactured. Even better, they can use this model to choose the better design parameters and suitable production machines. As a matter of fact, this model demonstrates that artificial intelligence can be applied to the product design and manufacturing in flexible cable business. And the more data we collect, the better design we get.

Nevertheless, the model shall be further improved, the set of parameters will be enlarged and the prediction shall become more precise and reliable. If you are interested to use this model to optimize your cable design or to cooperate to improve the deep learning algorithm by providing additional construction and/or test data, please contact CABX Cable Expert Dipl.-Ing. Jörg Bör.