Trying to build a model with using Adaline algorithm

I guessed I could find the right answer with using an output of the first array as new input to the next perceptron, so it detects a pattern of difference between target and out of first perceptron and updates weight with this pattern. But it didn't work. 

As you can see, the red line is a target. the Blue line is the output of 1st perceptron. It is quite close to the target line. But the thing is after the 1st line. The black, green, yellow(invisible) lines are far off from the target line. The equation of that i believe it'd perform is just simple. x = 2x. In the Adaline algorithm, I used SSE to cost function, Batch merge (I don't know for sure it can be said like that), Gradient Descent to minimize SSE and I didn't use quantizer(Because it is not about yes or no problem). I expected the more I repeat the way using output that is from the previous perceptron as new input to the next perceptron, the more SSE Converges on minimum value. But it wouldn't happen. 

As the beginner of Machine learning, I don't know what causes this result. Perhaps, It was a predictable result. Because the hypothesis is only in my head and never based on fundamental math or ML. I hope I will do better than this next time...